Build (Llama 4)
Browse filesThis view is limited to 50 files because it contains too many changes.
See raw diff
- build/torch25-cxx11-cu118-x86_64-linux/moe/__init__.py +2 -18
- build/torch25-cxx11-cu118-x86_64-linux/moe/{_moe_2f2wzwk42r5t2.abi3.so → _moe_21a4db0.abi3.so} +2 -2
- build/torch25-cxx11-cu118-x86_64-linux/moe/_ops.py +3 -3
- build/torch25-cxx11-cu118-x86_64-linux/moe/configs/E=160,N=192,device_name=NVIDIA_A800-SXM4-80GB.json +146 -0
- build/torch25-cxx11-cu118-x86_64-linux/moe/configs/E=256,N=1024,device_name=AMD_Instinct_MI325X,block_shape=[128,128].json +200 -0
- build/torch25-cxx11-cu118-x86_64-linux/moe/configs/E=256,N=1024,device_name=AMD_Instinct_MI325_OAM,dtype=fp8_w8a8,block_shape=[128,128].json +200 -0
- build/torch25-cxx11-cu118-x86_64-linux/moe/configs/E=256,N=128,device_name=NVIDIA_A100-SXM4-80GB,dtype=int8_w8a8,block_shape=[128,128].json +146 -0
- build/torch25-cxx11-cu118-x86_64-linux/moe/configs/E=256,N=128,device_name=NVIDIA_A100-SXM4-80GB,dtype=int8_w8a8.json +146 -0
- build/torch25-cxx11-cu118-x86_64-linux/moe/configs/E=256,N=128,device_name=NVIDIA_A800-SXM4-80GB,dtype=int8_w8a8,block_shape=[128,128].json +146 -0
- build/torch25-cxx11-cu118-x86_64-linux/moe/configs/E=256,N=128,device_name=NVIDIA_A800-SXM4-80GB,dtype=int8_w8a8.json +146 -0
- build/torch25-cxx11-cu118-x86_64-linux/moe/configs/E=256,N=128,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[128,128].json +146 -0
- build/torch25-cxx11-cu118-x86_64-linux/moe/configs/E=256,N=128,device_name=NVIDIA_L20Y,dtype=fp8_w8a8,block_shape=[128,128].json +146 -0
- build/torch25-cxx11-cu118-x86_64-linux/moe/configs/E=256,N=256,device_name=AMD_Instinct_MI300X,dtype=fp8_w8a8,block_shape=[128,128].json +164 -0
- build/torch25-cxx11-cu118-x86_64-linux/moe/configs/E=256,N=256,device_name=AMD_Instinct_MI325X,dtype=fp8_w8a8,block_shape=[128,128].json +200 -0
- build/torch25-cxx11-cu118-x86_64-linux/moe/configs/E=256,N=256,device_name=AMD_Instinct_MI325_OAM,dtype=fp8_w8a8,block_shape=[128,128].json +200 -0
- build/torch25-cxx11-cu118-x86_64-linux/moe/configs/E=256,N=256,device_name=NVIDIA_B200,dtype=fp8_w8a8,block_shape=[128,128].json +146 -0
- build/torch25-cxx11-cu118-x86_64-linux/moe/configs/E=256,N=256,device_name=NVIDIA_H20,dtype=int8_w8a8,block_shape=[128,128].json +146 -0
- build/torch25-cxx11-cu118-x86_64-linux/moe/configs/E=256,N=256,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[128,128].json +146 -0
- build/torch25-cxx11-cu118-x86_64-linux/moe/configs/E=256,N=256,device_name=NVIDIA_L20,dtype=fp8_w8a8,block_shape=[128,128].json +146 -0
- build/torch25-cxx11-cu118-x86_64-linux/moe/configs/E=256,N=512,device_name=AMD_Instinct_MI325_OAM,dtype=fp8_w8a8,block_shape=[128,128].json +200 -0
- build/torch25-cxx11-cu118-x86_64-linux/moe/configs/E=256,N=64,device_name=NVIDIA_A800-SXM4-80GB.json +146 -0
- build/torch25-cxx11-cu118-x86_64-linux/moe/configs/E=60,N=1408,device_name=AMD_Instinct_MI300X.json +200 -0
- build/torch25-cxx11-cu118-x86_64-linux/moe/configs/E=60,N=176,device_name=AMD_Instinct_MI300X.json +200 -0
- build/torch25-cxx11-cu118-x86_64-linux/moe/configs/E=60,N=352,device_name=AMD_Instinct_MI300X.json +200 -0
- build/torch25-cxx11-cu118-x86_64-linux/moe/configs/E=60,N=704,device_name=AMD_Instinct_MI300X.json +200 -0
- build/torch25-cxx11-cu118-x86_64-linux/moe/configs/E=64,N=1280,device_name=NVIDIA_A800-SXM4-80GB.json +146 -0
- build/torch25-cxx11-cu118-x86_64-linux/moe/configs/E=64,N=1280,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8.json +146 -0
- build/torch25-cxx11-cu118-x86_64-linux/moe/configs/E=64,N=1280,device_name=NVIDIA_H200,dtype=fp8_w8a8.json +146 -0
- build/torch25-cxx11-cu118-x86_64-linux/moe/configs/E=64,N=1280,device_name=NVIDIA_H200.json +146 -0
- build/torch25-cxx11-cu118-x86_64-linux/moe/configs/E=64,N=2560,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8.json +146 -0
- build/torch25-cxx11-cu118-x86_64-linux/moe/configs/E=64,N=2560,device_name=NVIDIA_H200,dtype=fp8_w8a8.json +146 -0
- build/torch25-cxx11-cu118-x86_64-linux/moe/configs/E=64,N=2560,device_name=NVIDIA_H200.json +146 -0
- build/torch25-cxx11-cu118-x86_64-linux/moe/configs/E=64,N=320,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8.json +146 -0
- build/torch25-cxx11-cu118-x86_64-linux/moe/configs/E=64,N=320,device_name=NVIDIA_H100_80GB_HBM3.json +146 -0
- build/torch25-cxx11-cu118-x86_64-linux/moe/configs/E=64,N=320,device_name=NVIDIA_H200,dtype=fp8_w8a8.json +146 -0
- build/torch25-cxx11-cu118-x86_64-linux/moe/configs/E=64,N=320,device_name=NVIDIA_H200.json +146 -0
- build/torch25-cxx11-cu118-x86_64-linux/moe/configs/E=64,N=640,device_name=NVIDIA_A800-SXM4-80GB.json +146 -0
- build/torch25-cxx11-cu118-x86_64-linux/moe/configs/E=64,N=640,device_name=NVIDIA_GeForce_RTX_4090,dtype=fp8_w8a8.json +146 -0
- build/torch25-cxx11-cu118-x86_64-linux/moe/configs/E=64,N=640,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8.json +146 -0
- build/torch25-cxx11-cu118-x86_64-linux/moe/configs/E=64,N=640,device_name=NVIDIA_H200,dtype=fp8_w8a8.json +146 -0
- build/torch25-cxx11-cu118-x86_64-linux/moe/configs/E=64,N=640,device_name=NVIDIA_H200.json +146 -0
- build/torch25-cxx11-cu118-x86_64-linux/moe/configs/E=8,N=14336,device_name=AMD_Instinct_MI300X,dtype=fp8_w8a8.json +164 -0
- build/torch25-cxx11-cu118-x86_64-linux/moe/configs/E=8,N=14336,device_name=AMD_Instinct_MI300X.json +52 -52
- build/torch25-cxx11-cu118-x86_64-linux/moe/configs/E=8,N=14336,device_name=AMD_Instinct_MI325X,dtype=fp8_w8a8.json +164 -0
- build/torch25-cxx11-cu118-x86_64-linux/moe/configs/E=8,N=14336,device_name=AMD_Instinct_MI325X.json +200 -0
- build/torch25-cxx11-cu118-x86_64-linux/moe/configs/E=8,N=14336,device_name=NVIDIA_H200,dtype=fp8_w8a8.json +146 -0
- build/torch25-cxx11-cu118-x86_64-linux/moe/configs/E=8,N=14336,device_name=NVIDIA_H200.json +146 -0
- build/torch25-cxx11-cu118-x86_64-linux/moe/configs/E=8,N=16384,device_name=AMD_Instinct_MI300X,dtype=fp8_w8a8.json +164 -0
- build/torch25-cxx11-cu118-x86_64-linux/moe/configs/E=8,N=16384,device_name=AMD_Instinct_MI300X.json +200 -0
- build/torch25-cxx11-cu118-x86_64-linux/moe/configs/E=8,N=16384,device_name=AMD_Instinct_MI325X,dtype=fp8_w8a8.json +164 -0
build/torch25-cxx11-cu118-x86_64-linux/moe/__init__.py
CHANGED
@@ -1,5 +1,6 @@
|
|
1 |
import torch
|
2 |
|
|
|
3 |
from ._ops import ops
|
4 |
from .fp8_utils import per_token_group_quant_fp8, w8a8_block_fp8_matmul
|
5 |
from .fused_marlin_moe import fused_marlin_moe
|
@@ -51,24 +52,6 @@ def moe_sum(input: torch.Tensor, output: torch.Tensor):
|
|
51 |
ops.moe_sum(input, output)
|
52 |
|
53 |
|
54 |
-
def moe_align_block_size(
|
55 |
-
topk_ids: torch.Tensor,
|
56 |
-
num_experts: int,
|
57 |
-
block_size: int,
|
58 |
-
sorted_token_ids: torch.Tensor,
|
59 |
-
experts_ids: torch.Tensor,
|
60 |
-
num_tokens_post_pad: torch.Tensor,
|
61 |
-
) -> None:
|
62 |
-
ops.moe_align_block_size(
|
63 |
-
topk_ids,
|
64 |
-
num_experts,
|
65 |
-
block_size,
|
66 |
-
sorted_token_ids,
|
67 |
-
experts_ids,
|
68 |
-
num_tokens_post_pad,
|
69 |
-
)
|
70 |
-
|
71 |
-
|
72 |
def topk_softmax(
|
73 |
topk_weights: torch.Tensor,
|
74 |
topk_ids: torch.Tensor,
|
@@ -87,6 +70,7 @@ __all__ = [
|
|
87 |
"fused_topk",
|
88 |
"gptq_marlin_moe_repack",
|
89 |
"grouped_topk",
|
|
|
90 |
"moe_align_block_size",
|
91 |
"moe_sum",
|
92 |
"per_token_group_quant_fp8",
|
|
|
1 |
import torch
|
2 |
|
3 |
+
from . import layers
|
4 |
from ._ops import ops
|
5 |
from .fp8_utils import per_token_group_quant_fp8, w8a8_block_fp8_matmul
|
6 |
from .fused_marlin_moe import fused_marlin_moe
|
|
|
52 |
ops.moe_sum(input, output)
|
53 |
|
54 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
55 |
def topk_softmax(
|
56 |
topk_weights: torch.Tensor,
|
57 |
topk_ids: torch.Tensor,
|
|
|
70 |
"fused_topk",
|
71 |
"gptq_marlin_moe_repack",
|
72 |
"grouped_topk",
|
73 |
+
"layers",
|
74 |
"moe_align_block_size",
|
75 |
"moe_sum",
|
76 |
"per_token_group_quant_fp8",
|
build/torch25-cxx11-cu118-x86_64-linux/moe/{_moe_2f2wzwk42r5t2.abi3.so → _moe_21a4db0.abi3.so}
RENAMED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
-
size
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:03298359c23e496df84a18978298d3372423f7733fb8185b2f6a535d25a64a7e
|
3 |
+
size 87060424
|
build/torch25-cxx11-cu118-x86_64-linux/moe/_ops.py
CHANGED
@@ -1,9 +1,9 @@
|
|
1 |
import torch
|
2 |
-
from . import
|
3 |
-
ops = torch.ops.
|
4 |
|
5 |
def add_op_namespace_prefix(op_name: str):
|
6 |
"""
|
7 |
Prefix op by namespace.
|
8 |
"""
|
9 |
-
return f"
|
|
|
1 |
import torch
|
2 |
+
from . import _moe_21a4db0
|
3 |
+
ops = torch.ops._moe_21a4db0
|
4 |
|
5 |
def add_op_namespace_prefix(op_name: str):
|
6 |
"""
|
7 |
Prefix op by namespace.
|
8 |
"""
|
9 |
+
return f"_moe_21a4db0::{op_name}"
|
build/torch25-cxx11-cu118-x86_64-linux/moe/configs/E=160,N=192,device_name=NVIDIA_A800-SXM4-80GB.json
ADDED
@@ -0,0 +1,146 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"1": {
|
3 |
+
"BLOCK_SIZE_M": 16,
|
4 |
+
"BLOCK_SIZE_N": 64,
|
5 |
+
"BLOCK_SIZE_K": 64,
|
6 |
+
"GROUP_SIZE_M": 64,
|
7 |
+
"num_warps": 4,
|
8 |
+
"num_stages": 4
|
9 |
+
},
|
10 |
+
"2": {
|
11 |
+
"BLOCK_SIZE_M": 16,
|
12 |
+
"BLOCK_SIZE_N": 64,
|
13 |
+
"BLOCK_SIZE_K": 64,
|
14 |
+
"GROUP_SIZE_M": 64,
|
15 |
+
"num_warps": 4,
|
16 |
+
"num_stages": 5
|
17 |
+
},
|
18 |
+
"4": {
|
19 |
+
"BLOCK_SIZE_M": 16,
|
20 |
+
"BLOCK_SIZE_N": 64,
|
21 |
+
"BLOCK_SIZE_K": 64,
|
22 |
+
"GROUP_SIZE_M": 64,
|
23 |
+
"num_warps": 4,
|
24 |
+
"num_stages": 5
|
25 |
+
},
|
26 |
+
"8": {
|
27 |
+
"BLOCK_SIZE_M": 16,
|
28 |
+
"BLOCK_SIZE_N": 128,
|
29 |
+
"BLOCK_SIZE_K": 128,
|
30 |
+
"GROUP_SIZE_M": 16,
|
31 |
+
"num_warps": 4,
|
32 |
+
"num_stages": 2
|
33 |
+
},
|
34 |
+
"16": {
|
35 |
+
"BLOCK_SIZE_M": 16,
|
36 |
+
"BLOCK_SIZE_N": 128,
|
37 |
+
"BLOCK_SIZE_K": 64,
|
38 |
+
"GROUP_SIZE_M": 1,
|
39 |
+
"num_warps": 4,
|
40 |
+
"num_stages": 5
|
41 |
+
},
|
42 |
+
"24": {
|
43 |
+
"BLOCK_SIZE_M": 16,
|
44 |
+
"BLOCK_SIZE_N": 64,
|
45 |
+
"BLOCK_SIZE_K": 128,
|
46 |
+
"GROUP_SIZE_M": 16,
|
47 |
+
"num_warps": 4,
|
48 |
+
"num_stages": 2
|
49 |
+
},
|
50 |
+
"32": {
|
51 |
+
"BLOCK_SIZE_M": 16,
|
52 |
+
"BLOCK_SIZE_N": 64,
|
53 |
+
"BLOCK_SIZE_K": 256,
|
54 |
+
"GROUP_SIZE_M": 16,
|
55 |
+
"num_warps": 4,
|
56 |
+
"num_stages": 2
|
57 |
+
},
|
58 |
+
"48": {
|
59 |
+
"BLOCK_SIZE_M": 16,
|
60 |
+
"BLOCK_SIZE_N": 128,
|
61 |
+
"BLOCK_SIZE_K": 128,
|
62 |
+
"GROUP_SIZE_M": 32,
|
63 |
+
"num_warps": 4,
|
64 |
+
"num_stages": 2
|
65 |
+
},
|
66 |
+
"64": {
|
67 |
+
"BLOCK_SIZE_M": 16,
|
68 |
+
"BLOCK_SIZE_N": 128,
|
69 |
+
"BLOCK_SIZE_K": 128,
|
70 |
+
"GROUP_SIZE_M": 16,
|
71 |
+
"num_warps": 4,
|
72 |
+
"num_stages": 3
|
73 |
+
},
|
74 |
+
"96": {
|
75 |
+
"BLOCK_SIZE_M": 16,
|
76 |
+
"BLOCK_SIZE_N": 128,
|
77 |
+
"BLOCK_SIZE_K": 128,
|
78 |
+
"GROUP_SIZE_M": 16,
|
79 |
+
"num_warps": 4,
|
80 |
+
"num_stages": 3
|
81 |
+
},
|
82 |
+
"128": {
|
83 |
+
"BLOCK_SIZE_M": 16,
|
84 |
+
"BLOCK_SIZE_N": 128,
|
85 |
+
"BLOCK_SIZE_K": 128,
|
86 |
+
"GROUP_SIZE_M": 16,
|
87 |
+
"num_warps": 4,
|
88 |
+
"num_stages": 3
|
89 |
+
},
|
90 |
+
"256": {
|
91 |
+
"BLOCK_SIZE_M": 16,
|
92 |
+
"BLOCK_SIZE_N": 128,
|
93 |
+
"BLOCK_SIZE_K": 128,
|
94 |
+
"GROUP_SIZE_M": 16,
|
95 |
+
"num_warps": 4,
|
96 |
+
"num_stages": 3
|
97 |
+
},
|
98 |
+
"512": {
|
99 |
+
"BLOCK_SIZE_M": 32,
|
100 |
+
"BLOCK_SIZE_N": 128,
|
101 |
+
"BLOCK_SIZE_K": 128,
|
102 |
+
"GROUP_SIZE_M": 16,
|
103 |
+
"num_warps": 4,
|
104 |
+
"num_stages": 2
|
105 |
+
},
|
106 |
+
"1024": {
|
107 |
+
"BLOCK_SIZE_M": 64,
|
108 |
+
"BLOCK_SIZE_N": 128,
|
109 |
+
"BLOCK_SIZE_K": 64,
|
110 |
+
"GROUP_SIZE_M": 16,
|
111 |
+
"num_warps": 4,
|
112 |
+
"num_stages": 4
|
113 |
+
},
|
114 |
+
"1536": {
|
115 |
+
"BLOCK_SIZE_M": 64,
|
116 |
+
"BLOCK_SIZE_N": 128,
|
117 |
+
"BLOCK_SIZE_K": 64,
|
118 |
+
"GROUP_SIZE_M": 16,
|
119 |
+
"num_warps": 8,
|
120 |
+
"num_stages": 3
|
121 |
+
},
|
122 |
+
"2048": {
|
123 |
+
"BLOCK_SIZE_M": 128,
|
124 |
+
"BLOCK_SIZE_N": 128,
|
125 |
+
"BLOCK_SIZE_K": 64,
|
126 |
+
"GROUP_SIZE_M": 1,
|
127 |
+
"num_warps": 4,
|
128 |
+
"num_stages": 3
|
129 |
+
},
|
130 |
+
"3072": {
|
131 |
+
"BLOCK_SIZE_M": 128,
|
132 |
+
"BLOCK_SIZE_N": 128,
|
133 |
+
"BLOCK_SIZE_K": 64,
|
134 |
+
"GROUP_SIZE_M": 1,
|
135 |
+
"num_warps": 4,
|
136 |
+
"num_stages": 3
|
137 |
+
},
|
138 |
+
"4096": {
|
139 |
+
"BLOCK_SIZE_M": 64,
|
140 |
+
"BLOCK_SIZE_N": 128,
|
141 |
+
"BLOCK_SIZE_K": 64,
|
142 |
+
"GROUP_SIZE_M": 16,
|
143 |
+
"num_warps": 4,
|
144 |
+
"num_stages": 4
|
145 |
+
}
|
146 |
+
}
|
build/torch25-cxx11-cu118-x86_64-linux/moe/configs/E=256,N=1024,device_name=AMD_Instinct_MI325X,block_shape=[128,128].json
ADDED
@@ -0,0 +1,200 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"1": {
|
3 |
+
"BLOCK_SIZE_M": 16,
|
4 |
+
"BLOCK_SIZE_N": 64,
|
5 |
+
"BLOCK_SIZE_K": 256,
|
6 |
+
"GROUP_SIZE_M": 1,
|
7 |
+
"num_warps": 4,
|
8 |
+
"num_stages": 2,
|
9 |
+
"waves_per_eu": 0,
|
10 |
+
"matrix_instr_nonkdim": 16,
|
11 |
+
"kpack": 2
|
12 |
+
},
|
13 |
+
"2": {
|
14 |
+
"BLOCK_SIZE_M": 16,
|
15 |
+
"BLOCK_SIZE_N": 128,
|
16 |
+
"BLOCK_SIZE_K": 64,
|
17 |
+
"GROUP_SIZE_M": 1,
|
18 |
+
"num_warps": 8,
|
19 |
+
"num_stages": 2,
|
20 |
+
"waves_per_eu": 0,
|
21 |
+
"matrix_instr_nonkdim": 16,
|
22 |
+
"kpack": 1
|
23 |
+
},
|
24 |
+
"4": {
|
25 |
+
"BLOCK_SIZE_M": 16,
|
26 |
+
"BLOCK_SIZE_N": 128,
|
27 |
+
"BLOCK_SIZE_K": 128,
|
28 |
+
"GROUP_SIZE_M": 1,
|
29 |
+
"num_warps": 8,
|
30 |
+
"num_stages": 2,
|
31 |
+
"waves_per_eu": 0,
|
32 |
+
"matrix_instr_nonkdim": 16,
|
33 |
+
"kpack": 2
|
34 |
+
},
|
35 |
+
"8": {
|
36 |
+
"BLOCK_SIZE_M": 16,
|
37 |
+
"BLOCK_SIZE_N": 64,
|
38 |
+
"BLOCK_SIZE_K": 128,
|
39 |
+
"GROUP_SIZE_M": 1,
|
40 |
+
"num_warps": 4,
|
41 |
+
"num_stages": 2,
|
42 |
+
"waves_per_eu": 0,
|
43 |
+
"matrix_instr_nonkdim": 16,
|
44 |
+
"kpack": 2
|
45 |
+
},
|
46 |
+
"16": {
|
47 |
+
"BLOCK_SIZE_M": 16,
|
48 |
+
"BLOCK_SIZE_N": 64,
|
49 |
+
"BLOCK_SIZE_K": 128,
|
50 |
+
"GROUP_SIZE_M": 1,
|
51 |
+
"num_warps": 4,
|
52 |
+
"num_stages": 2,
|
53 |
+
"waves_per_eu": 0,
|
54 |
+
"matrix_instr_nonkdim": 16,
|
55 |
+
"kpack": 2
|
56 |
+
},
|
57 |
+
"24": {
|
58 |
+
"BLOCK_SIZE_M": 16,
|
59 |
+
"BLOCK_SIZE_N": 64,
|
60 |
+
"BLOCK_SIZE_K": 128,
|
61 |
+
"GROUP_SIZE_M": 1,
|
62 |
+
"num_warps": 4,
|
63 |
+
"num_stages": 2,
|
64 |
+
"waves_per_eu": 0,
|
65 |
+
"matrix_instr_nonkdim": 16,
|
66 |
+
"kpack": 2
|
67 |
+
},
|
68 |
+
"32": {
|
69 |
+
"BLOCK_SIZE_M": 16,
|
70 |
+
"BLOCK_SIZE_N": 64,
|
71 |
+
"BLOCK_SIZE_K": 256,
|
72 |
+
"GROUP_SIZE_M": 16,
|
73 |
+
"num_warps": 4,
|
74 |
+
"num_stages": 2,
|
75 |
+
"waves_per_eu": 0,
|
76 |
+
"matrix_instr_nonkdim": 16,
|
77 |
+
"kpack": 2
|
78 |
+
},
|
79 |
+
"48": {
|
80 |
+
"BLOCK_SIZE_M": 16,
|
81 |
+
"BLOCK_SIZE_N": 64,
|
82 |
+
"BLOCK_SIZE_K": 256,
|
83 |
+
"GROUP_SIZE_M": 16,
|
84 |
+
"num_warps": 4,
|
85 |
+
"num_stages": 2,
|
86 |
+
"waves_per_eu": 0,
|
87 |
+
"matrix_instr_nonkdim": 16,
|
88 |
+
"kpack": 2
|
89 |
+
},
|
90 |
+
"64": {
|
91 |
+
"BLOCK_SIZE_M": 16,
|
92 |
+
"BLOCK_SIZE_N": 128,
|
93 |
+
"BLOCK_SIZE_K": 128,
|
94 |
+
"GROUP_SIZE_M": 32,
|
95 |
+
"num_warps": 8,
|
96 |
+
"num_stages": 2,
|
97 |
+
"waves_per_eu": 0,
|
98 |
+
"matrix_instr_nonkdim": 16,
|
99 |
+
"kpack": 1
|
100 |
+
},
|
101 |
+
"96": {
|
102 |
+
"BLOCK_SIZE_M": 16,
|
103 |
+
"BLOCK_SIZE_N": 64,
|
104 |
+
"BLOCK_SIZE_K": 128,
|
105 |
+
"GROUP_SIZE_M": 32,
|
106 |
+
"num_warps": 1,
|
107 |
+
"num_stages": 2,
|
108 |
+
"waves_per_eu": 0,
|
109 |
+
"matrix_instr_nonkdim": 16,
|
110 |
+
"kpack": 2
|
111 |
+
},
|
112 |
+
"128": {
|
113 |
+
"BLOCK_SIZE_M": 16,
|
114 |
+
"BLOCK_SIZE_N": 64,
|
115 |
+
"BLOCK_SIZE_K": 128,
|
116 |
+
"GROUP_SIZE_M": 32,
|
117 |
+
"num_warps": 1,
|
118 |
+
"num_stages": 2,
|
119 |
+
"waves_per_eu": 0,
|
120 |
+
"matrix_instr_nonkdim": 16,
|
121 |
+
"kpack": 2
|
122 |
+
},
|
123 |
+
"256": {
|
124 |
+
"BLOCK_SIZE_M": 64,
|
125 |
+
"BLOCK_SIZE_N": 64,
|
126 |
+
"BLOCK_SIZE_K": 128,
|
127 |
+
"GROUP_SIZE_M": 32,
|
128 |
+
"num_warps": 8,
|
129 |
+
"num_stages": 2,
|
130 |
+
"waves_per_eu": 0,
|
131 |
+
"matrix_instr_nonkdim": 16,
|
132 |
+
"kpack": 1
|
133 |
+
},
|
134 |
+
"512": {
|
135 |
+
"BLOCK_SIZE_M": 64,
|
136 |
+
"BLOCK_SIZE_N": 64,
|
137 |
+
"BLOCK_SIZE_K": 128,
|
138 |
+
"GROUP_SIZE_M": 8,
|
139 |
+
"num_warps": 8,
|
140 |
+
"num_stages": 2,
|
141 |
+
"waves_per_eu": 0,
|
142 |
+
"matrix_instr_nonkdim": 16,
|
143 |
+
"kpack": 1
|
144 |
+
},
|
145 |
+
"1024": {
|
146 |
+
"BLOCK_SIZE_M": 64,
|
147 |
+
"BLOCK_SIZE_N": 64,
|
148 |
+
"BLOCK_SIZE_K": 128,
|
149 |
+
"GROUP_SIZE_M": 8,
|
150 |
+
"num_warps": 8,
|
151 |
+
"num_stages": 2,
|
152 |
+
"waves_per_eu": 0,
|
153 |
+
"matrix_instr_nonkdim": 16,
|
154 |
+
"kpack": 2
|
155 |
+
},
|
156 |
+
"1536": {
|
157 |
+
"BLOCK_SIZE_M": 64,
|
158 |
+
"BLOCK_SIZE_N": 128,
|
159 |
+
"BLOCK_SIZE_K": 128,
|
160 |
+
"GROUP_SIZE_M": 8,
|
161 |
+
"num_warps": 8,
|
162 |
+
"num_stages": 2,
|
163 |
+
"waves_per_eu": 0,
|
164 |
+
"matrix_instr_nonkdim": 16,
|
165 |
+
"kpack": 2
|
166 |
+
},
|
167 |
+
"2048": {
|
168 |
+
"BLOCK_SIZE_M": 128,
|
169 |
+
"BLOCK_SIZE_N": 128,
|
170 |
+
"BLOCK_SIZE_K": 64,
|
171 |
+
"GROUP_SIZE_M": 16,
|
172 |
+
"num_warps": 8,
|
173 |
+
"num_stages": 2,
|
174 |
+
"waves_per_eu": 0,
|
175 |
+
"matrix_instr_nonkdim": 16,
|
176 |
+
"kpack": 2
|
177 |
+
},
|
178 |
+
"3072": {
|
179 |
+
"BLOCK_SIZE_M": 128,
|
180 |
+
"BLOCK_SIZE_N": 128,
|
181 |
+
"BLOCK_SIZE_K": 64,
|
182 |
+
"GROUP_SIZE_M": 8,
|
183 |
+
"num_warps": 8,
|
184 |
+
"num_stages": 2,
|
185 |
+
"waves_per_eu": 0,
|
186 |
+
"matrix_instr_nonkdim": 16,
|
187 |
+
"kpack": 2
|
188 |
+
},
|
189 |
+
"4096": {
|
190 |
+
"BLOCK_SIZE_M": 128,
|
191 |
+
"BLOCK_SIZE_N": 128,
|
192 |
+
"BLOCK_SIZE_K": 64,
|
193 |
+
"GROUP_SIZE_M": 8,
|
194 |
+
"num_warps": 8,
|
195 |
+
"num_stages": 2,
|
196 |
+
"waves_per_eu": 0,
|
197 |
+
"matrix_instr_nonkdim": 16,
|
198 |
+
"kpack": 2
|
199 |
+
}
|
200 |
+
}
|
build/torch25-cxx11-cu118-x86_64-linux/moe/configs/E=256,N=1024,device_name=AMD_Instinct_MI325_OAM,dtype=fp8_w8a8,block_shape=[128,128].json
ADDED
@@ -0,0 +1,200 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"1": {
|
3 |
+
"BLOCK_SIZE_M": 16,
|
4 |
+
"BLOCK_SIZE_N": 64,
|
5 |
+
"BLOCK_SIZE_K": 256,
|
6 |
+
"GROUP_SIZE_M": 1,
|
7 |
+
"num_warps": 4,
|
8 |
+
"num_stages": 2,
|
9 |
+
"waves_per_eu": 0,
|
10 |
+
"matrix_instr_nonkdim": 16,
|
11 |
+
"kpack": 2
|
12 |
+
},
|
13 |
+
"2": {
|
14 |
+
"BLOCK_SIZE_M": 16,
|
15 |
+
"BLOCK_SIZE_N": 128,
|
16 |
+
"BLOCK_SIZE_K": 64,
|
17 |
+
"GROUP_SIZE_M": 1,
|
18 |
+
"num_warps": 8,
|
19 |
+
"num_stages": 2,
|
20 |
+
"waves_per_eu": 0,
|
21 |
+
"matrix_instr_nonkdim": 16,
|
22 |
+
"kpack": 1
|
23 |
+
},
|
24 |
+
"4": {
|
25 |
+
"BLOCK_SIZE_M": 16,
|
26 |
+
"BLOCK_SIZE_N": 128,
|
27 |
+
"BLOCK_SIZE_K": 128,
|
28 |
+
"GROUP_SIZE_M": 1,
|
29 |
+
"num_warps": 8,
|
30 |
+
"num_stages": 2,
|
31 |
+
"waves_per_eu": 0,
|
32 |
+
"matrix_instr_nonkdim": 16,
|
33 |
+
"kpack": 2
|
34 |
+
},
|
35 |
+
"8": {
|
36 |
+
"BLOCK_SIZE_M": 16,
|
37 |
+
"BLOCK_SIZE_N": 64,
|
38 |
+
"BLOCK_SIZE_K": 128,
|
39 |
+
"GROUP_SIZE_M": 1,
|
40 |
+
"num_warps": 4,
|
41 |
+
"num_stages": 2,
|
42 |
+
"waves_per_eu": 0,
|
43 |
+
"matrix_instr_nonkdim": 16,
|
44 |
+
"kpack": 2
|
45 |
+
},
|
46 |
+
"16": {
|
47 |
+
"BLOCK_SIZE_M": 16,
|
48 |
+
"BLOCK_SIZE_N": 64,
|
49 |
+
"BLOCK_SIZE_K": 128,
|
50 |
+
"GROUP_SIZE_M": 1,
|
51 |
+
"num_warps": 4,
|
52 |
+
"num_stages": 2,
|
53 |
+
"waves_per_eu": 0,
|
54 |
+
"matrix_instr_nonkdim": 16,
|
55 |
+
"kpack": 2
|
56 |
+
},
|
57 |
+
"24": {
|
58 |
+
"BLOCK_SIZE_M": 16,
|
59 |
+
"BLOCK_SIZE_N": 64,
|
60 |
+
"BLOCK_SIZE_K": 128,
|
61 |
+
"GROUP_SIZE_M": 1,
|
62 |
+
"num_warps": 4,
|
63 |
+
"num_stages": 2,
|
64 |
+
"waves_per_eu": 0,
|
65 |
+
"matrix_instr_nonkdim": 16,
|
66 |
+
"kpack": 2
|
67 |
+
},
|
68 |
+
"32": {
|
69 |
+
"BLOCK_SIZE_M": 16,
|
70 |
+
"BLOCK_SIZE_N": 64,
|
71 |
+
"BLOCK_SIZE_K": 256,
|
72 |
+
"GROUP_SIZE_M": 16,
|
73 |
+
"num_warps": 4,
|
74 |
+
"num_stages": 2,
|
75 |
+
"waves_per_eu": 0,
|
76 |
+
"matrix_instr_nonkdim": 16,
|
77 |
+
"kpack": 2
|
78 |
+
},
|
79 |
+
"48": {
|
80 |
+
"BLOCK_SIZE_M": 16,
|
81 |
+
"BLOCK_SIZE_N": 64,
|
82 |
+
"BLOCK_SIZE_K": 256,
|
83 |
+
"GROUP_SIZE_M": 16,
|
84 |
+
"num_warps": 4,
|
85 |
+
"num_stages": 2,
|
86 |
+
"waves_per_eu": 0,
|
87 |
+
"matrix_instr_nonkdim": 16,
|
88 |
+
"kpack": 2
|
89 |
+
},
|
90 |
+
"64": {
|
91 |
+
"BLOCK_SIZE_M": 16,
|
92 |
+
"BLOCK_SIZE_N": 128,
|
93 |
+
"BLOCK_SIZE_K": 128,
|
94 |
+
"GROUP_SIZE_M": 32,
|
95 |
+
"num_warps": 8,
|
96 |
+
"num_stages": 2,
|
97 |
+
"waves_per_eu": 0,
|
98 |
+
"matrix_instr_nonkdim": 16,
|
99 |
+
"kpack": 1
|
100 |
+
},
|
101 |
+
"96": {
|
102 |
+
"BLOCK_SIZE_M": 16,
|
103 |
+
"BLOCK_SIZE_N": 64,
|
104 |
+
"BLOCK_SIZE_K": 128,
|
105 |
+
"GROUP_SIZE_M": 32,
|
106 |
+
"num_warps": 1,
|
107 |
+
"num_stages": 2,
|
108 |
+
"waves_per_eu": 0,
|
109 |
+
"matrix_instr_nonkdim": 16,
|
110 |
+
"kpack": 2
|
111 |
+
},
|
112 |
+
"128": {
|
113 |
+
"BLOCK_SIZE_M": 16,
|
114 |
+
"BLOCK_SIZE_N": 64,
|
115 |
+
"BLOCK_SIZE_K": 128,
|
116 |
+
"GROUP_SIZE_M": 32,
|
117 |
+
"num_warps": 1,
|
118 |
+
"num_stages": 2,
|
119 |
+
"waves_per_eu": 0,
|
120 |
+
"matrix_instr_nonkdim": 16,
|
121 |
+
"kpack": 2
|
122 |
+
},
|
123 |
+
"256": {
|
124 |
+
"BLOCK_SIZE_M": 64,
|
125 |
+
"BLOCK_SIZE_N": 64,
|
126 |
+
"BLOCK_SIZE_K": 128,
|
127 |
+
"GROUP_SIZE_M": 32,
|
128 |
+
"num_warps": 8,
|
129 |
+
"num_stages": 2,
|
130 |
+
"waves_per_eu": 0,
|
131 |
+
"matrix_instr_nonkdim": 16,
|
132 |
+
"kpack": 1
|
133 |
+
},
|
134 |
+
"512": {
|
135 |
+
"BLOCK_SIZE_M": 64,
|
136 |
+
"BLOCK_SIZE_N": 64,
|
137 |
+
"BLOCK_SIZE_K": 128,
|
138 |
+
"GROUP_SIZE_M": 8,
|
139 |
+
"num_warps": 8,
|
140 |
+
"num_stages": 2,
|
141 |
+
"waves_per_eu": 0,
|
142 |
+
"matrix_instr_nonkdim": 16,
|
143 |
+
"kpack": 1
|
144 |
+
},
|
145 |
+
"1024": {
|
146 |
+
"BLOCK_SIZE_M": 64,
|
147 |
+
"BLOCK_SIZE_N": 64,
|
148 |
+
"BLOCK_SIZE_K": 128,
|
149 |
+
"GROUP_SIZE_M": 8,
|
150 |
+
"num_warps": 8,
|
151 |
+
"num_stages": 2,
|
152 |
+
"waves_per_eu": 0,
|
153 |
+
"matrix_instr_nonkdim": 16,
|
154 |
+
"kpack": 2
|
155 |
+
},
|
156 |
+
"1536": {
|
157 |
+
"BLOCK_SIZE_M": 64,
|
158 |
+
"BLOCK_SIZE_N": 128,
|
159 |
+
"BLOCK_SIZE_K": 128,
|
160 |
+
"GROUP_SIZE_M": 8,
|
161 |
+
"num_warps": 8,
|
162 |
+
"num_stages": 2,
|
163 |
+
"waves_per_eu": 0,
|
164 |
+
"matrix_instr_nonkdim": 16,
|
165 |
+
"kpack": 2
|
166 |
+
},
|
167 |
+
"2048": {
|
168 |
+
"BLOCK_SIZE_M": 128,
|
169 |
+
"BLOCK_SIZE_N": 128,
|
170 |
+
"BLOCK_SIZE_K": 64,
|
171 |
+
"GROUP_SIZE_M": 16,
|
172 |
+
"num_warps": 8,
|
173 |
+
"num_stages": 2,
|
174 |
+
"waves_per_eu": 0,
|
175 |
+
"matrix_instr_nonkdim": 16,
|
176 |
+
"kpack": 2
|
177 |
+
},
|
178 |
+
"3072": {
|
179 |
+
"BLOCK_SIZE_M": 128,
|
180 |
+
"BLOCK_SIZE_N": 128,
|
181 |
+
"BLOCK_SIZE_K": 64,
|
182 |
+
"GROUP_SIZE_M": 8,
|
183 |
+
"num_warps": 8,
|
184 |
+
"num_stages": 2,
|
185 |
+
"waves_per_eu": 0,
|
186 |
+
"matrix_instr_nonkdim": 16,
|
187 |
+
"kpack": 2
|
188 |
+
},
|
189 |
+
"4096": {
|
190 |
+
"BLOCK_SIZE_M": 128,
|
191 |
+
"BLOCK_SIZE_N": 128,
|
192 |
+
"BLOCK_SIZE_K": 64,
|
193 |
+
"GROUP_SIZE_M": 8,
|
194 |
+
"num_warps": 8,
|
195 |
+
"num_stages": 2,
|
196 |
+
"waves_per_eu": 0,
|
197 |
+
"matrix_instr_nonkdim": 16,
|
198 |
+
"kpack": 2
|
199 |
+
}
|
200 |
+
}
|
build/torch25-cxx11-cu118-x86_64-linux/moe/configs/E=256,N=128,device_name=NVIDIA_A100-SXM4-80GB,dtype=int8_w8a8,block_shape=[128,128].json
ADDED
@@ -0,0 +1,146 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"1": {
|
3 |
+
"BLOCK_SIZE_M": 16,
|
4 |
+
"BLOCK_SIZE_N": 64,
|
5 |
+
"BLOCK_SIZE_K": 128,
|
6 |
+
"GROUP_SIZE_M": 32,
|
7 |
+
"num_warps": 4,
|
8 |
+
"num_stages": 4
|
9 |
+
},
|
10 |
+
"2": {
|
11 |
+
"BLOCK_SIZE_M": 16,
|
12 |
+
"BLOCK_SIZE_N": 64,
|
13 |
+
"BLOCK_SIZE_K": 128,
|
14 |
+
"GROUP_SIZE_M": 64,
|
15 |
+
"num_warps": 4,
|
16 |
+
"num_stages": 4
|
17 |
+
},
|
18 |
+
"4": {
|
19 |
+
"BLOCK_SIZE_M": 16,
|
20 |
+
"BLOCK_SIZE_N": 128,
|
21 |
+
"BLOCK_SIZE_K": 128,
|
22 |
+
"GROUP_SIZE_M": 64,
|
23 |
+
"num_warps": 4,
|
24 |
+
"num_stages": 3
|
25 |
+
},
|
26 |
+
"8": {
|
27 |
+
"BLOCK_SIZE_M": 16,
|
28 |
+
"BLOCK_SIZE_N": 128,
|
29 |
+
"BLOCK_SIZE_K": 128,
|
30 |
+
"GROUP_SIZE_M": 64,
|
31 |
+
"num_warps": 4,
|
32 |
+
"num_stages": 3
|
33 |
+
},
|
34 |
+
"16": {
|
35 |
+
"BLOCK_SIZE_M": 16,
|
36 |
+
"BLOCK_SIZE_N": 128,
|
37 |
+
"BLOCK_SIZE_K": 128,
|
38 |
+
"GROUP_SIZE_M": 16,
|
39 |
+
"num_warps": 4,
|
40 |
+
"num_stages": 4
|
41 |
+
},
|
42 |
+
"24": {
|
43 |
+
"BLOCK_SIZE_M": 16,
|
44 |
+
"BLOCK_SIZE_N": 128,
|
45 |
+
"BLOCK_SIZE_K": 128,
|
46 |
+
"GROUP_SIZE_M": 1,
|
47 |
+
"num_warps": 4,
|
48 |
+
"num_stages": 3
|
49 |
+
},
|
50 |
+
"32": {
|
51 |
+
"BLOCK_SIZE_M": 16,
|
52 |
+
"BLOCK_SIZE_N": 128,
|
53 |
+
"BLOCK_SIZE_K": 128,
|
54 |
+
"GROUP_SIZE_M": 16,
|
55 |
+
"num_warps": 4,
|
56 |
+
"num_stages": 2
|
57 |
+
},
|
58 |
+
"48": {
|
59 |
+
"BLOCK_SIZE_M": 16,
|
60 |
+
"BLOCK_SIZE_N": 128,
|
61 |
+
"BLOCK_SIZE_K": 128,
|
62 |
+
"GROUP_SIZE_M": 1,
|
63 |
+
"num_warps": 4,
|
64 |
+
"num_stages": 2
|
65 |
+
},
|
66 |
+
"64": {
|
67 |
+
"BLOCK_SIZE_M": 16,
|
68 |
+
"BLOCK_SIZE_N": 128,
|
69 |
+
"BLOCK_SIZE_K": 128,
|
70 |
+
"GROUP_SIZE_M": 64,
|
71 |
+
"num_warps": 4,
|
72 |
+
"num_stages": 2
|
73 |
+
},
|
74 |
+
"96": {
|
75 |
+
"BLOCK_SIZE_M": 16,
|
76 |
+
"BLOCK_SIZE_N": 128,
|
77 |
+
"BLOCK_SIZE_K": 128,
|
78 |
+
"GROUP_SIZE_M": 64,
|
79 |
+
"num_warps": 4,
|
80 |
+
"num_stages": 4
|
81 |
+
},
|
82 |
+
"128": {
|
83 |
+
"BLOCK_SIZE_M": 16,
|
84 |
+
"BLOCK_SIZE_N": 64,
|
85 |
+
"BLOCK_SIZE_K": 128,
|
86 |
+
"GROUP_SIZE_M": 64,
|
87 |
+
"num_warps": 4,
|
88 |
+
"num_stages": 5
|
89 |
+
},
|
90 |
+
"256": {
|
91 |
+
"BLOCK_SIZE_M": 16,
|
92 |
+
"BLOCK_SIZE_N": 128,
|
93 |
+
"BLOCK_SIZE_K": 128,
|
94 |
+
"GROUP_SIZE_M": 64,
|
95 |
+
"num_warps": 4,
|
96 |
+
"num_stages": 4
|
97 |
+
},
|
98 |
+
"512": {
|
99 |
+
"BLOCK_SIZE_M": 32,
|
100 |
+
"BLOCK_SIZE_N": 128,
|
101 |
+
"BLOCK_SIZE_K": 128,
|
102 |
+
"GROUP_SIZE_M": 1,
|
103 |
+
"num_warps": 4,
|
104 |
+
"num_stages": 3
|
105 |
+
},
|
106 |
+
"1024": {
|
107 |
+
"BLOCK_SIZE_M": 32,
|
108 |
+
"BLOCK_SIZE_N": 128,
|
109 |
+
"BLOCK_SIZE_K": 128,
|
110 |
+
"GROUP_SIZE_M": 64,
|
111 |
+
"num_warps": 4,
|
112 |
+
"num_stages": 3
|
113 |
+
},
|
114 |
+
"1536": {
|
115 |
+
"BLOCK_SIZE_M": 32,
|
116 |
+
"BLOCK_SIZE_N": 128,
|
117 |
+
"BLOCK_SIZE_K": 128,
|
118 |
+
"GROUP_SIZE_M": 16,
|
119 |
+
"num_warps": 4,
|
120 |
+
"num_stages": 3
|
121 |
+
},
|
122 |
+
"2048": {
|
123 |
+
"BLOCK_SIZE_M": 32,
|
124 |
+
"BLOCK_SIZE_N": 128,
|
125 |
+
"BLOCK_SIZE_K": 128,
|
126 |
+
"GROUP_SIZE_M": 16,
|
127 |
+
"num_warps": 4,
|
128 |
+
"num_stages": 3
|
129 |
+
},
|
130 |
+
"3072": {
|
131 |
+
"BLOCK_SIZE_M": 32,
|
132 |
+
"BLOCK_SIZE_N": 128,
|
133 |
+
"BLOCK_SIZE_K": 128,
|
134 |
+
"GROUP_SIZE_M": 16,
|
135 |
+
"num_warps": 4,
|
136 |
+
"num_stages": 3
|
137 |
+
},
|
138 |
+
"4096": {
|
139 |
+
"BLOCK_SIZE_M": 32,
|
140 |
+
"BLOCK_SIZE_N": 128,
|
141 |
+
"BLOCK_SIZE_K": 128,
|
142 |
+
"GROUP_SIZE_M": 16,
|
143 |
+
"num_warps": 4,
|
144 |
+
"num_stages": 3
|
145 |
+
}
|
146 |
+
}
|
build/torch25-cxx11-cu118-x86_64-linux/moe/configs/E=256,N=128,device_name=NVIDIA_A100-SXM4-80GB,dtype=int8_w8a8.json
ADDED
@@ -0,0 +1,146 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"1": {
|
3 |
+
"BLOCK_SIZE_M": 16,
|
4 |
+
"BLOCK_SIZE_N": 64,
|
5 |
+
"BLOCK_SIZE_K": 128,
|
6 |
+
"GROUP_SIZE_M": 64,
|
7 |
+
"num_warps": 4,
|
8 |
+
"num_stages": 3
|
9 |
+
},
|
10 |
+
"2": {
|
11 |
+
"BLOCK_SIZE_M": 16,
|
12 |
+
"BLOCK_SIZE_N": 64,
|
13 |
+
"BLOCK_SIZE_K": 128,
|
14 |
+
"GROUP_SIZE_M": 64,
|
15 |
+
"num_warps": 4,
|
16 |
+
"num_stages": 3
|
17 |
+
},
|
18 |
+
"4": {
|
19 |
+
"BLOCK_SIZE_M": 16,
|
20 |
+
"BLOCK_SIZE_N": 64,
|
21 |
+
"BLOCK_SIZE_K": 128,
|
22 |
+
"GROUP_SIZE_M": 16,
|
23 |
+
"num_warps": 4,
|
24 |
+
"num_stages": 3
|
25 |
+
},
|
26 |
+
"8": {
|
27 |
+
"BLOCK_SIZE_M": 16,
|
28 |
+
"BLOCK_SIZE_N": 128,
|
29 |
+
"BLOCK_SIZE_K": 128,
|
30 |
+
"GROUP_SIZE_M": 64,
|
31 |
+
"num_warps": 4,
|
32 |
+
"num_stages": 3
|
33 |
+
},
|
34 |
+
"16": {
|
35 |
+
"BLOCK_SIZE_M": 16,
|
36 |
+
"BLOCK_SIZE_N": 64,
|
37 |
+
"BLOCK_SIZE_K": 128,
|
38 |
+
"GROUP_SIZE_M": 1,
|
39 |
+
"num_warps": 4,
|
40 |
+
"num_stages": 3
|
41 |
+
},
|
42 |
+
"24": {
|
43 |
+
"BLOCK_SIZE_M": 16,
|
44 |
+
"BLOCK_SIZE_N": 64,
|
45 |
+
"BLOCK_SIZE_K": 128,
|
46 |
+
"GROUP_SIZE_M": 1,
|
47 |
+
"num_warps": 4,
|
48 |
+
"num_stages": 3
|
49 |
+
},
|
50 |
+
"32": {
|
51 |
+
"BLOCK_SIZE_M": 16,
|
52 |
+
"BLOCK_SIZE_N": 64,
|
53 |
+
"BLOCK_SIZE_K": 128,
|
54 |
+
"GROUP_SIZE_M": 32,
|
55 |
+
"num_warps": 4,
|
56 |
+
"num_stages": 2
|
57 |
+
},
|
58 |
+
"48": {
|
59 |
+
"BLOCK_SIZE_M": 16,
|
60 |
+
"BLOCK_SIZE_N": 128,
|
61 |
+
"BLOCK_SIZE_K": 128,
|
62 |
+
"GROUP_SIZE_M": 32,
|
63 |
+
"num_warps": 4,
|
64 |
+
"num_stages": 2
|
65 |
+
},
|
66 |
+
"64": {
|
67 |
+
"BLOCK_SIZE_M": 16,
|
68 |
+
"BLOCK_SIZE_N": 128,
|
69 |
+
"BLOCK_SIZE_K": 128,
|
70 |
+
"GROUP_SIZE_M": 1,
|
71 |
+
"num_warps": 4,
|
72 |
+
"num_stages": 2
|
73 |
+
},
|
74 |
+
"96": {
|
75 |
+
"BLOCK_SIZE_M": 16,
|
76 |
+
"BLOCK_SIZE_N": 128,
|
77 |
+
"BLOCK_SIZE_K": 128,
|
78 |
+
"GROUP_SIZE_M": 1,
|
79 |
+
"num_warps": 4,
|
80 |
+
"num_stages": 2
|
81 |
+
},
|
82 |
+
"128": {
|
83 |
+
"BLOCK_SIZE_M": 16,
|
84 |
+
"BLOCK_SIZE_N": 128,
|
85 |
+
"BLOCK_SIZE_K": 128,
|
86 |
+
"GROUP_SIZE_M": 64,
|
87 |
+
"num_warps": 4,
|
88 |
+
"num_stages": 2
|
89 |
+
},
|
90 |
+
"256": {
|
91 |
+
"BLOCK_SIZE_M": 16,
|
92 |
+
"BLOCK_SIZE_N": 128,
|
93 |
+
"BLOCK_SIZE_K": 128,
|
94 |
+
"GROUP_SIZE_M": 64,
|
95 |
+
"num_warps": 4,
|
96 |
+
"num_stages": 2
|
97 |
+
},
|
98 |
+
"512": {
|
99 |
+
"BLOCK_SIZE_M": 32,
|
100 |
+
"BLOCK_SIZE_N": 128,
|
101 |
+
"BLOCK_SIZE_K": 128,
|
102 |
+
"GROUP_SIZE_M": 1,
|
103 |
+
"num_warps": 4,
|
104 |
+
"num_stages": 3
|
105 |
+
},
|
106 |
+
"1024": {
|
107 |
+
"BLOCK_SIZE_M": 32,
|
108 |
+
"BLOCK_SIZE_N": 128,
|
109 |
+
"BLOCK_SIZE_K": 128,
|
110 |
+
"GROUP_SIZE_M": 64,
|
111 |
+
"num_warps": 4,
|
112 |
+
"num_stages": 3
|
113 |
+
},
|
114 |
+
"1536": {
|
115 |
+
"BLOCK_SIZE_M": 64,
|
116 |
+
"BLOCK_SIZE_N": 128,
|
117 |
+
"BLOCK_SIZE_K": 128,
|
118 |
+
"GROUP_SIZE_M": 64,
|
119 |
+
"num_warps": 4,
|
120 |
+
"num_stages": 3
|
121 |
+
},
|
122 |
+
"2048": {
|
123 |
+
"BLOCK_SIZE_M": 32,
|
124 |
+
"BLOCK_SIZE_N": 128,
|
125 |
+
"BLOCK_SIZE_K": 128,
|
126 |
+
"GROUP_SIZE_M": 16,
|
127 |
+
"num_warps": 4,
|
128 |
+
"num_stages": 3
|
129 |
+
},
|
130 |
+
"3072": {
|
131 |
+
"BLOCK_SIZE_M": 32,
|
132 |
+
"BLOCK_SIZE_N": 128,
|
133 |
+
"BLOCK_SIZE_K": 128,
|
134 |
+
"GROUP_SIZE_M": 16,
|
135 |
+
"num_warps": 4,
|
136 |
+
"num_stages": 3
|
137 |
+
},
|
138 |
+
"4096": {
|
139 |
+
"BLOCK_SIZE_M": 32,
|
140 |
+
"BLOCK_SIZE_N": 128,
|
141 |
+
"BLOCK_SIZE_K": 128,
|
142 |
+
"GROUP_SIZE_M": 16,
|
143 |
+
"num_warps": 4,
|
144 |
+
"num_stages": 3
|
145 |
+
}
|
146 |
+
}
|
build/torch25-cxx11-cu118-x86_64-linux/moe/configs/E=256,N=128,device_name=NVIDIA_A800-SXM4-80GB,dtype=int8_w8a8,block_shape=[128,128].json
ADDED
@@ -0,0 +1,146 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"1": {
|
3 |
+
"BLOCK_SIZE_M": 16,
|
4 |
+
"BLOCK_SIZE_N": 64,
|
5 |
+
"BLOCK_SIZE_K": 128,
|
6 |
+
"GROUP_SIZE_M": 32,
|
7 |
+
"num_warps": 4,
|
8 |
+
"num_stages": 3
|
9 |
+
},
|
10 |
+
"2": {
|
11 |
+
"BLOCK_SIZE_M": 16,
|
12 |
+
"BLOCK_SIZE_N": 64,
|
13 |
+
"BLOCK_SIZE_K": 128,
|
14 |
+
"GROUP_SIZE_M": 64,
|
15 |
+
"num_warps": 4,
|
16 |
+
"num_stages": 3
|
17 |
+
},
|
18 |
+
"4": {
|
19 |
+
"BLOCK_SIZE_M": 16,
|
20 |
+
"BLOCK_SIZE_N": 128,
|
21 |
+
"BLOCK_SIZE_K": 128,
|
22 |
+
"GROUP_SIZE_M": 64,
|
23 |
+
"num_warps": 4,
|
24 |
+
"num_stages": 3
|
25 |
+
},
|
26 |
+
"8": {
|
27 |
+
"BLOCK_SIZE_M": 16,
|
28 |
+
"BLOCK_SIZE_N": 128,
|
29 |
+
"BLOCK_SIZE_K": 128,
|
30 |
+
"GROUP_SIZE_M": 64,
|
31 |
+
"num_warps": 4,
|
32 |
+
"num_stages": 3
|
33 |
+
},
|
34 |
+
"16": {
|
35 |
+
"BLOCK_SIZE_M": 16,
|
36 |
+
"BLOCK_SIZE_N": 64,
|
37 |
+
"BLOCK_SIZE_K": 128,
|
38 |
+
"GROUP_SIZE_M": 1,
|
39 |
+
"num_warps": 4,
|
40 |
+
"num_stages": 3
|
41 |
+
},
|
42 |
+
"24": {
|
43 |
+
"BLOCK_SIZE_M": 16,
|
44 |
+
"BLOCK_SIZE_N": 128,
|
45 |
+
"BLOCK_SIZE_K": 128,
|
46 |
+
"GROUP_SIZE_M": 1,
|
47 |
+
"num_warps": 4,
|
48 |
+
"num_stages": 2
|
49 |
+
},
|
50 |
+
"32": {
|
51 |
+
"BLOCK_SIZE_M": 16,
|
52 |
+
"BLOCK_SIZE_N": 128,
|
53 |
+
"BLOCK_SIZE_K": 128,
|
54 |
+
"GROUP_SIZE_M": 1,
|
55 |
+
"num_warps": 4,
|
56 |
+
"num_stages": 2
|
57 |
+
},
|
58 |
+
"48": {
|
59 |
+
"BLOCK_SIZE_M": 16,
|
60 |
+
"BLOCK_SIZE_N": 128,
|
61 |
+
"BLOCK_SIZE_K": 128,
|
62 |
+
"GROUP_SIZE_M": 1,
|
63 |
+
"num_warps": 4,
|
64 |
+
"num_stages": 2
|
65 |
+
},
|
66 |
+
"64": {
|
67 |
+
"BLOCK_SIZE_M": 16,
|
68 |
+
"BLOCK_SIZE_N": 128,
|
69 |
+
"BLOCK_SIZE_K": 128,
|
70 |
+
"GROUP_SIZE_M": 1,
|
71 |
+
"num_warps": 4,
|
72 |
+
"num_stages": 3
|
73 |
+
},
|
74 |
+
"96": {
|
75 |
+
"BLOCK_SIZE_M": 16,
|
76 |
+
"BLOCK_SIZE_N": 128,
|
77 |
+
"BLOCK_SIZE_K": 128,
|
78 |
+
"GROUP_SIZE_M": 1,
|
79 |
+
"num_warps": 4,
|
80 |
+
"num_stages": 3
|
81 |
+
},
|
82 |
+
"128": {
|
83 |
+
"BLOCK_SIZE_M": 16,
|
84 |
+
"BLOCK_SIZE_N": 128,
|
85 |
+
"BLOCK_SIZE_K": 128,
|
86 |
+
"GROUP_SIZE_M": 1,
|
87 |
+
"num_warps": 4,
|
88 |
+
"num_stages": 3
|
89 |
+
},
|
90 |
+
"256": {
|
91 |
+
"BLOCK_SIZE_M": 16,
|
92 |
+
"BLOCK_SIZE_N": 128,
|
93 |
+
"BLOCK_SIZE_K": 128,
|
94 |
+
"GROUP_SIZE_M": 64,
|
95 |
+
"num_warps": 4,
|
96 |
+
"num_stages": 3
|
97 |
+
},
|
98 |
+
"512": {
|
99 |
+
"BLOCK_SIZE_M": 16,
|
100 |
+
"BLOCK_SIZE_N": 128,
|
101 |
+
"BLOCK_SIZE_K": 128,
|
102 |
+
"GROUP_SIZE_M": 64,
|
103 |
+
"num_warps": 4,
|
104 |
+
"num_stages": 2
|
105 |
+
},
|
106 |
+
"1024": {
|
107 |
+
"BLOCK_SIZE_M": 32,
|
108 |
+
"BLOCK_SIZE_N": 128,
|
109 |
+
"BLOCK_SIZE_K": 128,
|
110 |
+
"GROUP_SIZE_M": 64,
|
111 |
+
"num_warps": 4,
|
112 |
+
"num_stages": 3
|
113 |
+
},
|
114 |
+
"1536": {
|
115 |
+
"BLOCK_SIZE_M": 32,
|
116 |
+
"BLOCK_SIZE_N": 128,
|
117 |
+
"BLOCK_SIZE_K": 128,
|
118 |
+
"GROUP_SIZE_M": 64,
|
119 |
+
"num_warps": 4,
|
120 |
+
"num_stages": 3
|
121 |
+
},
|
122 |
+
"2048": {
|
123 |
+
"BLOCK_SIZE_M": 32,
|
124 |
+
"BLOCK_SIZE_N": 128,
|
125 |
+
"BLOCK_SIZE_K": 128,
|
126 |
+
"GROUP_SIZE_M": 32,
|
127 |
+
"num_warps": 4,
|
128 |
+
"num_stages": 3
|
129 |
+
},
|
130 |
+
"3072": {
|
131 |
+
"BLOCK_SIZE_M": 32,
|
132 |
+
"BLOCK_SIZE_N": 128,
|
133 |
+
"BLOCK_SIZE_K": 128,
|
134 |
+
"GROUP_SIZE_M": 16,
|
135 |
+
"num_warps": 4,
|
136 |
+
"num_stages": 3
|
137 |
+
},
|
138 |
+
"4096": {
|
139 |
+
"BLOCK_SIZE_M": 32,
|
140 |
+
"BLOCK_SIZE_N": 128,
|
141 |
+
"BLOCK_SIZE_K": 128,
|
142 |
+
"GROUP_SIZE_M": 16,
|
143 |
+
"num_warps": 4,
|
144 |
+
"num_stages": 3
|
145 |
+
}
|
146 |
+
}
|
build/torch25-cxx11-cu118-x86_64-linux/moe/configs/E=256,N=128,device_name=NVIDIA_A800-SXM4-80GB,dtype=int8_w8a8.json
ADDED
@@ -0,0 +1,146 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"1": {
|
3 |
+
"BLOCK_SIZE_M": 16,
|
4 |
+
"BLOCK_SIZE_N": 64,
|
5 |
+
"BLOCK_SIZE_K": 128,
|
6 |
+
"GROUP_SIZE_M": 64,
|
7 |
+
"num_warps": 4,
|
8 |
+
"num_stages": 3
|
9 |
+
},
|
10 |
+
"2": {
|
11 |
+
"BLOCK_SIZE_M": 16,
|
12 |
+
"BLOCK_SIZE_N": 64,
|
13 |
+
"BLOCK_SIZE_K": 128,
|
14 |
+
"GROUP_SIZE_M": 64,
|
15 |
+
"num_warps": 4,
|
16 |
+
"num_stages": 3
|
17 |
+
},
|
18 |
+
"4": {
|
19 |
+
"BLOCK_SIZE_M": 16,
|
20 |
+
"BLOCK_SIZE_N": 64,
|
21 |
+
"BLOCK_SIZE_K": 128,
|
22 |
+
"GROUP_SIZE_M": 1,
|
23 |
+
"num_warps": 4,
|
24 |
+
"num_stages": 3
|
25 |
+
},
|
26 |
+
"8": {
|
27 |
+
"BLOCK_SIZE_M": 16,
|
28 |
+
"BLOCK_SIZE_N": 128,
|
29 |
+
"BLOCK_SIZE_K": 128,
|
30 |
+
"GROUP_SIZE_M": 64,
|
31 |
+
"num_warps": 4,
|
32 |
+
"num_stages": 3
|
33 |
+
},
|
34 |
+
"16": {
|
35 |
+
"BLOCK_SIZE_M": 16,
|
36 |
+
"BLOCK_SIZE_N": 64,
|
37 |
+
"BLOCK_SIZE_K": 128,
|
38 |
+
"GROUP_SIZE_M": 1,
|
39 |
+
"num_warps": 4,
|
40 |
+
"num_stages": 3
|
41 |
+
},
|
42 |
+
"24": {
|
43 |
+
"BLOCK_SIZE_M": 16,
|
44 |
+
"BLOCK_SIZE_N": 64,
|
45 |
+
"BLOCK_SIZE_K": 128,
|
46 |
+
"GROUP_SIZE_M": 1,
|
47 |
+
"num_warps": 4,
|
48 |
+
"num_stages": 3
|
49 |
+
},
|
50 |
+
"32": {
|
51 |
+
"BLOCK_SIZE_M": 16,
|
52 |
+
"BLOCK_SIZE_N": 64,
|
53 |
+
"BLOCK_SIZE_K": 128,
|
54 |
+
"GROUP_SIZE_M": 16,
|
55 |
+
"num_warps": 4,
|
56 |
+
"num_stages": 2
|
57 |
+
},
|
58 |
+
"48": {
|
59 |
+
"BLOCK_SIZE_M": 16,
|
60 |
+
"BLOCK_SIZE_N": 128,
|
61 |
+
"BLOCK_SIZE_K": 128,
|
62 |
+
"GROUP_SIZE_M": 1,
|
63 |
+
"num_warps": 4,
|
64 |
+
"num_stages": 2
|
65 |
+
},
|
66 |
+
"64": {
|
67 |
+
"BLOCK_SIZE_M": 16,
|
68 |
+
"BLOCK_SIZE_N": 128,
|
69 |
+
"BLOCK_SIZE_K": 128,
|
70 |
+
"GROUP_SIZE_M": 1,
|
71 |
+
"num_warps": 4,
|
72 |
+
"num_stages": 2
|
73 |
+
},
|
74 |
+
"96": {
|
75 |
+
"BLOCK_SIZE_M": 16,
|
76 |
+
"BLOCK_SIZE_N": 128,
|
77 |
+
"BLOCK_SIZE_K": 128,
|
78 |
+
"GROUP_SIZE_M": 1,
|
79 |
+
"num_warps": 4,
|
80 |
+
"num_stages": 2
|
81 |
+
},
|
82 |
+
"128": {
|
83 |
+
"BLOCK_SIZE_M": 16,
|
84 |
+
"BLOCK_SIZE_N": 128,
|
85 |
+
"BLOCK_SIZE_K": 128,
|
86 |
+
"GROUP_SIZE_M": 64,
|
87 |
+
"num_warps": 4,
|
88 |
+
"num_stages": 2
|
89 |
+
},
|
90 |
+
"256": {
|
91 |
+
"BLOCK_SIZE_M": 16,
|
92 |
+
"BLOCK_SIZE_N": 128,
|
93 |
+
"BLOCK_SIZE_K": 128,
|
94 |
+
"GROUP_SIZE_M": 64,
|
95 |
+
"num_warps": 4,
|
96 |
+
"num_stages": 2
|
97 |
+
},
|
98 |
+
"512": {
|
99 |
+
"BLOCK_SIZE_M": 32,
|
100 |
+
"BLOCK_SIZE_N": 128,
|
101 |
+
"BLOCK_SIZE_K": 128,
|
102 |
+
"GROUP_SIZE_M": 1,
|
103 |
+
"num_warps": 4,
|
104 |
+
"num_stages": 3
|
105 |
+
},
|
106 |
+
"1024": {
|
107 |
+
"BLOCK_SIZE_M": 32,
|
108 |
+
"BLOCK_SIZE_N": 128,
|
109 |
+
"BLOCK_SIZE_K": 128,
|
110 |
+
"GROUP_SIZE_M": 32,
|
111 |
+
"num_warps": 4,
|
112 |
+
"num_stages": 3
|
113 |
+
},
|
114 |
+
"1536": {
|
115 |
+
"BLOCK_SIZE_M": 64,
|
116 |
+
"BLOCK_SIZE_N": 64,
|
117 |
+
"BLOCK_SIZE_K": 128,
|
118 |
+
"GROUP_SIZE_M": 16,
|
119 |
+
"num_warps": 4,
|
120 |
+
"num_stages": 3
|
121 |
+
},
|
122 |
+
"2048": {
|
123 |
+
"BLOCK_SIZE_M": 32,
|
124 |
+
"BLOCK_SIZE_N": 128,
|
125 |
+
"BLOCK_SIZE_K": 128,
|
126 |
+
"GROUP_SIZE_M": 1,
|
127 |
+
"num_warps": 4,
|
128 |
+
"num_stages": 3
|
129 |
+
},
|
130 |
+
"3072": {
|
131 |
+
"BLOCK_SIZE_M": 32,
|
132 |
+
"BLOCK_SIZE_N": 128,
|
133 |
+
"BLOCK_SIZE_K": 128,
|
134 |
+
"GROUP_SIZE_M": 16,
|
135 |
+
"num_warps": 4,
|
136 |
+
"num_stages": 3
|
137 |
+
},
|
138 |
+
"4096": {
|
139 |
+
"BLOCK_SIZE_M": 32,
|
140 |
+
"BLOCK_SIZE_N": 128,
|
141 |
+
"BLOCK_SIZE_K": 128,
|
142 |
+
"GROUP_SIZE_M": 16,
|
143 |
+
"num_warps": 4,
|
144 |
+
"num_stages": 3
|
145 |
+
}
|
146 |
+
}
|
build/torch25-cxx11-cu118-x86_64-linux/moe/configs/E=256,N=128,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[128,128].json
ADDED
@@ -0,0 +1,146 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"1": {
|
3 |
+
"BLOCK_SIZE_M": 64,
|
4 |
+
"BLOCK_SIZE_N": 64,
|
5 |
+
"BLOCK_SIZE_K": 128,
|
6 |
+
"GROUP_SIZE_M": 16,
|
7 |
+
"num_warps": 4,
|
8 |
+
"num_stages": 3
|
9 |
+
},
|
10 |
+
"2": {
|
11 |
+
"BLOCK_SIZE_M": 64,
|
12 |
+
"BLOCK_SIZE_N": 64,
|
13 |
+
"BLOCK_SIZE_K": 128,
|
14 |
+
"GROUP_SIZE_M": 1,
|
15 |
+
"num_warps": 4,
|
16 |
+
"num_stages": 3
|
17 |
+
},
|
18 |
+
"4": {
|
19 |
+
"BLOCK_SIZE_M": 64,
|
20 |
+
"BLOCK_SIZE_N": 64,
|
21 |
+
"BLOCK_SIZE_K": 128,
|
22 |
+
"GROUP_SIZE_M": 16,
|
23 |
+
"num_warps": 4,
|
24 |
+
"num_stages": 3
|
25 |
+
},
|
26 |
+
"8": {
|
27 |
+
"BLOCK_SIZE_M": 64,
|
28 |
+
"BLOCK_SIZE_N": 64,
|
29 |
+
"BLOCK_SIZE_K": 128,
|
30 |
+
"GROUP_SIZE_M": 32,
|
31 |
+
"num_warps": 4,
|
32 |
+
"num_stages": 3
|
33 |
+
},
|
34 |
+
"16": {
|
35 |
+
"BLOCK_SIZE_M": 64,
|
36 |
+
"BLOCK_SIZE_N": 128,
|
37 |
+
"BLOCK_SIZE_K": 128,
|
38 |
+
"GROUP_SIZE_M": 16,
|
39 |
+
"num_warps": 4,
|
40 |
+
"num_stages": 3
|
41 |
+
},
|
42 |
+
"24": {
|
43 |
+
"BLOCK_SIZE_M": 64,
|
44 |
+
"BLOCK_SIZE_N": 128,
|
45 |
+
"BLOCK_SIZE_K": 128,
|
46 |
+
"GROUP_SIZE_M": 1,
|
47 |
+
"num_warps": 4,
|
48 |
+
"num_stages": 3
|
49 |
+
},
|
50 |
+
"32": {
|
51 |
+
"BLOCK_SIZE_M": 64,
|
52 |
+
"BLOCK_SIZE_N": 64,
|
53 |
+
"BLOCK_SIZE_K": 128,
|
54 |
+
"GROUP_SIZE_M": 64,
|
55 |
+
"num_warps": 4,
|
56 |
+
"num_stages": 3
|
57 |
+
},
|
58 |
+
"48": {
|
59 |
+
"BLOCK_SIZE_M": 64,
|
60 |
+
"BLOCK_SIZE_N": 64,
|
61 |
+
"BLOCK_SIZE_K": 128,
|
62 |
+
"GROUP_SIZE_M": 64,
|
63 |
+
"num_warps": 4,
|
64 |
+
"num_stages": 3
|
65 |
+
},
|
66 |
+
"64": {
|
67 |
+
"BLOCK_SIZE_M": 64,
|
68 |
+
"BLOCK_SIZE_N": 128,
|
69 |
+
"BLOCK_SIZE_K": 128,
|
70 |
+
"GROUP_SIZE_M": 16,
|
71 |
+
"num_warps": 4,
|
72 |
+
"num_stages": 3
|
73 |
+
},
|
74 |
+
"96": {
|
75 |
+
"BLOCK_SIZE_M": 64,
|
76 |
+
"BLOCK_SIZE_N": 128,
|
77 |
+
"BLOCK_SIZE_K": 128,
|
78 |
+
"GROUP_SIZE_M": 64,
|
79 |
+
"num_warps": 4,
|
80 |
+
"num_stages": 3
|
81 |
+
},
|
82 |
+
"128": {
|
83 |
+
"BLOCK_SIZE_M": 64,
|
84 |
+
"BLOCK_SIZE_N": 128,
|
85 |
+
"BLOCK_SIZE_K": 128,
|
86 |
+
"GROUP_SIZE_M": 64,
|
87 |
+
"num_warps": 4,
|
88 |
+
"num_stages": 3
|
89 |
+
},
|
90 |
+
"256": {
|
91 |
+
"BLOCK_SIZE_M": 64,
|
92 |
+
"BLOCK_SIZE_N": 128,
|
93 |
+
"BLOCK_SIZE_K": 128,
|
94 |
+
"GROUP_SIZE_M": 32,
|
95 |
+
"num_warps": 4,
|
96 |
+
"num_stages": 3
|
97 |
+
},
|
98 |
+
"512": {
|
99 |
+
"BLOCK_SIZE_M": 64,
|
100 |
+
"BLOCK_SIZE_N": 128,
|
101 |
+
"BLOCK_SIZE_K": 128,
|
102 |
+
"GROUP_SIZE_M": 64,
|
103 |
+
"num_warps": 4,
|
104 |
+
"num_stages": 3
|
105 |
+
},
|
106 |
+
"1024": {
|
107 |
+
"BLOCK_SIZE_M": 64,
|
108 |
+
"BLOCK_SIZE_N": 128,
|
109 |
+
"BLOCK_SIZE_K": 128,
|
110 |
+
"GROUP_SIZE_M": 64,
|
111 |
+
"num_warps": 4,
|
112 |
+
"num_stages": 3
|
113 |
+
},
|
114 |
+
"1536": {
|
115 |
+
"BLOCK_SIZE_M": 64,
|
116 |
+
"BLOCK_SIZE_N": 128,
|
117 |
+
"BLOCK_SIZE_K": 128,
|
118 |
+
"GROUP_SIZE_M": 64,
|
119 |
+
"num_warps": 4,
|
120 |
+
"num_stages": 3
|
121 |
+
},
|
122 |
+
"2048": {
|
123 |
+
"BLOCK_SIZE_M": 64,
|
124 |
+
"BLOCK_SIZE_N": 128,
|
125 |
+
"BLOCK_SIZE_K": 128,
|
126 |
+
"GROUP_SIZE_M": 32,
|
127 |
+
"num_warps": 4,
|
128 |
+
"num_stages": 3
|
129 |
+
},
|
130 |
+
"3072": {
|
131 |
+
"BLOCK_SIZE_M": 128,
|
132 |
+
"BLOCK_SIZE_N": 64,
|
133 |
+
"BLOCK_SIZE_K": 128,
|
134 |
+
"GROUP_SIZE_M": 16,
|
135 |
+
"num_warps": 4,
|
136 |
+
"num_stages": 3
|
137 |
+
},
|
138 |
+
"4096": {
|
139 |
+
"BLOCK_SIZE_M": 64,
|
140 |
+
"BLOCK_SIZE_N": 128,
|
141 |
+
"BLOCK_SIZE_K": 128,
|
142 |
+
"GROUP_SIZE_M": 32,
|
143 |
+
"num_warps": 4,
|
144 |
+
"num_stages": 3
|
145 |
+
}
|
146 |
+
}
|
build/torch25-cxx11-cu118-x86_64-linux/moe/configs/E=256,N=128,device_name=NVIDIA_L20Y,dtype=fp8_w8a8,block_shape=[128,128].json
ADDED
@@ -0,0 +1,146 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"1": {
|
3 |
+
"BLOCK_SIZE_M": 64,
|
4 |
+
"BLOCK_SIZE_N": 64,
|
5 |
+
"BLOCK_SIZE_K": 128,
|
6 |
+
"GROUP_SIZE_M": 16,
|
7 |
+
"num_warps": 4,
|
8 |
+
"num_stages": 3
|
9 |
+
},
|
10 |
+
"2": {
|
11 |
+
"BLOCK_SIZE_M": 64,
|
12 |
+
"BLOCK_SIZE_N": 64,
|
13 |
+
"BLOCK_SIZE_K": 128,
|
14 |
+
"GROUP_SIZE_M": 16,
|
15 |
+
"num_warps": 4,
|
16 |
+
"num_stages": 3
|
17 |
+
},
|
18 |
+
"4": {
|
19 |
+
"BLOCK_SIZE_M": 16,
|
20 |
+
"BLOCK_SIZE_N": 128,
|
21 |
+
"BLOCK_SIZE_K": 128,
|
22 |
+
"GROUP_SIZE_M": 16,
|
23 |
+
"num_warps": 4,
|
24 |
+
"num_stages": 3
|
25 |
+
},
|
26 |
+
"8": {
|
27 |
+
"BLOCK_SIZE_M": 64,
|
28 |
+
"BLOCK_SIZE_N": 64,
|
29 |
+
"BLOCK_SIZE_K": 128,
|
30 |
+
"GROUP_SIZE_M": 32,
|
31 |
+
"num_warps": 4,
|
32 |
+
"num_stages": 3
|
33 |
+
},
|
34 |
+
"16": {
|
35 |
+
"BLOCK_SIZE_M": 64,
|
36 |
+
"BLOCK_SIZE_N": 128,
|
37 |
+
"BLOCK_SIZE_K": 128,
|
38 |
+
"GROUP_SIZE_M": 16,
|
39 |
+
"num_warps": 4,
|
40 |
+
"num_stages": 3
|
41 |
+
},
|
42 |
+
"24": {
|
43 |
+
"BLOCK_SIZE_M": 64,
|
44 |
+
"BLOCK_SIZE_N": 64,
|
45 |
+
"BLOCK_SIZE_K": 128,
|
46 |
+
"GROUP_SIZE_M": 1,
|
47 |
+
"num_warps": 4,
|
48 |
+
"num_stages": 3
|
49 |
+
},
|
50 |
+
"32": {
|
51 |
+
"BLOCK_SIZE_M": 64,
|
52 |
+
"BLOCK_SIZE_N": 64,
|
53 |
+
"BLOCK_SIZE_K": 128,
|
54 |
+
"GROUP_SIZE_M": 64,
|
55 |
+
"num_warps": 4,
|
56 |
+
"num_stages": 3
|
57 |
+
},
|
58 |
+
"48": {
|
59 |
+
"BLOCK_SIZE_M": 64,
|
60 |
+
"BLOCK_SIZE_N": 64,
|
61 |
+
"BLOCK_SIZE_K": 128,
|
62 |
+
"GROUP_SIZE_M": 1,
|
63 |
+
"num_warps": 4,
|
64 |
+
"num_stages": 3
|
65 |
+
},
|
66 |
+
"64": {
|
67 |
+
"BLOCK_SIZE_M": 64,
|
68 |
+
"BLOCK_SIZE_N": 128,
|
69 |
+
"BLOCK_SIZE_K": 128,
|
70 |
+
"GROUP_SIZE_M": 16,
|
71 |
+
"num_warps": 4,
|
72 |
+
"num_stages": 3
|
73 |
+
},
|
74 |
+
"96": {
|
75 |
+
"BLOCK_SIZE_M": 64,
|
76 |
+
"BLOCK_SIZE_N": 128,
|
77 |
+
"BLOCK_SIZE_K": 128,
|
78 |
+
"GROUP_SIZE_M": 16,
|
79 |
+
"num_warps": 4,
|
80 |
+
"num_stages": 3
|
81 |
+
},
|
82 |
+
"128": {
|
83 |
+
"BLOCK_SIZE_M": 64,
|
84 |
+
"BLOCK_SIZE_N": 128,
|
85 |
+
"BLOCK_SIZE_K": 128,
|
86 |
+
"GROUP_SIZE_M": 16,
|
87 |
+
"num_warps": 4,
|
88 |
+
"num_stages": 3
|
89 |
+
},
|
90 |
+
"256": {
|
91 |
+
"BLOCK_SIZE_M": 64,
|
92 |
+
"BLOCK_SIZE_N": 128,
|
93 |
+
"BLOCK_SIZE_K": 128,
|
94 |
+
"GROUP_SIZE_M": 32,
|
95 |
+
"num_warps": 4,
|
96 |
+
"num_stages": 3
|
97 |
+
},
|
98 |
+
"512": {
|
99 |
+
"BLOCK_SIZE_M": 64,
|
100 |
+
"BLOCK_SIZE_N": 128,
|
101 |
+
"BLOCK_SIZE_K": 128,
|
102 |
+
"GROUP_SIZE_M": 16,
|
103 |
+
"num_warps": 4,
|
104 |
+
"num_stages": 3
|
105 |
+
},
|
106 |
+
"1024": {
|
107 |
+
"BLOCK_SIZE_M": 64,
|
108 |
+
"BLOCK_SIZE_N": 128,
|
109 |
+
"BLOCK_SIZE_K": 128,
|
110 |
+
"GROUP_SIZE_M": 64,
|
111 |
+
"num_warps": 4,
|
112 |
+
"num_stages": 3
|
113 |
+
},
|
114 |
+
"1536": {
|
115 |
+
"BLOCK_SIZE_M": 64,
|
116 |
+
"BLOCK_SIZE_N": 128,
|
117 |
+
"BLOCK_SIZE_K": 128,
|
118 |
+
"GROUP_SIZE_M": 32,
|
119 |
+
"num_warps": 4,
|
120 |
+
"num_stages": 3
|
121 |
+
},
|
122 |
+
"2048": {
|
123 |
+
"BLOCK_SIZE_M": 64,
|
124 |
+
"BLOCK_SIZE_N": 128,
|
125 |
+
"BLOCK_SIZE_K": 128,
|
126 |
+
"GROUP_SIZE_M": 64,
|
127 |
+
"num_warps": 4,
|
128 |
+
"num_stages": 3
|
129 |
+
},
|
130 |
+
"3072": {
|
131 |
+
"BLOCK_SIZE_M": 128,
|
132 |
+
"BLOCK_SIZE_N": 64,
|
133 |
+
"BLOCK_SIZE_K": 128,
|
134 |
+
"GROUP_SIZE_M": 16,
|
135 |
+
"num_warps": 4,
|
136 |
+
"num_stages": 3
|
137 |
+
},
|
138 |
+
"4096": {
|
139 |
+
"BLOCK_SIZE_M": 64,
|
140 |
+
"BLOCK_SIZE_N": 128,
|
141 |
+
"BLOCK_SIZE_K": 128,
|
142 |
+
"GROUP_SIZE_M": 64,
|
143 |
+
"num_warps": 4,
|
144 |
+
"num_stages": 3
|
145 |
+
}
|
146 |
+
}
|
build/torch25-cxx11-cu118-x86_64-linux/moe/configs/E=256,N=256,device_name=AMD_Instinct_MI300X,dtype=fp8_w8a8,block_shape=[128,128].json
ADDED
@@ -0,0 +1,164 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"1": {
|
3 |
+
"BLOCK_SIZE_M": 16,
|
4 |
+
"BLOCK_SIZE_N": 128,
|
5 |
+
"BLOCK_SIZE_K": 256,
|
6 |
+
"GROUP_SIZE_M": 1,
|
7 |
+
"num_warps": 8,
|
8 |
+
"num_stages": 2,
|
9 |
+
"waves_per_eu": 0
|
10 |
+
},
|
11 |
+
"2": {
|
12 |
+
"BLOCK_SIZE_M": 16,
|
13 |
+
"BLOCK_SIZE_N": 128,
|
14 |
+
"BLOCK_SIZE_K": 256,
|
15 |
+
"GROUP_SIZE_M": 1,
|
16 |
+
"num_warps": 8,
|
17 |
+
"num_stages": 2,
|
18 |
+
"waves_per_eu": 0
|
19 |
+
},
|
20 |
+
"4": {
|
21 |
+
"BLOCK_SIZE_M": 16,
|
22 |
+
"BLOCK_SIZE_N": 128,
|
23 |
+
"BLOCK_SIZE_K": 256,
|
24 |
+
"GROUP_SIZE_M": 1,
|
25 |
+
"num_warps": 8,
|
26 |
+
"num_stages": 2,
|
27 |
+
"waves_per_eu": 0
|
28 |
+
},
|
29 |
+
"8": {
|
30 |
+
"BLOCK_SIZE_M": 16,
|
31 |
+
"BLOCK_SIZE_N": 128,
|
32 |
+
"BLOCK_SIZE_K": 128,
|
33 |
+
"GROUP_SIZE_M": 1,
|
34 |
+
"num_warps": 8,
|
35 |
+
"num_stages": 2,
|
36 |
+
"waves_per_eu": 0
|
37 |
+
},
|
38 |
+
"16": {
|
39 |
+
"BLOCK_SIZE_M": 16,
|
40 |
+
"BLOCK_SIZE_N": 128,
|
41 |
+
"BLOCK_SIZE_K": 128,
|
42 |
+
"GROUP_SIZE_M": 1,
|
43 |
+
"num_warps": 2,
|
44 |
+
"num_stages": 2,
|
45 |
+
"waves_per_eu": 0
|
46 |
+
},
|
47 |
+
"24": {
|
48 |
+
"BLOCK_SIZE_M": 16,
|
49 |
+
"BLOCK_SIZE_N": 128,
|
50 |
+
"BLOCK_SIZE_K": 128,
|
51 |
+
"GROUP_SIZE_M": 1,
|
52 |
+
"num_warps": 2,
|
53 |
+
"num_stages": 2,
|
54 |
+
"waves_per_eu": 0
|
55 |
+
},
|
56 |
+
"32": {
|
57 |
+
"BLOCK_SIZE_M": 16,
|
58 |
+
"BLOCK_SIZE_N": 128,
|
59 |
+
"BLOCK_SIZE_K": 128,
|
60 |
+
"GROUP_SIZE_M": 4,
|
61 |
+
"num_warps": 2,
|
62 |
+
"num_stages": 2,
|
63 |
+
"waves_per_eu": 0
|
64 |
+
},
|
65 |
+
"48": {
|
66 |
+
"BLOCK_SIZE_M": 16,
|
67 |
+
"BLOCK_SIZE_N": 128,
|
68 |
+
"BLOCK_SIZE_K": 128,
|
69 |
+
"GROUP_SIZE_M": 4,
|
70 |
+
"num_warps": 2,
|
71 |
+
"num_stages": 2,
|
72 |
+
"waves_per_eu": 0
|
73 |
+
},
|
74 |
+
"64": {
|
75 |
+
"BLOCK_SIZE_M": 16,
|
76 |
+
"BLOCK_SIZE_N": 128,
|
77 |
+
"BLOCK_SIZE_K": 128,
|
78 |
+
"GROUP_SIZE_M": 1,
|
79 |
+
"num_warps": 2,
|
80 |
+
"num_stages": 2,
|
81 |
+
"waves_per_eu": 0
|
82 |
+
},
|
83 |
+
"96": {
|
84 |
+
"BLOCK_SIZE_M": 16,
|
85 |
+
"BLOCK_SIZE_N": 128,
|
86 |
+
"BLOCK_SIZE_K": 128,
|
87 |
+
"GROUP_SIZE_M": 8,
|
88 |
+
"num_warps": 8,
|
89 |
+
"num_stages": 2,
|
90 |
+
"waves_per_eu": 0
|
91 |
+
},
|
92 |
+
"128": {
|
93 |
+
"BLOCK_SIZE_M": 16,
|
94 |
+
"BLOCK_SIZE_N": 128,
|
95 |
+
"BLOCK_SIZE_K": 128,
|
96 |
+
"GROUP_SIZE_M": 4,
|
97 |
+
"num_warps": 4,
|
98 |
+
"num_stages": 2,
|
99 |
+
"waves_per_eu": 0
|
100 |
+
},
|
101 |
+
"256": {
|
102 |
+
"BLOCK_SIZE_M": 16,
|
103 |
+
"BLOCK_SIZE_N": 128,
|
104 |
+
"BLOCK_SIZE_K": 128,
|
105 |
+
"GROUP_SIZE_M": 8,
|
106 |
+
"num_warps": 4,
|
107 |
+
"num_stages": 2,
|
108 |
+
"waves_per_eu": 0
|
109 |
+
},
|
110 |
+
"512": {
|
111 |
+
"BLOCK_SIZE_M": 32,
|
112 |
+
"BLOCK_SIZE_N": 128,
|
113 |
+
"BLOCK_SIZE_K": 128,
|
114 |
+
"GROUP_SIZE_M": 8,
|
115 |
+
"num_warps": 4,
|
116 |
+
"num_stages": 2,
|
117 |
+
"waves_per_eu": 0
|
118 |
+
},
|
119 |
+
"1024": {
|
120 |
+
"BLOCK_SIZE_M": 64,
|
121 |
+
"BLOCK_SIZE_N": 128,
|
122 |
+
"BLOCK_SIZE_K": 128,
|
123 |
+
"GROUP_SIZE_M": 8,
|
124 |
+
"num_warps": 2,
|
125 |
+
"num_stages": 2,
|
126 |
+
"waves_per_eu": 0
|
127 |
+
},
|
128 |
+
"1536": {
|
129 |
+
"BLOCK_SIZE_M": 64,
|
130 |
+
"BLOCK_SIZE_N": 128,
|
131 |
+
"BLOCK_SIZE_K": 128,
|
132 |
+
"GROUP_SIZE_M": 4,
|
133 |
+
"num_warps": 2,
|
134 |
+
"num_stages": 2,
|
135 |
+
"waves_per_eu": 0
|
136 |
+
},
|
137 |
+
"2048": {
|
138 |
+
"BLOCK_SIZE_M": 128,
|
139 |
+
"BLOCK_SIZE_N": 256,
|
140 |
+
"BLOCK_SIZE_K": 128,
|
141 |
+
"GROUP_SIZE_M": 8,
|
142 |
+
"num_warps": 4,
|
143 |
+
"num_stages": 2,
|
144 |
+
"waves_per_eu": 0
|
145 |
+
},
|
146 |
+
"3072": {
|
147 |
+
"BLOCK_SIZE_M": 128,
|
148 |
+
"BLOCK_SIZE_N": 256,
|
149 |
+
"BLOCK_SIZE_K": 128,
|
150 |
+
"GROUP_SIZE_M": 8,
|
151 |
+
"num_warps": 4,
|
152 |
+
"num_stages": 2,
|
153 |
+
"waves_per_eu": 0
|
154 |
+
},
|
155 |
+
"4096": {
|
156 |
+
"BLOCK_SIZE_M": 128,
|
157 |
+
"BLOCK_SIZE_N": 256,
|
158 |
+
"BLOCK_SIZE_K": 128,
|
159 |
+
"GROUP_SIZE_M": 4,
|
160 |
+
"num_warps": 4,
|
161 |
+
"num_stages": 2,
|
162 |
+
"waves_per_eu": 0
|
163 |
+
}
|
164 |
+
}
|
build/torch25-cxx11-cu118-x86_64-linux/moe/configs/E=256,N=256,device_name=AMD_Instinct_MI325X,dtype=fp8_w8a8,block_shape=[128,128].json
ADDED
@@ -0,0 +1,200 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"1": {
|
3 |
+
"BLOCK_SIZE_M": 16,
|
4 |
+
"BLOCK_SIZE_N": 16,
|
5 |
+
"BLOCK_SIZE_K": 256,
|
6 |
+
"GROUP_SIZE_M": 1,
|
7 |
+
"num_warps": 4,
|
8 |
+
"num_stages": 2,
|
9 |
+
"waves_per_eu": 0,
|
10 |
+
"matrix_instr_nonkdim": 16,
|
11 |
+
"kpack": 1
|
12 |
+
},
|
13 |
+
"2": {
|
14 |
+
"BLOCK_SIZE_M": 16,
|
15 |
+
"BLOCK_SIZE_N": 64,
|
16 |
+
"BLOCK_SIZE_K": 128,
|
17 |
+
"GROUP_SIZE_M": 1,
|
18 |
+
"num_warps": 4,
|
19 |
+
"num_stages": 2,
|
20 |
+
"waves_per_eu": 0,
|
21 |
+
"matrix_instr_nonkdim": 16,
|
22 |
+
"kpack": 1
|
23 |
+
},
|
24 |
+
"4": {
|
25 |
+
"BLOCK_SIZE_M": 64,
|
26 |
+
"BLOCK_SIZE_N": 64,
|
27 |
+
"BLOCK_SIZE_K": 128,
|
28 |
+
"GROUP_SIZE_M": 1,
|
29 |
+
"num_warps": 8,
|
30 |
+
"num_stages": 2,
|
31 |
+
"waves_per_eu": 0,
|
32 |
+
"matrix_instr_nonkdim": 16,
|
33 |
+
"kpack": 1
|
34 |
+
},
|
35 |
+
"8": {
|
36 |
+
"BLOCK_SIZE_M": 16,
|
37 |
+
"BLOCK_SIZE_N": 64,
|
38 |
+
"BLOCK_SIZE_K": 128,
|
39 |
+
"GROUP_SIZE_M": 1,
|
40 |
+
"num_warps": 4,
|
41 |
+
"num_stages": 2,
|
42 |
+
"waves_per_eu": 0,
|
43 |
+
"matrix_instr_nonkdim": 16,
|
44 |
+
"kpack": 1
|
45 |
+
},
|
46 |
+
"16": {
|
47 |
+
"BLOCK_SIZE_M": 16,
|
48 |
+
"BLOCK_SIZE_N": 64,
|
49 |
+
"BLOCK_SIZE_K": 128,
|
50 |
+
"GROUP_SIZE_M": 1,
|
51 |
+
"num_warps": 4,
|
52 |
+
"num_stages": 2,
|
53 |
+
"waves_per_eu": 0,
|
54 |
+
"matrix_instr_nonkdim": 16,
|
55 |
+
"kpack": 1
|
56 |
+
},
|
57 |
+
"24": {
|
58 |
+
"BLOCK_SIZE_M": 32,
|
59 |
+
"BLOCK_SIZE_N": 128,
|
60 |
+
"BLOCK_SIZE_K": 64,
|
61 |
+
"GROUP_SIZE_M": 1,
|
62 |
+
"num_warps": 8,
|
63 |
+
"num_stages": 2,
|
64 |
+
"waves_per_eu": 0,
|
65 |
+
"matrix_instr_nonkdim": 16,
|
66 |
+
"kpack": 2
|
67 |
+
},
|
68 |
+
"32": {
|
69 |
+
"BLOCK_SIZE_M": 16,
|
70 |
+
"BLOCK_SIZE_N": 64,
|
71 |
+
"BLOCK_SIZE_K": 128,
|
72 |
+
"GROUP_SIZE_M": 1,
|
73 |
+
"num_warps": 4,
|
74 |
+
"num_stages": 2,
|
75 |
+
"waves_per_eu": 0,
|
76 |
+
"matrix_instr_nonkdim": 16,
|
77 |
+
"kpack": 2
|
78 |
+
},
|
79 |
+
"48": {
|
80 |
+
"BLOCK_SIZE_M": 16,
|
81 |
+
"BLOCK_SIZE_N": 64,
|
82 |
+
"BLOCK_SIZE_K": 128,
|
83 |
+
"GROUP_SIZE_M": 1,
|
84 |
+
"num_warps": 2,
|
85 |
+
"num_stages": 2,
|
86 |
+
"waves_per_eu": 0,
|
87 |
+
"matrix_instr_nonkdim": 16,
|
88 |
+
"kpack": 2
|
89 |
+
},
|
90 |
+
"64": {
|
91 |
+
"BLOCK_SIZE_M": 16,
|
92 |
+
"BLOCK_SIZE_N": 64,
|
93 |
+
"BLOCK_SIZE_K": 128,
|
94 |
+
"GROUP_SIZE_M": 1,
|
95 |
+
"num_warps": 2,
|
96 |
+
"num_stages": 2,
|
97 |
+
"waves_per_eu": 0,
|
98 |
+
"matrix_instr_nonkdim": 16,
|
99 |
+
"kpack": 2
|
100 |
+
},
|
101 |
+
"96": {
|
102 |
+
"BLOCK_SIZE_M": 16,
|
103 |
+
"BLOCK_SIZE_N": 32,
|
104 |
+
"BLOCK_SIZE_K": 128,
|
105 |
+
"GROUP_SIZE_M": 4,
|
106 |
+
"num_warps": 2,
|
107 |
+
"num_stages": 2,
|
108 |
+
"waves_per_eu": 0,
|
109 |
+
"matrix_instr_nonkdim": 16,
|
110 |
+
"kpack": 2
|
111 |
+
},
|
112 |
+
"128": {
|
113 |
+
"BLOCK_SIZE_M": 16,
|
114 |
+
"BLOCK_SIZE_N": 32,
|
115 |
+
"BLOCK_SIZE_K": 128,
|
116 |
+
"GROUP_SIZE_M": 4,
|
117 |
+
"num_warps": 1,
|
118 |
+
"num_stages": 2,
|
119 |
+
"waves_per_eu": 0,
|
120 |
+
"matrix_instr_nonkdim": 16,
|
121 |
+
"kpack": 2
|
122 |
+
},
|
123 |
+
"256": {
|
124 |
+
"BLOCK_SIZE_M": 16,
|
125 |
+
"BLOCK_SIZE_N": 32,
|
126 |
+
"BLOCK_SIZE_K": 128,
|
127 |
+
"GROUP_SIZE_M": 1,
|
128 |
+
"num_warps": 2,
|
129 |
+
"num_stages": 2,
|
130 |
+
"waves_per_eu": 0,
|
131 |
+
"matrix_instr_nonkdim": 16,
|
132 |
+
"kpack": 2
|
133 |
+
},
|
134 |
+
"512": {
|
135 |
+
"BLOCK_SIZE_M": 32,
|
136 |
+
"BLOCK_SIZE_N": 64,
|
137 |
+
"BLOCK_SIZE_K": 128,
|
138 |
+
"GROUP_SIZE_M": 8,
|
139 |
+
"num_warps": 8,
|
140 |
+
"num_stages": 2,
|
141 |
+
"waves_per_eu": 0,
|
142 |
+
"matrix_instr_nonkdim": 16,
|
143 |
+
"kpack": 2
|
144 |
+
},
|
145 |
+
"1024": {
|
146 |
+
"BLOCK_SIZE_M": 64,
|
147 |
+
"BLOCK_SIZE_N": 128,
|
148 |
+
"BLOCK_SIZE_K": 128,
|
149 |
+
"GROUP_SIZE_M": 8,
|
150 |
+
"num_warps": 8,
|
151 |
+
"num_stages": 2,
|
152 |
+
"waves_per_eu": 0,
|
153 |
+
"matrix_instr_nonkdim": 16,
|
154 |
+
"kpack": 2
|
155 |
+
},
|
156 |
+
"1536": {
|
157 |
+
"BLOCK_SIZE_M": 64,
|
158 |
+
"BLOCK_SIZE_N": 128,
|
159 |
+
"BLOCK_SIZE_K": 128,
|
160 |
+
"GROUP_SIZE_M": 8,
|
161 |
+
"num_warps": 8,
|
162 |
+
"num_stages": 2,
|
163 |
+
"waves_per_eu": 0,
|
164 |
+
"matrix_instr_nonkdim": 16,
|
165 |
+
"kpack": 2
|
166 |
+
},
|
167 |
+
"2048": {
|
168 |
+
"BLOCK_SIZE_M": 128,
|
169 |
+
"BLOCK_SIZE_N": 128,
|
170 |
+
"BLOCK_SIZE_K": 64,
|
171 |
+
"GROUP_SIZE_M": 8,
|
172 |
+
"num_warps": 8,
|
173 |
+
"num_stages": 2,
|
174 |
+
"waves_per_eu": 0,
|
175 |
+
"matrix_instr_nonkdim": 16,
|
176 |
+
"kpack": 2
|
177 |
+
},
|
178 |
+
"3072": {
|
179 |
+
"BLOCK_SIZE_M": 128,
|
180 |
+
"BLOCK_SIZE_N": 128,
|
181 |
+
"BLOCK_SIZE_K": 64,
|
182 |
+
"GROUP_SIZE_M": 8,
|
183 |
+
"num_warps": 8,
|
184 |
+
"num_stages": 2,
|
185 |
+
"waves_per_eu": 0,
|
186 |
+
"matrix_instr_nonkdim": 16,
|
187 |
+
"kpack": 2
|
188 |
+
},
|
189 |
+
"4096": {
|
190 |
+
"BLOCK_SIZE_M": 128,
|
191 |
+
"BLOCK_SIZE_N": 128,
|
192 |
+
"BLOCK_SIZE_K": 64,
|
193 |
+
"GROUP_SIZE_M": 8,
|
194 |
+
"num_warps": 8,
|
195 |
+
"num_stages": 2,
|
196 |
+
"waves_per_eu": 0,
|
197 |
+
"matrix_instr_nonkdim": 16,
|
198 |
+
"kpack": 2
|
199 |
+
}
|
200 |
+
}
|
build/torch25-cxx11-cu118-x86_64-linux/moe/configs/E=256,N=256,device_name=AMD_Instinct_MI325_OAM,dtype=fp8_w8a8,block_shape=[128,128].json
ADDED
@@ -0,0 +1,200 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"1": {
|
3 |
+
"BLOCK_SIZE_M": 16,
|
4 |
+
"BLOCK_SIZE_N": 16,
|
5 |
+
"BLOCK_SIZE_K": 256,
|
6 |
+
"GROUP_SIZE_M": 1,
|
7 |
+
"num_warps": 4,
|
8 |
+
"num_stages": 2,
|
9 |
+
"waves_per_eu": 0,
|
10 |
+
"matrix_instr_nonkdim": 16,
|
11 |
+
"kpack": 1
|
12 |
+
},
|
13 |
+
"2": {
|
14 |
+
"BLOCK_SIZE_M": 16,
|
15 |
+
"BLOCK_SIZE_N": 64,
|
16 |
+
"BLOCK_SIZE_K": 128,
|
17 |
+
"GROUP_SIZE_M": 1,
|
18 |
+
"num_warps": 4,
|
19 |
+
"num_stages": 2,
|
20 |
+
"waves_per_eu": 0,
|
21 |
+
"matrix_instr_nonkdim": 16,
|
22 |
+
"kpack": 1
|
23 |
+
},
|
24 |
+
"4": {
|
25 |
+
"BLOCK_SIZE_M": 64,
|
26 |
+
"BLOCK_SIZE_N": 64,
|
27 |
+
"BLOCK_SIZE_K": 128,
|
28 |
+
"GROUP_SIZE_M": 1,
|
29 |
+
"num_warps": 8,
|
30 |
+
"num_stages": 2,
|
31 |
+
"waves_per_eu": 0,
|
32 |
+
"matrix_instr_nonkdim": 16,
|
33 |
+
"kpack": 1
|
34 |
+
},
|
35 |
+
"8": {
|
36 |
+
"BLOCK_SIZE_M": 16,
|
37 |
+
"BLOCK_SIZE_N": 64,
|
38 |
+
"BLOCK_SIZE_K": 128,
|
39 |
+
"GROUP_SIZE_M": 1,
|
40 |
+
"num_warps": 4,
|
41 |
+
"num_stages": 2,
|
42 |
+
"waves_per_eu": 0,
|
43 |
+
"matrix_instr_nonkdim": 16,
|
44 |
+
"kpack": 1
|
45 |
+
},
|
46 |
+
"16": {
|
47 |
+
"BLOCK_SIZE_M": 16,
|
48 |
+
"BLOCK_SIZE_N": 64,
|
49 |
+
"BLOCK_SIZE_K": 128,
|
50 |
+
"GROUP_SIZE_M": 1,
|
51 |
+
"num_warps": 4,
|
52 |
+
"num_stages": 2,
|
53 |
+
"waves_per_eu": 0,
|
54 |
+
"matrix_instr_nonkdim": 16,
|
55 |
+
"kpack": 1
|
56 |
+
},
|
57 |
+
"24": {
|
58 |
+
"BLOCK_SIZE_M": 32,
|
59 |
+
"BLOCK_SIZE_N": 128,
|
60 |
+
"BLOCK_SIZE_K": 64,
|
61 |
+
"GROUP_SIZE_M": 1,
|
62 |
+
"num_warps": 8,
|
63 |
+
"num_stages": 2,
|
64 |
+
"waves_per_eu": 0,
|
65 |
+
"matrix_instr_nonkdim": 16,
|
66 |
+
"kpack": 2
|
67 |
+
},
|
68 |
+
"32": {
|
69 |
+
"BLOCK_SIZE_M": 16,
|
70 |
+
"BLOCK_SIZE_N": 64,
|
71 |
+
"BLOCK_SIZE_K": 128,
|
72 |
+
"GROUP_SIZE_M": 1,
|
73 |
+
"num_warps": 4,
|
74 |
+
"num_stages": 2,
|
75 |
+
"waves_per_eu": 0,
|
76 |
+
"matrix_instr_nonkdim": 16,
|
77 |
+
"kpack": 2
|
78 |
+
},
|
79 |
+
"48": {
|
80 |
+
"BLOCK_SIZE_M": 16,
|
81 |
+
"BLOCK_SIZE_N": 64,
|
82 |
+
"BLOCK_SIZE_K": 128,
|
83 |
+
"GROUP_SIZE_M": 1,
|
84 |
+
"num_warps": 2,
|
85 |
+
"num_stages": 2,
|
86 |
+
"waves_per_eu": 0,
|
87 |
+
"matrix_instr_nonkdim": 16,
|
88 |
+
"kpack": 2
|
89 |
+
},
|
90 |
+
"64": {
|
91 |
+
"BLOCK_SIZE_M": 16,
|
92 |
+
"BLOCK_SIZE_N": 64,
|
93 |
+
"BLOCK_SIZE_K": 128,
|
94 |
+
"GROUP_SIZE_M": 1,
|
95 |
+
"num_warps": 2,
|
96 |
+
"num_stages": 2,
|
97 |
+
"waves_per_eu": 0,
|
98 |
+
"matrix_instr_nonkdim": 16,
|
99 |
+
"kpack": 2
|
100 |
+
},
|
101 |
+
"96": {
|
102 |
+
"BLOCK_SIZE_M": 16,
|
103 |
+
"BLOCK_SIZE_N": 32,
|
104 |
+
"BLOCK_SIZE_K": 128,
|
105 |
+
"GROUP_SIZE_M": 4,
|
106 |
+
"num_warps": 2,
|
107 |
+
"num_stages": 2,
|
108 |
+
"waves_per_eu": 0,
|
109 |
+
"matrix_instr_nonkdim": 16,
|
110 |
+
"kpack": 2
|
111 |
+
},
|
112 |
+
"128": {
|
113 |
+
"BLOCK_SIZE_M": 16,
|
114 |
+
"BLOCK_SIZE_N": 32,
|
115 |
+
"BLOCK_SIZE_K": 128,
|
116 |
+
"GROUP_SIZE_M": 4,
|
117 |
+
"num_warps": 1,
|
118 |
+
"num_stages": 2,
|
119 |
+
"waves_per_eu": 0,
|
120 |
+
"matrix_instr_nonkdim": 16,
|
121 |
+
"kpack": 2
|
122 |
+
},
|
123 |
+
"256": {
|
124 |
+
"BLOCK_SIZE_M": 16,
|
125 |
+
"BLOCK_SIZE_N": 32,
|
126 |
+
"BLOCK_SIZE_K": 128,
|
127 |
+
"GROUP_SIZE_M": 1,
|
128 |
+
"num_warps": 2,
|
129 |
+
"num_stages": 2,
|
130 |
+
"waves_per_eu": 0,
|
131 |
+
"matrix_instr_nonkdim": 16,
|
132 |
+
"kpack": 2
|
133 |
+
},
|
134 |
+
"512": {
|
135 |
+
"BLOCK_SIZE_M": 32,
|
136 |
+
"BLOCK_SIZE_N": 64,
|
137 |
+
"BLOCK_SIZE_K": 128,
|
138 |
+
"GROUP_SIZE_M": 8,
|
139 |
+
"num_warps": 8,
|
140 |
+
"num_stages": 2,
|
141 |
+
"waves_per_eu": 0,
|
142 |
+
"matrix_instr_nonkdim": 16,
|
143 |
+
"kpack": 2
|
144 |
+
},
|
145 |
+
"1024": {
|
146 |
+
"BLOCK_SIZE_M": 64,
|
147 |
+
"BLOCK_SIZE_N": 128,
|
148 |
+
"BLOCK_SIZE_K": 128,
|
149 |
+
"GROUP_SIZE_M": 8,
|
150 |
+
"num_warps": 8,
|
151 |
+
"num_stages": 2,
|
152 |
+
"waves_per_eu": 0,
|
153 |
+
"matrix_instr_nonkdim": 16,
|
154 |
+
"kpack": 2
|
155 |
+
},
|
156 |
+
"1536": {
|
157 |
+
"BLOCK_SIZE_M": 64,
|
158 |
+
"BLOCK_SIZE_N": 128,
|
159 |
+
"BLOCK_SIZE_K": 128,
|
160 |
+
"GROUP_SIZE_M": 8,
|
161 |
+
"num_warps": 8,
|
162 |
+
"num_stages": 2,
|
163 |
+
"waves_per_eu": 0,
|
164 |
+
"matrix_instr_nonkdim": 16,
|
165 |
+
"kpack": 2
|
166 |
+
},
|
167 |
+
"2048": {
|
168 |
+
"BLOCK_SIZE_M": 128,
|
169 |
+
"BLOCK_SIZE_N": 128,
|
170 |
+
"BLOCK_SIZE_K": 64,
|
171 |
+
"GROUP_SIZE_M": 8,
|
172 |
+
"num_warps": 8,
|
173 |
+
"num_stages": 2,
|
174 |
+
"waves_per_eu": 0,
|
175 |
+
"matrix_instr_nonkdim": 16,
|
176 |
+
"kpack": 2
|
177 |
+
},
|
178 |
+
"3072": {
|
179 |
+
"BLOCK_SIZE_M": 128,
|
180 |
+
"BLOCK_SIZE_N": 128,
|
181 |
+
"BLOCK_SIZE_K": 64,
|
182 |
+
"GROUP_SIZE_M": 8,
|
183 |
+
"num_warps": 8,
|
184 |
+
"num_stages": 2,
|
185 |
+
"waves_per_eu": 0,
|
186 |
+
"matrix_instr_nonkdim": 16,
|
187 |
+
"kpack": 2
|
188 |
+
},
|
189 |
+
"4096": {
|
190 |
+
"BLOCK_SIZE_M": 128,
|
191 |
+
"BLOCK_SIZE_N": 128,
|
192 |
+
"BLOCK_SIZE_K": 64,
|
193 |
+
"GROUP_SIZE_M": 8,
|
194 |
+
"num_warps": 8,
|
195 |
+
"num_stages": 2,
|
196 |
+
"waves_per_eu": 0,
|
197 |
+
"matrix_instr_nonkdim": 16,
|
198 |
+
"kpack": 2
|
199 |
+
}
|
200 |
+
}
|
build/torch25-cxx11-cu118-x86_64-linux/moe/configs/E=256,N=256,device_name=NVIDIA_B200,dtype=fp8_w8a8,block_shape=[128,128].json
ADDED
@@ -0,0 +1,146 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"1": {
|
3 |
+
"BLOCK_SIZE_M": 16,
|
4 |
+
"BLOCK_SIZE_N": 64,
|
5 |
+
"BLOCK_SIZE_K": 128,
|
6 |
+
"GROUP_SIZE_M": 32,
|
7 |
+
"num_warps": 4,
|
8 |
+
"num_stages": 4
|
9 |
+
},
|
10 |
+
"2": {
|
11 |
+
"BLOCK_SIZE_M": 16,
|
12 |
+
"BLOCK_SIZE_N": 128,
|
13 |
+
"BLOCK_SIZE_K": 128,
|
14 |
+
"GROUP_SIZE_M": 16,
|
15 |
+
"num_warps": 4,
|
16 |
+
"num_stages": 4
|
17 |
+
},
|
18 |
+
"4": {
|
19 |
+
"BLOCK_SIZE_M": 16,
|
20 |
+
"BLOCK_SIZE_N": 256,
|
21 |
+
"BLOCK_SIZE_K": 128,
|
22 |
+
"GROUP_SIZE_M": 64,
|
23 |
+
"num_warps": 8,
|
24 |
+
"num_stages": 4
|
25 |
+
},
|
26 |
+
"8": {
|
27 |
+
"BLOCK_SIZE_M": 16,
|
28 |
+
"BLOCK_SIZE_N": 256,
|
29 |
+
"BLOCK_SIZE_K": 128,
|
30 |
+
"GROUP_SIZE_M": 32,
|
31 |
+
"num_warps": 8,
|
32 |
+
"num_stages": 4
|
33 |
+
},
|
34 |
+
"16": {
|
35 |
+
"BLOCK_SIZE_M": 16,
|
36 |
+
"BLOCK_SIZE_N": 128,
|
37 |
+
"BLOCK_SIZE_K": 128,
|
38 |
+
"GROUP_SIZE_M": 64,
|
39 |
+
"num_warps": 4,
|
40 |
+
"num_stages": 5
|
41 |
+
},
|
42 |
+
"24": {
|
43 |
+
"BLOCK_SIZE_M": 16,
|
44 |
+
"BLOCK_SIZE_N": 128,
|
45 |
+
"BLOCK_SIZE_K": 128,
|
46 |
+
"GROUP_SIZE_M": 64,
|
47 |
+
"num_warps": 4,
|
48 |
+
"num_stages": 3
|
49 |
+
},
|
50 |
+
"32": {
|
51 |
+
"BLOCK_SIZE_M": 16,
|
52 |
+
"BLOCK_SIZE_N": 256,
|
53 |
+
"BLOCK_SIZE_K": 128,
|
54 |
+
"GROUP_SIZE_M": 1,
|
55 |
+
"num_warps": 4,
|
56 |
+
"num_stages": 3
|
57 |
+
},
|
58 |
+
"48": {
|
59 |
+
"BLOCK_SIZE_M": 16,
|
60 |
+
"BLOCK_SIZE_N": 256,
|
61 |
+
"BLOCK_SIZE_K": 128,
|
62 |
+
"GROUP_SIZE_M": 1,
|
63 |
+
"num_warps": 4,
|
64 |
+
"num_stages": 3
|
65 |
+
},
|
66 |
+
"64": {
|
67 |
+
"BLOCK_SIZE_M": 16,
|
68 |
+
"BLOCK_SIZE_N": 256,
|
69 |
+
"BLOCK_SIZE_K": 128,
|
70 |
+
"GROUP_SIZE_M": 32,
|
71 |
+
"num_warps": 4,
|
72 |
+
"num_stages": 3
|
73 |
+
},
|
74 |
+
"96": {
|
75 |
+
"BLOCK_SIZE_M": 16,
|
76 |
+
"BLOCK_SIZE_N": 128,
|
77 |
+
"BLOCK_SIZE_K": 64,
|
78 |
+
"GROUP_SIZE_M": 1,
|
79 |
+
"num_warps": 4,
|
80 |
+
"num_stages": 4
|
81 |
+
},
|
82 |
+
"128": {
|
83 |
+
"BLOCK_SIZE_M": 16,
|
84 |
+
"BLOCK_SIZE_N": 128,
|
85 |
+
"BLOCK_SIZE_K": 64,
|
86 |
+
"GROUP_SIZE_M": 1,
|
87 |
+
"num_warps": 4,
|
88 |
+
"num_stages": 4
|
89 |
+
},
|
90 |
+
"256": {
|
91 |
+
"BLOCK_SIZE_M": 16,
|
92 |
+
"BLOCK_SIZE_N": 128,
|
93 |
+
"BLOCK_SIZE_K": 64,
|
94 |
+
"GROUP_SIZE_M": 1,
|
95 |
+
"num_warps": 4,
|
96 |
+
"num_stages": 4
|
97 |
+
},
|
98 |
+
"512": {
|
99 |
+
"BLOCK_SIZE_M": 16,
|
100 |
+
"BLOCK_SIZE_N": 256,
|
101 |
+
"BLOCK_SIZE_K": 128,
|
102 |
+
"GROUP_SIZE_M": 32,
|
103 |
+
"num_warps": 4,
|
104 |
+
"num_stages": 3
|
105 |
+
},
|
106 |
+
"1024": {
|
107 |
+
"BLOCK_SIZE_M": 16,
|
108 |
+
"BLOCK_SIZE_N": 256,
|
109 |
+
"BLOCK_SIZE_K": 128,
|
110 |
+
"GROUP_SIZE_M": 1,
|
111 |
+
"num_warps": 4,
|
112 |
+
"num_stages": 3
|
113 |
+
},
|
114 |
+
"1536": {
|
115 |
+
"BLOCK_SIZE_M": 16,
|
116 |
+
"BLOCK_SIZE_N": 256,
|
117 |
+
"BLOCK_SIZE_K": 128,
|
118 |
+
"GROUP_SIZE_M": 1,
|
119 |
+
"num_warps": 4,
|
120 |
+
"num_stages": 3
|
121 |
+
},
|
122 |
+
"2048": {
|
123 |
+
"BLOCK_SIZE_M": 16,
|
124 |
+
"BLOCK_SIZE_N": 256,
|
125 |
+
"BLOCK_SIZE_K": 128,
|
126 |
+
"GROUP_SIZE_M": 1,
|
127 |
+
"num_warps": 4,
|
128 |
+
"num_stages": 3
|
129 |
+
},
|
130 |
+
"3072": {
|
131 |
+
"BLOCK_SIZE_M": 32,
|
132 |
+
"BLOCK_SIZE_N": 256,
|
133 |
+
"BLOCK_SIZE_K": 128,
|
134 |
+
"GROUP_SIZE_M": 16,
|
135 |
+
"num_warps": 4,
|
136 |
+
"num_stages": 3
|
137 |
+
},
|
138 |
+
"4096": {
|
139 |
+
"BLOCK_SIZE_M": 32,
|
140 |
+
"BLOCK_SIZE_N": 256,
|
141 |
+
"BLOCK_SIZE_K": 128,
|
142 |
+
"GROUP_SIZE_M": 16,
|
143 |
+
"num_warps": 4,
|
144 |
+
"num_stages": 3
|
145 |
+
}
|
146 |
+
}
|
build/torch25-cxx11-cu118-x86_64-linux/moe/configs/E=256,N=256,device_name=NVIDIA_H20,dtype=int8_w8a8,block_shape=[128,128].json
ADDED
@@ -0,0 +1,146 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"1": {
|
3 |
+
"BLOCK_SIZE_M": 16,
|
4 |
+
"BLOCK_SIZE_N": 128,
|
5 |
+
"BLOCK_SIZE_K": 128,
|
6 |
+
"GROUP_SIZE_M": 1,
|
7 |
+
"num_warps": 4,
|
8 |
+
"num_stages": 3
|
9 |
+
},
|
10 |
+
"2": {
|
11 |
+
"BLOCK_SIZE_M": 16,
|
12 |
+
"BLOCK_SIZE_N": 128,
|
13 |
+
"BLOCK_SIZE_K": 128,
|
14 |
+
"GROUP_SIZE_M": 32,
|
15 |
+
"num_warps": 4,
|
16 |
+
"num_stages": 3
|
17 |
+
},
|
18 |
+
"4": {
|
19 |
+
"BLOCK_SIZE_M": 16,
|
20 |
+
"BLOCK_SIZE_N": 128,
|
21 |
+
"BLOCK_SIZE_K": 128,
|
22 |
+
"GROUP_SIZE_M": 1,
|
23 |
+
"num_warps": 4,
|
24 |
+
"num_stages": 3
|
25 |
+
},
|
26 |
+
"8": {
|
27 |
+
"BLOCK_SIZE_M": 16,
|
28 |
+
"BLOCK_SIZE_N": 128,
|
29 |
+
"BLOCK_SIZE_K": 128,
|
30 |
+
"GROUP_SIZE_M": 32,
|
31 |
+
"num_warps": 4,
|
32 |
+
"num_stages": 3
|
33 |
+
},
|
34 |
+
"16": {
|
35 |
+
"BLOCK_SIZE_M": 16,
|
36 |
+
"BLOCK_SIZE_N": 128,
|
37 |
+
"BLOCK_SIZE_K": 128,
|
38 |
+
"GROUP_SIZE_M": 1,
|
39 |
+
"num_warps": 4,
|
40 |
+
"num_stages": 3
|
41 |
+
},
|
42 |
+
"24": {
|
43 |
+
"BLOCK_SIZE_M": 16,
|
44 |
+
"BLOCK_SIZE_N": 128,
|
45 |
+
"BLOCK_SIZE_K": 128,
|
46 |
+
"GROUP_SIZE_M": 64,
|
47 |
+
"num_warps": 4,
|
48 |
+
"num_stages": 3
|
49 |
+
},
|
50 |
+
"32": {
|
51 |
+
"BLOCK_SIZE_M": 16,
|
52 |
+
"BLOCK_SIZE_N": 128,
|
53 |
+
"BLOCK_SIZE_K": 128,
|
54 |
+
"GROUP_SIZE_M": 64,
|
55 |
+
"num_warps": 4,
|
56 |
+
"num_stages": 3
|
57 |
+
},
|
58 |
+
"48": {
|
59 |
+
"BLOCK_SIZE_M": 16,
|
60 |
+
"BLOCK_SIZE_N": 128,
|
61 |
+
"BLOCK_SIZE_K": 128,
|
62 |
+
"GROUP_SIZE_M": 32,
|
63 |
+
"num_warps": 4,
|
64 |
+
"num_stages": 3
|
65 |
+
},
|
66 |
+
"64": {
|
67 |
+
"BLOCK_SIZE_M": 16,
|
68 |
+
"BLOCK_SIZE_N": 128,
|
69 |
+
"BLOCK_SIZE_K": 128,
|
70 |
+
"GROUP_SIZE_M": 1,
|
71 |
+
"num_warps": 4,
|
72 |
+
"num_stages": 3
|
73 |
+
},
|
74 |
+
"96": {
|
75 |
+
"BLOCK_SIZE_M": 16,
|
76 |
+
"BLOCK_SIZE_N": 128,
|
77 |
+
"BLOCK_SIZE_K": 128,
|
78 |
+
"GROUP_SIZE_M": 64,
|
79 |
+
"num_warps": 4,
|
80 |
+
"num_stages": 3
|
81 |
+
},
|
82 |
+
"128": {
|
83 |
+
"BLOCK_SIZE_M": 16,
|
84 |
+
"BLOCK_SIZE_N": 128,
|
85 |
+
"BLOCK_SIZE_K": 128,
|
86 |
+
"GROUP_SIZE_M": 32,
|
87 |
+
"num_warps": 4,
|
88 |
+
"num_stages": 3
|
89 |
+
},
|
90 |
+
"256": {
|
91 |
+
"BLOCK_SIZE_M": 16,
|
92 |
+
"BLOCK_SIZE_N": 128,
|
93 |
+
"BLOCK_SIZE_K": 128,
|
94 |
+
"GROUP_SIZE_M": 16,
|
95 |
+
"num_warps": 4,
|
96 |
+
"num_stages": 3
|
97 |
+
},
|
98 |
+
"512": {
|
99 |
+
"BLOCK_SIZE_M": 16,
|
100 |
+
"BLOCK_SIZE_N": 128,
|
101 |
+
"BLOCK_SIZE_K": 128,
|
102 |
+
"GROUP_SIZE_M": 16,
|
103 |
+
"num_warps": 4,
|
104 |
+
"num_stages": 3
|
105 |
+
},
|
106 |
+
"1024": {
|
107 |
+
"BLOCK_SIZE_M": 16,
|
108 |
+
"BLOCK_SIZE_N": 128,
|
109 |
+
"BLOCK_SIZE_K": 128,
|
110 |
+
"GROUP_SIZE_M": 16,
|
111 |
+
"num_warps": 4,
|
112 |
+
"num_stages": 3
|
113 |
+
},
|
114 |
+
"1536": {
|
115 |
+
"BLOCK_SIZE_M": 64,
|
116 |
+
"BLOCK_SIZE_N": 64,
|
117 |
+
"BLOCK_SIZE_K": 128,
|
118 |
+
"GROUP_SIZE_M": 16,
|
119 |
+
"num_warps": 4,
|
120 |
+
"num_stages": 3
|
121 |
+
},
|
122 |
+
"2048": {
|
123 |
+
"BLOCK_SIZE_M": 16,
|
124 |
+
"BLOCK_SIZE_N": 128,
|
125 |
+
"BLOCK_SIZE_K": 128,
|
126 |
+
"GROUP_SIZE_M": 32,
|
127 |
+
"num_warps": 4,
|
128 |
+
"num_stages": 3
|
129 |
+
},
|
130 |
+
"3072": {
|
131 |
+
"BLOCK_SIZE_M": 128,
|
132 |
+
"BLOCK_SIZE_N": 64,
|
133 |
+
"BLOCK_SIZE_K": 128,
|
134 |
+
"GROUP_SIZE_M": 16,
|
135 |
+
"num_warps": 4,
|
136 |
+
"num_stages": 4
|
137 |
+
},
|
138 |
+
"4096": {
|
139 |
+
"BLOCK_SIZE_M": 64,
|
140 |
+
"BLOCK_SIZE_N": 64,
|
141 |
+
"BLOCK_SIZE_K": 128,
|
142 |
+
"GROUP_SIZE_M": 16,
|
143 |
+
"num_warps": 4,
|
144 |
+
"num_stages": 3
|
145 |
+
}
|
146 |
+
}
|
build/torch25-cxx11-cu118-x86_64-linux/moe/configs/E=256,N=256,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[128,128].json
ADDED
@@ -0,0 +1,146 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"1": {
|
3 |
+
"BLOCK_SIZE_M": 16,
|
4 |
+
"BLOCK_SIZE_N": 64,
|
5 |
+
"BLOCK_SIZE_K": 128,
|
6 |
+
"GROUP_SIZE_M": 1,
|
7 |
+
"num_warps": 4,
|
8 |
+
"num_stages": 5
|
9 |
+
},
|
10 |
+
"2": {
|
11 |
+
"BLOCK_SIZE_M": 16,
|
12 |
+
"BLOCK_SIZE_N": 128,
|
13 |
+
"BLOCK_SIZE_K": 128,
|
14 |
+
"GROUP_SIZE_M": 16,
|
15 |
+
"num_warps": 4,
|
16 |
+
"num_stages": 3
|
17 |
+
},
|
18 |
+
"4": {
|
19 |
+
"BLOCK_SIZE_M": 64,
|
20 |
+
"BLOCK_SIZE_N": 64,
|
21 |
+
"BLOCK_SIZE_K": 128,
|
22 |
+
"GROUP_SIZE_M": 64,
|
23 |
+
"num_warps": 4,
|
24 |
+
"num_stages": 4
|
25 |
+
},
|
26 |
+
"8": {
|
27 |
+
"BLOCK_SIZE_M": 64,
|
28 |
+
"BLOCK_SIZE_N": 128,
|
29 |
+
"BLOCK_SIZE_K": 128,
|
30 |
+
"GROUP_SIZE_M": 64,
|
31 |
+
"num_warps": 4,
|
32 |
+
"num_stages": 3
|
33 |
+
},
|
34 |
+
"16": {
|
35 |
+
"BLOCK_SIZE_M": 16,
|
36 |
+
"BLOCK_SIZE_N": 256,
|
37 |
+
"BLOCK_SIZE_K": 64,
|
38 |
+
"GROUP_SIZE_M": 32,
|
39 |
+
"num_warps": 4,
|
40 |
+
"num_stages": 3
|
41 |
+
},
|
42 |
+
"24": {
|
43 |
+
"BLOCK_SIZE_M": 64,
|
44 |
+
"BLOCK_SIZE_N": 128,
|
45 |
+
"BLOCK_SIZE_K": 128,
|
46 |
+
"GROUP_SIZE_M": 16,
|
47 |
+
"num_warps": 4,
|
48 |
+
"num_stages": 3
|
49 |
+
},
|
50 |
+
"32": {
|
51 |
+
"BLOCK_SIZE_M": 64,
|
52 |
+
"BLOCK_SIZE_N": 128,
|
53 |
+
"BLOCK_SIZE_K": 128,
|
54 |
+
"GROUP_SIZE_M": 1,
|
55 |
+
"num_warps": 4,
|
56 |
+
"num_stages": 3
|
57 |
+
},
|
58 |
+
"48": {
|
59 |
+
"BLOCK_SIZE_M": 64,
|
60 |
+
"BLOCK_SIZE_N": 128,
|
61 |
+
"BLOCK_SIZE_K": 128,
|
62 |
+
"GROUP_SIZE_M": 32,
|
63 |
+
"num_warps": 4,
|
64 |
+
"num_stages": 3
|
65 |
+
},
|
66 |
+
"64": {
|
67 |
+
"BLOCK_SIZE_M": 64,
|
68 |
+
"BLOCK_SIZE_N": 128,
|
69 |
+
"BLOCK_SIZE_K": 128,
|
70 |
+
"GROUP_SIZE_M": 32,
|
71 |
+
"num_warps": 4,
|
72 |
+
"num_stages": 3
|
73 |
+
},
|
74 |
+
"96": {
|
75 |
+
"BLOCK_SIZE_M": 64,
|
76 |
+
"BLOCK_SIZE_N": 128,
|
77 |
+
"BLOCK_SIZE_K": 128,
|
78 |
+
"GROUP_SIZE_M": 32,
|
79 |
+
"num_warps": 4,
|
80 |
+
"num_stages": 3
|
81 |
+
},
|
82 |
+
"128": {
|
83 |
+
"BLOCK_SIZE_M": 64,
|
84 |
+
"BLOCK_SIZE_N": 128,
|
85 |
+
"BLOCK_SIZE_K": 128,
|
86 |
+
"GROUP_SIZE_M": 16,
|
87 |
+
"num_warps": 4,
|
88 |
+
"num_stages": 3
|
89 |
+
},
|
90 |
+
"256": {
|
91 |
+
"BLOCK_SIZE_M": 64,
|
92 |
+
"BLOCK_SIZE_N": 128,
|
93 |
+
"BLOCK_SIZE_K": 128,
|
94 |
+
"GROUP_SIZE_M": 32,
|
95 |
+
"num_warps": 4,
|
96 |
+
"num_stages": 3
|
97 |
+
},
|
98 |
+
"512": {
|
99 |
+
"BLOCK_SIZE_M": 64,
|
100 |
+
"BLOCK_SIZE_N": 128,
|
101 |
+
"BLOCK_SIZE_K": 128,
|
102 |
+
"GROUP_SIZE_M": 32,
|
103 |
+
"num_warps": 4,
|
104 |
+
"num_stages": 3
|
105 |
+
},
|
106 |
+
"1024": {
|
107 |
+
"BLOCK_SIZE_M": 64,
|
108 |
+
"BLOCK_SIZE_N": 128,
|
109 |
+
"BLOCK_SIZE_K": 128,
|
110 |
+
"GROUP_SIZE_M": 32,
|
111 |
+
"num_warps": 4,
|
112 |
+
"num_stages": 3
|
113 |
+
},
|
114 |
+
"1536": {
|
115 |
+
"BLOCK_SIZE_M": 64,
|
116 |
+
"BLOCK_SIZE_N": 128,
|
117 |
+
"BLOCK_SIZE_K": 128,
|
118 |
+
"GROUP_SIZE_M": 32,
|
119 |
+
"num_warps": 4,
|
120 |
+
"num_stages": 3
|
121 |
+
},
|
122 |
+
"2048": {
|
123 |
+
"BLOCK_SIZE_M": 64,
|
124 |
+
"BLOCK_SIZE_N": 128,
|
125 |
+
"BLOCK_SIZE_K": 128,
|
126 |
+
"GROUP_SIZE_M": 64,
|
127 |
+
"num_warps": 4,
|
128 |
+
"num_stages": 3
|
129 |
+
},
|
130 |
+
"3072": {
|
131 |
+
"BLOCK_SIZE_M": 128,
|
132 |
+
"BLOCK_SIZE_N": 64,
|
133 |
+
"BLOCK_SIZE_K": 128,
|
134 |
+
"GROUP_SIZE_M": 16,
|
135 |
+
"num_warps": 4,
|
136 |
+
"num_stages": 4
|
137 |
+
},
|
138 |
+
"4096": {
|
139 |
+
"BLOCK_SIZE_M": 64,
|
140 |
+
"BLOCK_SIZE_N": 128,
|
141 |
+
"BLOCK_SIZE_K": 128,
|
142 |
+
"GROUP_SIZE_M": 64,
|
143 |
+
"num_warps": 4,
|
144 |
+
"num_stages": 3
|
145 |
+
}
|
146 |
+
}
|
build/torch25-cxx11-cu118-x86_64-linux/moe/configs/E=256,N=256,device_name=NVIDIA_L20,dtype=fp8_w8a8,block_shape=[128,128].json
ADDED
@@ -0,0 +1,146 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"1": {
|
3 |
+
"BLOCK_SIZE_M": 16,
|
4 |
+
"BLOCK_SIZE_N": 128,
|
5 |
+
"BLOCK_SIZE_K": 128,
|
6 |
+
"GROUP_SIZE_M": 1,
|
7 |
+
"num_warps": 8,
|
8 |
+
"num_stages": 3
|
9 |
+
},
|
10 |
+
"2": {
|
11 |
+
"BLOCK_SIZE_M": 32,
|
12 |
+
"BLOCK_SIZE_N": 128,
|
13 |
+
"BLOCK_SIZE_K": 128,
|
14 |
+
"GROUP_SIZE_M": 1,
|
15 |
+
"num_warps": 4,
|
16 |
+
"num_stages": 3
|
17 |
+
},
|
18 |
+
"4": {
|
19 |
+
"BLOCK_SIZE_M": 16,
|
20 |
+
"BLOCK_SIZE_N": 128,
|
21 |
+
"BLOCK_SIZE_K": 128,
|
22 |
+
"GROUP_SIZE_M": 64,
|
23 |
+
"num_warps": 8,
|
24 |
+
"num_stages": 2
|
25 |
+
},
|
26 |
+
"8": {
|
27 |
+
"BLOCK_SIZE_M": 16,
|
28 |
+
"BLOCK_SIZE_N": 128,
|
29 |
+
"BLOCK_SIZE_K": 256,
|
30 |
+
"GROUP_SIZE_M": 32,
|
31 |
+
"num_warps": 8,
|
32 |
+
"num_stages": 2
|
33 |
+
},
|
34 |
+
"16": {
|
35 |
+
"BLOCK_SIZE_M": 16,
|
36 |
+
"BLOCK_SIZE_N": 128,
|
37 |
+
"BLOCK_SIZE_K": 256,
|
38 |
+
"GROUP_SIZE_M": 16,
|
39 |
+
"num_warps": 8,
|
40 |
+
"num_stages": 2
|
41 |
+
},
|
42 |
+
"24": {
|
43 |
+
"BLOCK_SIZE_M": 16,
|
44 |
+
"BLOCK_SIZE_N": 128,
|
45 |
+
"BLOCK_SIZE_K": 256,
|
46 |
+
"GROUP_SIZE_M": 16,
|
47 |
+
"num_warps": 8,
|
48 |
+
"num_stages": 2
|
49 |
+
},
|
50 |
+
"32": {
|
51 |
+
"BLOCK_SIZE_M": 16,
|
52 |
+
"BLOCK_SIZE_N": 128,
|
53 |
+
"BLOCK_SIZE_K": 256,
|
54 |
+
"GROUP_SIZE_M": 16,
|
55 |
+
"num_warps": 8,
|
56 |
+
"num_stages": 2
|
57 |
+
},
|
58 |
+
"48": {
|
59 |
+
"BLOCK_SIZE_M": 16,
|
60 |
+
"BLOCK_SIZE_N": 128,
|
61 |
+
"BLOCK_SIZE_K": 256,
|
62 |
+
"GROUP_SIZE_M": 16,
|
63 |
+
"num_warps": 8,
|
64 |
+
"num_stages": 2
|
65 |
+
},
|
66 |
+
"64": {
|
67 |
+
"BLOCK_SIZE_M": 16,
|
68 |
+
"BLOCK_SIZE_N": 128,
|
69 |
+
"BLOCK_SIZE_K": 256,
|
70 |
+
"GROUP_SIZE_M": 64,
|
71 |
+
"num_warps": 8,
|
72 |
+
"num_stages": 2
|
73 |
+
},
|
74 |
+
"96": {
|
75 |
+
"BLOCK_SIZE_M": 16,
|
76 |
+
"BLOCK_SIZE_N": 128,
|
77 |
+
"BLOCK_SIZE_K": 256,
|
78 |
+
"GROUP_SIZE_M": 16,
|
79 |
+
"num_warps": 8,
|
80 |
+
"num_stages": 2
|
81 |
+
},
|
82 |
+
"128": {
|
83 |
+
"BLOCK_SIZE_M": 16,
|
84 |
+
"BLOCK_SIZE_N": 128,
|
85 |
+
"BLOCK_SIZE_K": 256,
|
86 |
+
"GROUP_SIZE_M": 16,
|
87 |
+
"num_warps": 8,
|
88 |
+
"num_stages": 2
|
89 |
+
},
|
90 |
+
"256": {
|
91 |
+
"BLOCK_SIZE_M": 16,
|
92 |
+
"BLOCK_SIZE_N": 128,
|
93 |
+
"BLOCK_SIZE_K": 256,
|
94 |
+
"GROUP_SIZE_M": 32,
|
95 |
+
"num_warps": 8,
|
96 |
+
"num_stages": 2
|
97 |
+
},
|
98 |
+
"512": {
|
99 |
+
"BLOCK_SIZE_M": 32,
|
100 |
+
"BLOCK_SIZE_N": 128,
|
101 |
+
"BLOCK_SIZE_K": 256,
|
102 |
+
"GROUP_SIZE_M": 32,
|
103 |
+
"num_warps": 8,
|
104 |
+
"num_stages": 2
|
105 |
+
},
|
106 |
+
"1024": {
|
107 |
+
"BLOCK_SIZE_M": 64,
|
108 |
+
"BLOCK_SIZE_N": 128,
|
109 |
+
"BLOCK_SIZE_K": 128,
|
110 |
+
"GROUP_SIZE_M": 1,
|
111 |
+
"num_warps": 4,
|
112 |
+
"num_stages": 2
|
113 |
+
},
|
114 |
+
"1536": {
|
115 |
+
"BLOCK_SIZE_M": 64,
|
116 |
+
"BLOCK_SIZE_N": 128,
|
117 |
+
"BLOCK_SIZE_K": 256,
|
118 |
+
"GROUP_SIZE_M": 1,
|
119 |
+
"num_warps": 8,
|
120 |
+
"num_stages": 2
|
121 |
+
},
|
122 |
+
"2048": {
|
123 |
+
"BLOCK_SIZE_M": 64,
|
124 |
+
"BLOCK_SIZE_N": 128,
|
125 |
+
"BLOCK_SIZE_K": 128,
|
126 |
+
"GROUP_SIZE_M": 1,
|
127 |
+
"num_warps": 4,
|
128 |
+
"num_stages": 2
|
129 |
+
},
|
130 |
+
"3072": {
|
131 |
+
"BLOCK_SIZE_M": 64,
|
132 |
+
"BLOCK_SIZE_N": 128,
|
133 |
+
"BLOCK_SIZE_K": 128,
|
134 |
+
"GROUP_SIZE_M": 16,
|
135 |
+
"num_warps": 4,
|
136 |
+
"num_stages": 2
|
137 |
+
},
|
138 |
+
"4096": {
|
139 |
+
"BLOCK_SIZE_M": 64,
|
140 |
+
"BLOCK_SIZE_N": 128,
|
141 |
+
"BLOCK_SIZE_K": 128,
|
142 |
+
"GROUP_SIZE_M": 16,
|
143 |
+
"num_warps": 4,
|
144 |
+
"num_stages": 2
|
145 |
+
}
|
146 |
+
}
|
build/torch25-cxx11-cu118-x86_64-linux/moe/configs/E=256,N=512,device_name=AMD_Instinct_MI325_OAM,dtype=fp8_w8a8,block_shape=[128,128].json
ADDED
@@ -0,0 +1,200 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"1": {
|
3 |
+
"BLOCK_SIZE_M": 16,
|
4 |
+
"BLOCK_SIZE_N": 16,
|
5 |
+
"BLOCK_SIZE_K": 256,
|
6 |
+
"GROUP_SIZE_M": 1,
|
7 |
+
"num_warps": 4,
|
8 |
+
"num_stages": 2,
|
9 |
+
"waves_per_eu": 0,
|
10 |
+
"matrix_instr_nonkdim": 16,
|
11 |
+
"kpack": 1
|
12 |
+
},
|
13 |
+
"2": {
|
14 |
+
"BLOCK_SIZE_M": 16,
|
15 |
+
"BLOCK_SIZE_N": 64,
|
16 |
+
"BLOCK_SIZE_K": 128,
|
17 |
+
"GROUP_SIZE_M": 1,
|
18 |
+
"num_warps": 4,
|
19 |
+
"num_stages": 2,
|
20 |
+
"waves_per_eu": 0,
|
21 |
+
"matrix_instr_nonkdim": 16,
|
22 |
+
"kpack": 1
|
23 |
+
},
|
24 |
+
"4": {
|
25 |
+
"BLOCK_SIZE_M": 64,
|
26 |
+
"BLOCK_SIZE_N": 64,
|
27 |
+
"BLOCK_SIZE_K": 128,
|
28 |
+
"GROUP_SIZE_M": 1,
|
29 |
+
"num_warps": 8,
|
30 |
+
"num_stages": 2,
|
31 |
+
"waves_per_eu": 0,
|
32 |
+
"matrix_instr_nonkdim": 16,
|
33 |
+
"kpack": 1
|
34 |
+
},
|
35 |
+
"8": {
|
36 |
+
"BLOCK_SIZE_M": 16,
|
37 |
+
"BLOCK_SIZE_N": 64,
|
38 |
+
"BLOCK_SIZE_K": 128,
|
39 |
+
"GROUP_SIZE_M": 1,
|
40 |
+
"num_warps": 4,
|
41 |
+
"num_stages": 2,
|
42 |
+
"waves_per_eu": 0,
|
43 |
+
"matrix_instr_nonkdim": 16,
|
44 |
+
"kpack": 1
|
45 |
+
},
|
46 |
+
"16": {
|
47 |
+
"BLOCK_SIZE_M": 16,
|
48 |
+
"BLOCK_SIZE_N": 64,
|
49 |
+
"BLOCK_SIZE_K": 128,
|
50 |
+
"GROUP_SIZE_M": 1,
|
51 |
+
"num_warps": 4,
|
52 |
+
"num_stages": 2,
|
53 |
+
"waves_per_eu": 0,
|
54 |
+
"matrix_instr_nonkdim": 16,
|
55 |
+
"kpack": 1
|
56 |
+
},
|
57 |
+
"24": {
|
58 |
+
"BLOCK_SIZE_M": 32,
|
59 |
+
"BLOCK_SIZE_N": 128,
|
60 |
+
"BLOCK_SIZE_K": 64,
|
61 |
+
"GROUP_SIZE_M": 1,
|
62 |
+
"num_warps": 8,
|
63 |
+
"num_stages": 2,
|
64 |
+
"waves_per_eu": 0,
|
65 |
+
"matrix_instr_nonkdim": 16,
|
66 |
+
"kpack": 2
|
67 |
+
},
|
68 |
+
"32": {
|
69 |
+
"BLOCK_SIZE_M": 16,
|
70 |
+
"BLOCK_SIZE_N": 64,
|
71 |
+
"BLOCK_SIZE_K": 128,
|
72 |
+
"GROUP_SIZE_M": 1,
|
73 |
+
"num_warps": 4,
|
74 |
+
"num_stages": 2,
|
75 |
+
"waves_per_eu": 0,
|
76 |
+
"matrix_instr_nonkdim": 16,
|
77 |
+
"kpack": 2
|
78 |
+
},
|
79 |
+
"48": {
|
80 |
+
"BLOCK_SIZE_M": 16,
|
81 |
+
"BLOCK_SIZE_N": 64,
|
82 |
+
"BLOCK_SIZE_K": 128,
|
83 |
+
"GROUP_SIZE_M": 1,
|
84 |
+
"num_warps": 2,
|
85 |
+
"num_stages": 2,
|
86 |
+
"waves_per_eu": 0,
|
87 |
+
"matrix_instr_nonkdim": 16,
|
88 |
+
"kpack": 2
|
89 |
+
},
|
90 |
+
"64": {
|
91 |
+
"BLOCK_SIZE_M": 16,
|
92 |
+
"BLOCK_SIZE_N": 64,
|
93 |
+
"BLOCK_SIZE_K": 128,
|
94 |
+
"GROUP_SIZE_M": 1,
|
95 |
+
"num_warps": 2,
|
96 |
+
"num_stages": 2,
|
97 |
+
"waves_per_eu": 0,
|
98 |
+
"matrix_instr_nonkdim": 16,
|
99 |
+
"kpack": 2
|
100 |
+
},
|
101 |
+
"96": {
|
102 |
+
"BLOCK_SIZE_M": 16,
|
103 |
+
"BLOCK_SIZE_N": 32,
|
104 |
+
"BLOCK_SIZE_K": 128,
|
105 |
+
"GROUP_SIZE_M": 4,
|
106 |
+
"num_warps": 2,
|
107 |
+
"num_stages": 2,
|
108 |
+
"waves_per_eu": 0,
|
109 |
+
"matrix_instr_nonkdim": 16,
|
110 |
+
"kpack": 2
|
111 |
+
},
|
112 |
+
"128": {
|
113 |
+
"BLOCK_SIZE_M": 16,
|
114 |
+
"BLOCK_SIZE_N": 32,
|
115 |
+
"BLOCK_SIZE_K": 128,
|
116 |
+
"GROUP_SIZE_M": 4,
|
117 |
+
"num_warps": 1,
|
118 |
+
"num_stages": 2,
|
119 |
+
"waves_per_eu": 0,
|
120 |
+
"matrix_instr_nonkdim": 16,
|
121 |
+
"kpack": 2
|
122 |
+
},
|
123 |
+
"256": {
|
124 |
+
"BLOCK_SIZE_M": 16,
|
125 |
+
"BLOCK_SIZE_N": 32,
|
126 |
+
"BLOCK_SIZE_K": 128,
|
127 |
+
"GROUP_SIZE_M": 1,
|
128 |
+
"num_warps": 2,
|
129 |
+
"num_stages": 2,
|
130 |
+
"waves_per_eu": 0,
|
131 |
+
"matrix_instr_nonkdim": 16,
|
132 |
+
"kpack": 2
|
133 |
+
},
|
134 |
+
"512": {
|
135 |
+
"BLOCK_SIZE_M": 32,
|
136 |
+
"BLOCK_SIZE_N": 64,
|
137 |
+
"BLOCK_SIZE_K": 128,
|
138 |
+
"GROUP_SIZE_M": 8,
|
139 |
+
"num_warps": 8,
|
140 |
+
"num_stages": 2,
|
141 |
+
"waves_per_eu": 0,
|
142 |
+
"matrix_instr_nonkdim": 16,
|
143 |
+
"kpack": 2
|
144 |
+
},
|
145 |
+
"1024": {
|
146 |
+
"BLOCK_SIZE_M": 64,
|
147 |
+
"BLOCK_SIZE_N": 128,
|
148 |
+
"BLOCK_SIZE_K": 128,
|
149 |
+
"GROUP_SIZE_M": 8,
|
150 |
+
"num_warps": 8,
|
151 |
+
"num_stages": 2,
|
152 |
+
"waves_per_eu": 0,
|
153 |
+
"matrix_instr_nonkdim": 16,
|
154 |
+
"kpack": 2
|
155 |
+
},
|
156 |
+
"1536": {
|
157 |
+
"BLOCK_SIZE_M": 64,
|
158 |
+
"BLOCK_SIZE_N": 128,
|
159 |
+
"BLOCK_SIZE_K": 128,
|
160 |
+
"GROUP_SIZE_M": 8,
|
161 |
+
"num_warps": 8,
|
162 |
+
"num_stages": 2,
|
163 |
+
"waves_per_eu": 0,
|
164 |
+
"matrix_instr_nonkdim": 16,
|
165 |
+
"kpack": 2
|
166 |
+
},
|
167 |
+
"2048": {
|
168 |
+
"BLOCK_SIZE_M": 128,
|
169 |
+
"BLOCK_SIZE_N": 128,
|
170 |
+
"BLOCK_SIZE_K": 64,
|
171 |
+
"GROUP_SIZE_M": 8,
|
172 |
+
"num_warps": 8,
|
173 |
+
"num_stages": 2,
|
174 |
+
"waves_per_eu": 0,
|
175 |
+
"matrix_instr_nonkdim": 16,
|
176 |
+
"kpack": 2
|
177 |
+
},
|
178 |
+
"3072": {
|
179 |
+
"BLOCK_SIZE_M": 128,
|
180 |
+
"BLOCK_SIZE_N": 128,
|
181 |
+
"BLOCK_SIZE_K": 64,
|
182 |
+
"GROUP_SIZE_M": 8,
|
183 |
+
"num_warps": 8,
|
184 |
+
"num_stages": 2,
|
185 |
+
"waves_per_eu": 0,
|
186 |
+
"matrix_instr_nonkdim": 16,
|
187 |
+
"kpack": 2
|
188 |
+
},
|
189 |
+
"4096": {
|
190 |
+
"BLOCK_SIZE_M": 128,
|
191 |
+
"BLOCK_SIZE_N": 128,
|
192 |
+
"BLOCK_SIZE_K": 64,
|
193 |
+
"GROUP_SIZE_M": 8,
|
194 |
+
"num_warps": 8,
|
195 |
+
"num_stages": 2,
|
196 |
+
"waves_per_eu": 0,
|
197 |
+
"matrix_instr_nonkdim": 16,
|
198 |
+
"kpack": 2
|
199 |
+
}
|
200 |
+
}
|
build/torch25-cxx11-cu118-x86_64-linux/moe/configs/E=256,N=64,device_name=NVIDIA_A800-SXM4-80GB.json
ADDED
@@ -0,0 +1,146 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"1": {
|
3 |
+
"BLOCK_SIZE_M": 16,
|
4 |
+
"BLOCK_SIZE_N": 64,
|
5 |
+
"BLOCK_SIZE_K": 64,
|
6 |
+
"GROUP_SIZE_M": 16,
|
7 |
+
"num_warps": 4,
|
8 |
+
"num_stages": 3
|
9 |
+
},
|
10 |
+
"2": {
|
11 |
+
"BLOCK_SIZE_M": 16,
|
12 |
+
"BLOCK_SIZE_N": 64,
|
13 |
+
"BLOCK_SIZE_K": 64,
|
14 |
+
"GROUP_SIZE_M": 64,
|
15 |
+
"num_warps": 4,
|
16 |
+
"num_stages": 4
|
17 |
+
},
|
18 |
+
"4": {
|
19 |
+
"BLOCK_SIZE_M": 16,
|
20 |
+
"BLOCK_SIZE_N": 64,
|
21 |
+
"BLOCK_SIZE_K": 64,
|
22 |
+
"GROUP_SIZE_M": 64,
|
23 |
+
"num_warps": 4,
|
24 |
+
"num_stages": 4
|
25 |
+
},
|
26 |
+
"8": {
|
27 |
+
"BLOCK_SIZE_M": 16,
|
28 |
+
"BLOCK_SIZE_N": 64,
|
29 |
+
"BLOCK_SIZE_K": 64,
|
30 |
+
"GROUP_SIZE_M": 32,
|
31 |
+
"num_warps": 4,
|
32 |
+
"num_stages": 3
|
33 |
+
},
|
34 |
+
"16": {
|
35 |
+
"BLOCK_SIZE_M": 16,
|
36 |
+
"BLOCK_SIZE_N": 128,
|
37 |
+
"BLOCK_SIZE_K": 64,
|
38 |
+
"GROUP_SIZE_M": 64,
|
39 |
+
"num_warps": 4,
|
40 |
+
"num_stages": 3
|
41 |
+
},
|
42 |
+
"24": {
|
43 |
+
"BLOCK_SIZE_M": 16,
|
44 |
+
"BLOCK_SIZE_N": 64,
|
45 |
+
"BLOCK_SIZE_K": 64,
|
46 |
+
"GROUP_SIZE_M": 1,
|
47 |
+
"num_warps": 4,
|
48 |
+
"num_stages": 3
|
49 |
+
},
|
50 |
+
"32": {
|
51 |
+
"BLOCK_SIZE_M": 16,
|
52 |
+
"BLOCK_SIZE_N": 64,
|
53 |
+
"BLOCK_SIZE_K": 64,
|
54 |
+
"GROUP_SIZE_M": 1,
|
55 |
+
"num_warps": 4,
|
56 |
+
"num_stages": 3
|
57 |
+
},
|
58 |
+
"48": {
|
59 |
+
"BLOCK_SIZE_M": 16,
|
60 |
+
"BLOCK_SIZE_N": 64,
|
61 |
+
"BLOCK_SIZE_K": 64,
|
62 |
+
"GROUP_SIZE_M": 1,
|
63 |
+
"num_warps": 4,
|
64 |
+
"num_stages": 3
|
65 |
+
},
|
66 |
+
"64": {
|
67 |
+
"BLOCK_SIZE_M": 16,
|
68 |
+
"BLOCK_SIZE_N": 64,
|
69 |
+
"BLOCK_SIZE_K": 64,
|
70 |
+
"GROUP_SIZE_M": 1,
|
71 |
+
"num_warps": 4,
|
72 |
+
"num_stages": 3
|
73 |
+
},
|
74 |
+
"96": {
|
75 |
+
"BLOCK_SIZE_M": 16,
|
76 |
+
"BLOCK_SIZE_N": 128,
|
77 |
+
"BLOCK_SIZE_K": 64,
|
78 |
+
"GROUP_SIZE_M": 32,
|
79 |
+
"num_warps": 4,
|
80 |
+
"num_stages": 3
|
81 |
+
},
|
82 |
+
"128": {
|
83 |
+
"BLOCK_SIZE_M": 16,
|
84 |
+
"BLOCK_SIZE_N": 128,
|
85 |
+
"BLOCK_SIZE_K": 64,
|
86 |
+
"GROUP_SIZE_M": 64,
|
87 |
+
"num_warps": 4,
|
88 |
+
"num_stages": 3
|
89 |
+
},
|
90 |
+
"256": {
|
91 |
+
"BLOCK_SIZE_M": 16,
|
92 |
+
"BLOCK_SIZE_N": 128,
|
93 |
+
"BLOCK_SIZE_K": 64,
|
94 |
+
"GROUP_SIZE_M": 16,
|
95 |
+
"num_warps": 4,
|
96 |
+
"num_stages": 3
|
97 |
+
},
|
98 |
+
"512": {
|
99 |
+
"BLOCK_SIZE_M": 32,
|
100 |
+
"BLOCK_SIZE_N": 128,
|
101 |
+
"BLOCK_SIZE_K": 64,
|
102 |
+
"GROUP_SIZE_M": 64,
|
103 |
+
"num_warps": 4,
|
104 |
+
"num_stages": 3
|
105 |
+
},
|
106 |
+
"1024": {
|
107 |
+
"BLOCK_SIZE_M": 32,
|
108 |
+
"BLOCK_SIZE_N": 128,
|
109 |
+
"BLOCK_SIZE_K": 64,
|
110 |
+
"GROUP_SIZE_M": 64,
|
111 |
+
"num_warps": 4,
|
112 |
+
"num_stages": 2
|
113 |
+
},
|
114 |
+
"1536": {
|
115 |
+
"BLOCK_SIZE_M": 64,
|
116 |
+
"BLOCK_SIZE_N": 128,
|
117 |
+
"BLOCK_SIZE_K": 64,
|
118 |
+
"GROUP_SIZE_M": 64,
|
119 |
+
"num_warps": 4,
|
120 |
+
"num_stages": 2
|
121 |
+
},
|
122 |
+
"2048": {
|
123 |
+
"BLOCK_SIZE_M": 32,
|
124 |
+
"BLOCK_SIZE_N": 128,
|
125 |
+
"BLOCK_SIZE_K": 64,
|
126 |
+
"GROUP_SIZE_M": 64,
|
127 |
+
"num_warps": 4,
|
128 |
+
"num_stages": 2
|
129 |
+
},
|
130 |
+
"3072": {
|
131 |
+
"BLOCK_SIZE_M": 128,
|
132 |
+
"BLOCK_SIZE_N": 128,
|
133 |
+
"BLOCK_SIZE_K": 64,
|
134 |
+
"GROUP_SIZE_M": 1,
|
135 |
+
"num_warps": 4,
|
136 |
+
"num_stages": 3
|
137 |
+
},
|
138 |
+
"4096": {
|
139 |
+
"BLOCK_SIZE_M": 64,
|
140 |
+
"BLOCK_SIZE_N": 128,
|
141 |
+
"BLOCK_SIZE_K": 64,
|
142 |
+
"GROUP_SIZE_M": 64,
|
143 |
+
"num_warps": 4,
|
144 |
+
"num_stages": 2
|
145 |
+
}
|
146 |
+
}
|
build/torch25-cxx11-cu118-x86_64-linux/moe/configs/E=60,N=1408,device_name=AMD_Instinct_MI300X.json
ADDED
@@ -0,0 +1,200 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"1": {
|
3 |
+
"BLOCK_SIZE_M": 16,
|
4 |
+
"BLOCK_SIZE_N": 64,
|
5 |
+
"BLOCK_SIZE_K": 128,
|
6 |
+
"GROUP_SIZE_M": 1,
|
7 |
+
"num_warps": 4,
|
8 |
+
"num_stages": 2,
|
9 |
+
"waves_per_eu": 0,
|
10 |
+
"matrix_instr_nonkdim": 16,
|
11 |
+
"kpack": 1
|
12 |
+
},
|
13 |
+
"2": {
|
14 |
+
"BLOCK_SIZE_M": 16,
|
15 |
+
"BLOCK_SIZE_N": 128,
|
16 |
+
"BLOCK_SIZE_K": 128,
|
17 |
+
"GROUP_SIZE_M": 1,
|
18 |
+
"num_warps": 8,
|
19 |
+
"num_stages": 2,
|
20 |
+
"waves_per_eu": 0,
|
21 |
+
"matrix_instr_nonkdim": 16,
|
22 |
+
"kpack": 1
|
23 |
+
},
|
24 |
+
"4": {
|
25 |
+
"BLOCK_SIZE_M": 16,
|
26 |
+
"BLOCK_SIZE_N": 32,
|
27 |
+
"BLOCK_SIZE_K": 64,
|
28 |
+
"GROUP_SIZE_M": 1,
|
29 |
+
"num_warps": 2,
|
30 |
+
"num_stages": 2,
|
31 |
+
"waves_per_eu": 0,
|
32 |
+
"matrix_instr_nonkdim": 16,
|
33 |
+
"kpack": 1
|
34 |
+
},
|
35 |
+
"8": {
|
36 |
+
"BLOCK_SIZE_M": 32,
|
37 |
+
"BLOCK_SIZE_N": 256,
|
38 |
+
"BLOCK_SIZE_K": 64,
|
39 |
+
"GROUP_SIZE_M": 1,
|
40 |
+
"num_warps": 8,
|
41 |
+
"num_stages": 2,
|
42 |
+
"waves_per_eu": 0,
|
43 |
+
"matrix_instr_nonkdim": 32,
|
44 |
+
"kpack": 2
|
45 |
+
},
|
46 |
+
"16": {
|
47 |
+
"BLOCK_SIZE_M": 32,
|
48 |
+
"BLOCK_SIZE_N": 128,
|
49 |
+
"BLOCK_SIZE_K": 64,
|
50 |
+
"GROUP_SIZE_M": 1,
|
51 |
+
"num_warps": 2,
|
52 |
+
"num_stages": 2,
|
53 |
+
"waves_per_eu": 0,
|
54 |
+
"matrix_instr_nonkdim": 32,
|
55 |
+
"kpack": 2
|
56 |
+
},
|
57 |
+
"24": {
|
58 |
+
"BLOCK_SIZE_M": 16,
|
59 |
+
"BLOCK_SIZE_N": 64,
|
60 |
+
"BLOCK_SIZE_K": 256,
|
61 |
+
"GROUP_SIZE_M": 1,
|
62 |
+
"num_warps": 4,
|
63 |
+
"num_stages": 2,
|
64 |
+
"waves_per_eu": 0,
|
65 |
+
"matrix_instr_nonkdim": 16,
|
66 |
+
"kpack": 2
|
67 |
+
},
|
68 |
+
"32": {
|
69 |
+
"BLOCK_SIZE_M": 16,
|
70 |
+
"BLOCK_SIZE_N": 64,
|
71 |
+
"BLOCK_SIZE_K": 256,
|
72 |
+
"GROUP_SIZE_M": 1,
|
73 |
+
"num_warps": 4,
|
74 |
+
"num_stages": 2,
|
75 |
+
"waves_per_eu": 0,
|
76 |
+
"matrix_instr_nonkdim": 16,
|
77 |
+
"kpack": 1
|
78 |
+
},
|
79 |
+
"48": {
|
80 |
+
"BLOCK_SIZE_M": 16,
|
81 |
+
"BLOCK_SIZE_N": 64,
|
82 |
+
"BLOCK_SIZE_K": 256,
|
83 |
+
"GROUP_SIZE_M": 4,
|
84 |
+
"num_warps": 4,
|
85 |
+
"num_stages": 2,
|
86 |
+
"waves_per_eu": 0,
|
87 |
+
"matrix_instr_nonkdim": 16,
|
88 |
+
"kpack": 2
|
89 |
+
},
|
90 |
+
"64": {
|
91 |
+
"BLOCK_SIZE_M": 16,
|
92 |
+
"BLOCK_SIZE_N": 64,
|
93 |
+
"BLOCK_SIZE_K": 256,
|
94 |
+
"GROUP_SIZE_M": 4,
|
95 |
+
"num_warps": 4,
|
96 |
+
"num_stages": 2,
|
97 |
+
"waves_per_eu": 0,
|
98 |
+
"matrix_instr_nonkdim": 16,
|
99 |
+
"kpack": 2
|
100 |
+
},
|
101 |
+
"96": {
|
102 |
+
"BLOCK_SIZE_M": 16,
|
103 |
+
"BLOCK_SIZE_N": 64,
|
104 |
+
"BLOCK_SIZE_K": 256,
|
105 |
+
"GROUP_SIZE_M": 4,
|
106 |
+
"num_warps": 4,
|
107 |
+
"num_stages": 2,
|
108 |
+
"waves_per_eu": 0,
|
109 |
+
"matrix_instr_nonkdim": 16,
|
110 |
+
"kpack": 2
|
111 |
+
},
|
112 |
+
"128": {
|
113 |
+
"BLOCK_SIZE_M": 16,
|
114 |
+
"BLOCK_SIZE_N": 64,
|
115 |
+
"BLOCK_SIZE_K": 256,
|
116 |
+
"GROUP_SIZE_M": 4,
|
117 |
+
"num_warps": 4,
|
118 |
+
"num_stages": 2,
|
119 |
+
"waves_per_eu": 0,
|
120 |
+
"matrix_instr_nonkdim": 16,
|
121 |
+
"kpack": 2
|
122 |
+
},
|
123 |
+
"256": {
|
124 |
+
"BLOCK_SIZE_M": 32,
|
125 |
+
"BLOCK_SIZE_N": 32,
|
126 |
+
"BLOCK_SIZE_K": 256,
|
127 |
+
"GROUP_SIZE_M": 4,
|
128 |
+
"num_warps": 4,
|
129 |
+
"num_stages": 2,
|
130 |
+
"waves_per_eu": 0,
|
131 |
+
"matrix_instr_nonkdim": 16,
|
132 |
+
"kpack": 2
|
133 |
+
},
|
134 |
+
"512": {
|
135 |
+
"BLOCK_SIZE_M": 64,
|
136 |
+
"BLOCK_SIZE_N": 64,
|
137 |
+
"BLOCK_SIZE_K": 128,
|
138 |
+
"GROUP_SIZE_M": 32,
|
139 |
+
"num_warps": 8,
|
140 |
+
"num_stages": 2,
|
141 |
+
"waves_per_eu": 0,
|
142 |
+
"matrix_instr_nonkdim": 16,
|
143 |
+
"kpack": 1
|
144 |
+
},
|
145 |
+
"1024": {
|
146 |
+
"BLOCK_SIZE_M": 128,
|
147 |
+
"BLOCK_SIZE_N": 128,
|
148 |
+
"BLOCK_SIZE_K": 64,
|
149 |
+
"GROUP_SIZE_M": 32,
|
150 |
+
"num_warps": 8,
|
151 |
+
"num_stages": 2,
|
152 |
+
"waves_per_eu": 0,
|
153 |
+
"matrix_instr_nonkdim": 16,
|
154 |
+
"kpack": 2
|
155 |
+
},
|
156 |
+
"1536": {
|
157 |
+
"BLOCK_SIZE_M": 128,
|
158 |
+
"BLOCK_SIZE_N": 128,
|
159 |
+
"BLOCK_SIZE_K": 64,
|
160 |
+
"GROUP_SIZE_M": 32,
|
161 |
+
"num_warps": 8,
|
162 |
+
"num_stages": 2,
|
163 |
+
"waves_per_eu": 0,
|
164 |
+
"matrix_instr_nonkdim": 16,
|
165 |
+
"kpack": 2
|
166 |
+
},
|
167 |
+
"2048": {
|
168 |
+
"BLOCK_SIZE_M": 128,
|
169 |
+
"BLOCK_SIZE_N": 128,
|
170 |
+
"BLOCK_SIZE_K": 64,
|
171 |
+
"GROUP_SIZE_M": 4,
|
172 |
+
"num_warps": 8,
|
173 |
+
"num_stages": 2,
|
174 |
+
"waves_per_eu": 0,
|
175 |
+
"matrix_instr_nonkdim": 16,
|
176 |
+
"kpack": 2
|
177 |
+
},
|
178 |
+
"3072": {
|
179 |
+
"BLOCK_SIZE_M": 128,
|
180 |
+
"BLOCK_SIZE_N": 128,
|
181 |
+
"BLOCK_SIZE_K": 64,
|
182 |
+
"GROUP_SIZE_M": 8,
|
183 |
+
"num_warps": 8,
|
184 |
+
"num_stages": 2,
|
185 |
+
"waves_per_eu": 0,
|
186 |
+
"matrix_instr_nonkdim": 16,
|
187 |
+
"kpack": 2
|
188 |
+
},
|
189 |
+
"4096": {
|
190 |
+
"BLOCK_SIZE_M": 128,
|
191 |
+
"BLOCK_SIZE_N": 128,
|
192 |
+
"BLOCK_SIZE_K": 64,
|
193 |
+
"GROUP_SIZE_M": 4,
|
194 |
+
"num_warps": 8,
|
195 |
+
"num_stages": 2,
|
196 |
+
"waves_per_eu": 0,
|
197 |
+
"matrix_instr_nonkdim": 16,
|
198 |
+
"kpack": 2
|
199 |
+
}
|
200 |
+
}
|
build/torch25-cxx11-cu118-x86_64-linux/moe/configs/E=60,N=176,device_name=AMD_Instinct_MI300X.json
ADDED
@@ -0,0 +1,200 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"1": {
|
3 |
+
"BLOCK_SIZE_M": 16,
|
4 |
+
"BLOCK_SIZE_N": 32,
|
5 |
+
"BLOCK_SIZE_K": 256,
|
6 |
+
"GROUP_SIZE_M": 1,
|
7 |
+
"num_warps": 2,
|
8 |
+
"num_stages": 2,
|
9 |
+
"waves_per_eu": 0,
|
10 |
+
"matrix_instr_nonkdim": 16,
|
11 |
+
"kpack": 2
|
12 |
+
},
|
13 |
+
"2": {
|
14 |
+
"BLOCK_SIZE_M": 16,
|
15 |
+
"BLOCK_SIZE_N": 16,
|
16 |
+
"BLOCK_SIZE_K": 256,
|
17 |
+
"GROUP_SIZE_M": 1,
|
18 |
+
"num_warps": 4,
|
19 |
+
"num_stages": 2,
|
20 |
+
"waves_per_eu": 0,
|
21 |
+
"matrix_instr_nonkdim": 16,
|
22 |
+
"kpack": 2
|
23 |
+
},
|
24 |
+
"4": {
|
25 |
+
"BLOCK_SIZE_M": 16,
|
26 |
+
"BLOCK_SIZE_N": 16,
|
27 |
+
"BLOCK_SIZE_K": 128,
|
28 |
+
"GROUP_SIZE_M": 1,
|
29 |
+
"num_warps": 4,
|
30 |
+
"num_stages": 2,
|
31 |
+
"waves_per_eu": 0,
|
32 |
+
"matrix_instr_nonkdim": 16,
|
33 |
+
"kpack": 2
|
34 |
+
},
|
35 |
+
"8": {
|
36 |
+
"BLOCK_SIZE_M": 16,
|
37 |
+
"BLOCK_SIZE_N": 64,
|
38 |
+
"BLOCK_SIZE_K": 128,
|
39 |
+
"GROUP_SIZE_M": 1,
|
40 |
+
"num_warps": 4,
|
41 |
+
"num_stages": 2,
|
42 |
+
"waves_per_eu": 0,
|
43 |
+
"matrix_instr_nonkdim": 16,
|
44 |
+
"kpack": 2
|
45 |
+
},
|
46 |
+
"16": {
|
47 |
+
"BLOCK_SIZE_M": 16,
|
48 |
+
"BLOCK_SIZE_N": 16,
|
49 |
+
"BLOCK_SIZE_K": 64,
|
50 |
+
"GROUP_SIZE_M": 4,
|
51 |
+
"num_warps": 2,
|
52 |
+
"num_stages": 2,
|
53 |
+
"waves_per_eu": 0,
|
54 |
+
"matrix_instr_nonkdim": 16,
|
55 |
+
"kpack": 1
|
56 |
+
},
|
57 |
+
"24": {
|
58 |
+
"BLOCK_SIZE_M": 16,
|
59 |
+
"BLOCK_SIZE_N": 16,
|
60 |
+
"BLOCK_SIZE_K": 64,
|
61 |
+
"GROUP_SIZE_M": 4,
|
62 |
+
"num_warps": 2,
|
63 |
+
"num_stages": 2,
|
64 |
+
"waves_per_eu": 0,
|
65 |
+
"matrix_instr_nonkdim": 16,
|
66 |
+
"kpack": 2
|
67 |
+
},
|
68 |
+
"32": {
|
69 |
+
"BLOCK_SIZE_M": 16,
|
70 |
+
"BLOCK_SIZE_N": 32,
|
71 |
+
"BLOCK_SIZE_K": 64,
|
72 |
+
"GROUP_SIZE_M": 8,
|
73 |
+
"num_warps": 2,
|
74 |
+
"num_stages": 2,
|
75 |
+
"waves_per_eu": 0,
|
76 |
+
"matrix_instr_nonkdim": 16,
|
77 |
+
"kpack": 2
|
78 |
+
},
|
79 |
+
"48": {
|
80 |
+
"BLOCK_SIZE_M": 16,
|
81 |
+
"BLOCK_SIZE_N": 32,
|
82 |
+
"BLOCK_SIZE_K": 64,
|
83 |
+
"GROUP_SIZE_M": 8,
|
84 |
+
"num_warps": 2,
|
85 |
+
"num_stages": 2,
|
86 |
+
"waves_per_eu": 0,
|
87 |
+
"matrix_instr_nonkdim": 16,
|
88 |
+
"kpack": 1
|
89 |
+
},
|
90 |
+
"64": {
|
91 |
+
"BLOCK_SIZE_M": 16,
|
92 |
+
"BLOCK_SIZE_N": 32,
|
93 |
+
"BLOCK_SIZE_K": 64,
|
94 |
+
"GROUP_SIZE_M": 8,
|
95 |
+
"num_warps": 2,
|
96 |
+
"num_stages": 2,
|
97 |
+
"waves_per_eu": 0,
|
98 |
+
"matrix_instr_nonkdim": 16,
|
99 |
+
"kpack": 1
|
100 |
+
},
|
101 |
+
"96": {
|
102 |
+
"BLOCK_SIZE_M": 16,
|
103 |
+
"BLOCK_SIZE_N": 16,
|
104 |
+
"BLOCK_SIZE_K": 64,
|
105 |
+
"GROUP_SIZE_M": 16,
|
106 |
+
"num_warps": 1,
|
107 |
+
"num_stages": 2,
|
108 |
+
"waves_per_eu": 0,
|
109 |
+
"matrix_instr_nonkdim": 16,
|
110 |
+
"kpack": 1
|
111 |
+
},
|
112 |
+
"128": {
|
113 |
+
"BLOCK_SIZE_M": 16,
|
114 |
+
"BLOCK_SIZE_N": 16,
|
115 |
+
"BLOCK_SIZE_K": 64,
|
116 |
+
"GROUP_SIZE_M": 16,
|
117 |
+
"num_warps": 1,
|
118 |
+
"num_stages": 2,
|
119 |
+
"waves_per_eu": 0,
|
120 |
+
"matrix_instr_nonkdim": 16,
|
121 |
+
"kpack": 1
|
122 |
+
},
|
123 |
+
"256": {
|
124 |
+
"BLOCK_SIZE_M": 32,
|
125 |
+
"BLOCK_SIZE_N": 64,
|
126 |
+
"BLOCK_SIZE_K": 64,
|
127 |
+
"GROUP_SIZE_M": 8,
|
128 |
+
"num_warps": 4,
|
129 |
+
"num_stages": 2,
|
130 |
+
"waves_per_eu": 0,
|
131 |
+
"matrix_instr_nonkdim": 16,
|
132 |
+
"kpack": 1
|
133 |
+
},
|
134 |
+
"512": {
|
135 |
+
"BLOCK_SIZE_M": 64,
|
136 |
+
"BLOCK_SIZE_N": 128,
|
137 |
+
"BLOCK_SIZE_K": 64,
|
138 |
+
"GROUP_SIZE_M": 32,
|
139 |
+
"num_warps": 8,
|
140 |
+
"num_stages": 2,
|
141 |
+
"waves_per_eu": 0,
|
142 |
+
"matrix_instr_nonkdim": 16,
|
143 |
+
"kpack": 1
|
144 |
+
},
|
145 |
+
"1024": {
|
146 |
+
"BLOCK_SIZE_M": 128,
|
147 |
+
"BLOCK_SIZE_N": 64,
|
148 |
+
"BLOCK_SIZE_K": 64,
|
149 |
+
"GROUP_SIZE_M": 16,
|
150 |
+
"num_warps": 4,
|
151 |
+
"num_stages": 2,
|
152 |
+
"waves_per_eu": 0,
|
153 |
+
"matrix_instr_nonkdim": 16,
|
154 |
+
"kpack": 2
|
155 |
+
},
|
156 |
+
"1536": {
|
157 |
+
"BLOCK_SIZE_M": 128,
|
158 |
+
"BLOCK_SIZE_N": 128,
|
159 |
+
"BLOCK_SIZE_K": 64,
|
160 |
+
"GROUP_SIZE_M": 8,
|
161 |
+
"num_warps": 8,
|
162 |
+
"num_stages": 2,
|
163 |
+
"waves_per_eu": 0,
|
164 |
+
"matrix_instr_nonkdim": 16,
|
165 |
+
"kpack": 2
|
166 |
+
},
|
167 |
+
"2048": {
|
168 |
+
"BLOCK_SIZE_M": 64,
|
169 |
+
"BLOCK_SIZE_N": 128,
|
170 |
+
"BLOCK_SIZE_K": 64,
|
171 |
+
"GROUP_SIZE_M": 8,
|
172 |
+
"num_warps": 8,
|
173 |
+
"num_stages": 2,
|
174 |
+
"waves_per_eu": 0,
|
175 |
+
"matrix_instr_nonkdim": 16,
|
176 |
+
"kpack": 1
|
177 |
+
},
|
178 |
+
"3072": {
|
179 |
+
"BLOCK_SIZE_M": 256,
|
180 |
+
"BLOCK_SIZE_N": 128,
|
181 |
+
"BLOCK_SIZE_K": 64,
|
182 |
+
"GROUP_SIZE_M": 4,
|
183 |
+
"num_warps": 8,
|
184 |
+
"num_stages": 2,
|
185 |
+
"waves_per_eu": 0,
|
186 |
+
"matrix_instr_nonkdim": 16,
|
187 |
+
"kpack": 2
|
188 |
+
},
|
189 |
+
"4096": {
|
190 |
+
"BLOCK_SIZE_M": 128,
|
191 |
+
"BLOCK_SIZE_N": 128,
|
192 |
+
"BLOCK_SIZE_K": 64,
|
193 |
+
"GROUP_SIZE_M": 8,
|
194 |
+
"num_warps": 8,
|
195 |
+
"num_stages": 2,
|
196 |
+
"waves_per_eu": 0,
|
197 |
+
"matrix_instr_nonkdim": 16,
|
198 |
+
"kpack": 2
|
199 |
+
}
|
200 |
+
}
|
build/torch25-cxx11-cu118-x86_64-linux/moe/configs/E=60,N=352,device_name=AMD_Instinct_MI300X.json
ADDED
@@ -0,0 +1,200 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"1": {
|
3 |
+
"BLOCK_SIZE_M": 16,
|
4 |
+
"BLOCK_SIZE_N": 16,
|
5 |
+
"BLOCK_SIZE_K": 256,
|
6 |
+
"GROUP_SIZE_M": 1,
|
7 |
+
"num_warps": 4,
|
8 |
+
"num_stages": 2,
|
9 |
+
"waves_per_eu": 0,
|
10 |
+
"matrix_instr_nonkdim": 16,
|
11 |
+
"kpack": 2
|
12 |
+
},
|
13 |
+
"2": {
|
14 |
+
"BLOCK_SIZE_M": 16,
|
15 |
+
"BLOCK_SIZE_N": 32,
|
16 |
+
"BLOCK_SIZE_K": 128,
|
17 |
+
"GROUP_SIZE_M": 1,
|
18 |
+
"num_warps": 4,
|
19 |
+
"num_stages": 2,
|
20 |
+
"waves_per_eu": 0,
|
21 |
+
"matrix_instr_nonkdim": 16,
|
22 |
+
"kpack": 2
|
23 |
+
},
|
24 |
+
"4": {
|
25 |
+
"BLOCK_SIZE_M": 16,
|
26 |
+
"BLOCK_SIZE_N": 64,
|
27 |
+
"BLOCK_SIZE_K": 128,
|
28 |
+
"GROUP_SIZE_M": 1,
|
29 |
+
"num_warps": 4,
|
30 |
+
"num_stages": 2,
|
31 |
+
"waves_per_eu": 0,
|
32 |
+
"matrix_instr_nonkdim": 16,
|
33 |
+
"kpack": 2
|
34 |
+
},
|
35 |
+
"8": {
|
36 |
+
"BLOCK_SIZE_M": 16,
|
37 |
+
"BLOCK_SIZE_N": 64,
|
38 |
+
"BLOCK_SIZE_K": 128,
|
39 |
+
"GROUP_SIZE_M": 1,
|
40 |
+
"num_warps": 4,
|
41 |
+
"num_stages": 2,
|
42 |
+
"waves_per_eu": 0,
|
43 |
+
"matrix_instr_nonkdim": 16,
|
44 |
+
"kpack": 2
|
45 |
+
},
|
46 |
+
"16": {
|
47 |
+
"BLOCK_SIZE_M": 32,
|
48 |
+
"BLOCK_SIZE_N": 64,
|
49 |
+
"BLOCK_SIZE_K": 64,
|
50 |
+
"GROUP_SIZE_M": 1,
|
51 |
+
"num_warps": 4,
|
52 |
+
"num_stages": 2,
|
53 |
+
"waves_per_eu": 0,
|
54 |
+
"matrix_instr_nonkdim": 16,
|
55 |
+
"kpack": 1
|
56 |
+
},
|
57 |
+
"24": {
|
58 |
+
"BLOCK_SIZE_M": 16,
|
59 |
+
"BLOCK_SIZE_N": 64,
|
60 |
+
"BLOCK_SIZE_K": 64,
|
61 |
+
"GROUP_SIZE_M": 1,
|
62 |
+
"num_warps": 2,
|
63 |
+
"num_stages": 2,
|
64 |
+
"waves_per_eu": 0,
|
65 |
+
"matrix_instr_nonkdim": 16,
|
66 |
+
"kpack": 2
|
67 |
+
},
|
68 |
+
"32": {
|
69 |
+
"BLOCK_SIZE_M": 16,
|
70 |
+
"BLOCK_SIZE_N": 64,
|
71 |
+
"BLOCK_SIZE_K": 64,
|
72 |
+
"GROUP_SIZE_M": 1,
|
73 |
+
"num_warps": 2,
|
74 |
+
"num_stages": 2,
|
75 |
+
"waves_per_eu": 0,
|
76 |
+
"matrix_instr_nonkdim": 16,
|
77 |
+
"kpack": 2
|
78 |
+
},
|
79 |
+
"48": {
|
80 |
+
"BLOCK_SIZE_M": 16,
|
81 |
+
"BLOCK_SIZE_N": 32,
|
82 |
+
"BLOCK_SIZE_K": 64,
|
83 |
+
"GROUP_SIZE_M": 8,
|
84 |
+
"num_warps": 1,
|
85 |
+
"num_stages": 2,
|
86 |
+
"waves_per_eu": 0,
|
87 |
+
"matrix_instr_nonkdim": 16,
|
88 |
+
"kpack": 1
|
89 |
+
},
|
90 |
+
"64": {
|
91 |
+
"BLOCK_SIZE_M": 16,
|
92 |
+
"BLOCK_SIZE_N": 32,
|
93 |
+
"BLOCK_SIZE_K": 64,
|
94 |
+
"GROUP_SIZE_M": 8,
|
95 |
+
"num_warps": 2,
|
96 |
+
"num_stages": 2,
|
97 |
+
"waves_per_eu": 0,
|
98 |
+
"matrix_instr_nonkdim": 16,
|
99 |
+
"kpack": 1
|
100 |
+
},
|
101 |
+
"96": {
|
102 |
+
"BLOCK_SIZE_M": 16,
|
103 |
+
"BLOCK_SIZE_N": 32,
|
104 |
+
"BLOCK_SIZE_K": 64,
|
105 |
+
"GROUP_SIZE_M": 16,
|
106 |
+
"num_warps": 2,
|
107 |
+
"num_stages": 2,
|
108 |
+
"waves_per_eu": 0,
|
109 |
+
"matrix_instr_nonkdim": 16,
|
110 |
+
"kpack": 2
|
111 |
+
},
|
112 |
+
"128": {
|
113 |
+
"BLOCK_SIZE_M": 16,
|
114 |
+
"BLOCK_SIZE_N": 32,
|
115 |
+
"BLOCK_SIZE_K": 64,
|
116 |
+
"GROUP_SIZE_M": 4,
|
117 |
+
"num_warps": 2,
|
118 |
+
"num_stages": 2,
|
119 |
+
"waves_per_eu": 0,
|
120 |
+
"matrix_instr_nonkdim": 16,
|
121 |
+
"kpack": 2
|
122 |
+
},
|
123 |
+
"256": {
|
124 |
+
"BLOCK_SIZE_M": 32,
|
125 |
+
"BLOCK_SIZE_N": 32,
|
126 |
+
"BLOCK_SIZE_K": 64,
|
127 |
+
"GROUP_SIZE_M": 8,
|
128 |
+
"num_warps": 4,
|
129 |
+
"num_stages": 2,
|
130 |
+
"waves_per_eu": 0,
|
131 |
+
"matrix_instr_nonkdim": 16,
|
132 |
+
"kpack": 1
|
133 |
+
},
|
134 |
+
"512": {
|
135 |
+
"BLOCK_SIZE_M": 64,
|
136 |
+
"BLOCK_SIZE_N": 64,
|
137 |
+
"BLOCK_SIZE_K": 64,
|
138 |
+
"GROUP_SIZE_M": 32,
|
139 |
+
"num_warps": 4,
|
140 |
+
"num_stages": 2,
|
141 |
+
"waves_per_eu": 0,
|
142 |
+
"matrix_instr_nonkdim": 16,
|
143 |
+
"kpack": 2
|
144 |
+
},
|
145 |
+
"1024": {
|
146 |
+
"BLOCK_SIZE_M": 128,
|
147 |
+
"BLOCK_SIZE_N": 128,
|
148 |
+
"BLOCK_SIZE_K": 64,
|
149 |
+
"GROUP_SIZE_M": 4,
|
150 |
+
"num_warps": 4,
|
151 |
+
"num_stages": 2,
|
152 |
+
"waves_per_eu": 0,
|
153 |
+
"matrix_instr_nonkdim": 16,
|
154 |
+
"kpack": 2
|
155 |
+
},
|
156 |
+
"1536": {
|
157 |
+
"BLOCK_SIZE_M": 128,
|
158 |
+
"BLOCK_SIZE_N": 128,
|
159 |
+
"BLOCK_SIZE_K": 64,
|
160 |
+
"GROUP_SIZE_M": 16,
|
161 |
+
"num_warps": 4,
|
162 |
+
"num_stages": 2,
|
163 |
+
"waves_per_eu": 0,
|
164 |
+
"matrix_instr_nonkdim": 16,
|
165 |
+
"kpack": 1
|
166 |
+
},
|
167 |
+
"2048": {
|
168 |
+
"BLOCK_SIZE_M": 256,
|
169 |
+
"BLOCK_SIZE_N": 256,
|
170 |
+
"BLOCK_SIZE_K": 32,
|
171 |
+
"GROUP_SIZE_M": 16,
|
172 |
+
"num_warps": 8,
|
173 |
+
"num_stages": 2,
|
174 |
+
"waves_per_eu": 0,
|
175 |
+
"matrix_instr_nonkdim": 16,
|
176 |
+
"kpack": 1
|
177 |
+
},
|
178 |
+
"3072": {
|
179 |
+
"BLOCK_SIZE_M": 256,
|
180 |
+
"BLOCK_SIZE_N": 256,
|
181 |
+
"BLOCK_SIZE_K": 32,
|
182 |
+
"GROUP_SIZE_M": 16,
|
183 |
+
"num_warps": 8,
|
184 |
+
"num_stages": 2,
|
185 |
+
"waves_per_eu": 0,
|
186 |
+
"matrix_instr_nonkdim": 16,
|
187 |
+
"kpack": 1
|
188 |
+
},
|
189 |
+
"4096": {
|
190 |
+
"BLOCK_SIZE_M": 128,
|
191 |
+
"BLOCK_SIZE_N": 128,
|
192 |
+
"BLOCK_SIZE_K": 64,
|
193 |
+
"GROUP_SIZE_M": 8,
|
194 |
+
"num_warps": 4,
|
195 |
+
"num_stages": 2,
|
196 |
+
"waves_per_eu": 0,
|
197 |
+
"matrix_instr_nonkdim": 16,
|
198 |
+
"kpack": 2
|
199 |
+
}
|
200 |
+
}
|
build/torch25-cxx11-cu118-x86_64-linux/moe/configs/E=60,N=704,device_name=AMD_Instinct_MI300X.json
ADDED
@@ -0,0 +1,200 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"1": {
|
3 |
+
"BLOCK_SIZE_M": 16,
|
4 |
+
"BLOCK_SIZE_N": 16,
|
5 |
+
"BLOCK_SIZE_K": 128,
|
6 |
+
"GROUP_SIZE_M": 1,
|
7 |
+
"num_warps": 4,
|
8 |
+
"num_stages": 2,
|
9 |
+
"waves_per_eu": 0,
|
10 |
+
"matrix_instr_nonkdim": 16,
|
11 |
+
"kpack": 1
|
12 |
+
},
|
13 |
+
"2": {
|
14 |
+
"BLOCK_SIZE_M": 16,
|
15 |
+
"BLOCK_SIZE_N": 64,
|
16 |
+
"BLOCK_SIZE_K": 128,
|
17 |
+
"GROUP_SIZE_M": 1,
|
18 |
+
"num_warps": 4,
|
19 |
+
"num_stages": 2,
|
20 |
+
"waves_per_eu": 0,
|
21 |
+
"matrix_instr_nonkdim": 16,
|
22 |
+
"kpack": 2
|
23 |
+
},
|
24 |
+
"4": {
|
25 |
+
"BLOCK_SIZE_M": 16,
|
26 |
+
"BLOCK_SIZE_N": 128,
|
27 |
+
"BLOCK_SIZE_K": 128,
|
28 |
+
"GROUP_SIZE_M": 1,
|
29 |
+
"num_warps": 8,
|
30 |
+
"num_stages": 2,
|
31 |
+
"waves_per_eu": 0,
|
32 |
+
"matrix_instr_nonkdim": 16,
|
33 |
+
"kpack": 2
|
34 |
+
},
|
35 |
+
"8": {
|
36 |
+
"BLOCK_SIZE_M": 32,
|
37 |
+
"BLOCK_SIZE_N": 64,
|
38 |
+
"BLOCK_SIZE_K": 64,
|
39 |
+
"GROUP_SIZE_M": 1,
|
40 |
+
"num_warps": 4,
|
41 |
+
"num_stages": 2,
|
42 |
+
"waves_per_eu": 0,
|
43 |
+
"matrix_instr_nonkdim": 16,
|
44 |
+
"kpack": 1
|
45 |
+
},
|
46 |
+
"16": {
|
47 |
+
"BLOCK_SIZE_M": 16,
|
48 |
+
"BLOCK_SIZE_N": 16,
|
49 |
+
"BLOCK_SIZE_K": 64,
|
50 |
+
"GROUP_SIZE_M": 1,
|
51 |
+
"num_warps": 2,
|
52 |
+
"num_stages": 2,
|
53 |
+
"waves_per_eu": 0,
|
54 |
+
"matrix_instr_nonkdim": 16,
|
55 |
+
"kpack": 1
|
56 |
+
},
|
57 |
+
"24": {
|
58 |
+
"BLOCK_SIZE_M": 16,
|
59 |
+
"BLOCK_SIZE_N": 64,
|
60 |
+
"BLOCK_SIZE_K": 256,
|
61 |
+
"GROUP_SIZE_M": 4,
|
62 |
+
"num_warps": 4,
|
63 |
+
"num_stages": 2,
|
64 |
+
"waves_per_eu": 0,
|
65 |
+
"matrix_instr_nonkdim": 16,
|
66 |
+
"kpack": 2
|
67 |
+
},
|
68 |
+
"32": {
|
69 |
+
"BLOCK_SIZE_M": 16,
|
70 |
+
"BLOCK_SIZE_N": 64,
|
71 |
+
"BLOCK_SIZE_K": 256,
|
72 |
+
"GROUP_SIZE_M": 1,
|
73 |
+
"num_warps": 4,
|
74 |
+
"num_stages": 2,
|
75 |
+
"waves_per_eu": 0,
|
76 |
+
"matrix_instr_nonkdim": 16,
|
77 |
+
"kpack": 2
|
78 |
+
},
|
79 |
+
"48": {
|
80 |
+
"BLOCK_SIZE_M": 16,
|
81 |
+
"BLOCK_SIZE_N": 64,
|
82 |
+
"BLOCK_SIZE_K": 256,
|
83 |
+
"GROUP_SIZE_M": 4,
|
84 |
+
"num_warps": 4,
|
85 |
+
"num_stages": 2,
|
86 |
+
"waves_per_eu": 0,
|
87 |
+
"matrix_instr_nonkdim": 16,
|
88 |
+
"kpack": 2
|
89 |
+
},
|
90 |
+
"64": {
|
91 |
+
"BLOCK_SIZE_M": 16,
|
92 |
+
"BLOCK_SIZE_N": 64,
|
93 |
+
"BLOCK_SIZE_K": 256,
|
94 |
+
"GROUP_SIZE_M": 4,
|
95 |
+
"num_warps": 4,
|
96 |
+
"num_stages": 2,
|
97 |
+
"waves_per_eu": 0,
|
98 |
+
"matrix_instr_nonkdim": 16,
|
99 |
+
"kpack": 2
|
100 |
+
},
|
101 |
+
"96": {
|
102 |
+
"BLOCK_SIZE_M": 16,
|
103 |
+
"BLOCK_SIZE_N": 64,
|
104 |
+
"BLOCK_SIZE_K": 256,
|
105 |
+
"GROUP_SIZE_M": 4,
|
106 |
+
"num_warps": 4,
|
107 |
+
"num_stages": 2,
|
108 |
+
"waves_per_eu": 0,
|
109 |
+
"matrix_instr_nonkdim": 16,
|
110 |
+
"kpack": 2
|
111 |
+
},
|
112 |
+
"128": {
|
113 |
+
"BLOCK_SIZE_M": 16,
|
114 |
+
"BLOCK_SIZE_N": 64,
|
115 |
+
"BLOCK_SIZE_K": 256,
|
116 |
+
"GROUP_SIZE_M": 4,
|
117 |
+
"num_warps": 4,
|
118 |
+
"num_stages": 2,
|
119 |
+
"waves_per_eu": 0,
|
120 |
+
"matrix_instr_nonkdim": 16,
|
121 |
+
"kpack": 2
|
122 |
+
},
|
123 |
+
"256": {
|
124 |
+
"BLOCK_SIZE_M": 32,
|
125 |
+
"BLOCK_SIZE_N": 32,
|
126 |
+
"BLOCK_SIZE_K": 256,
|
127 |
+
"GROUP_SIZE_M": 4,
|
128 |
+
"num_warps": 4,
|
129 |
+
"num_stages": 2,
|
130 |
+
"waves_per_eu": 0,
|
131 |
+
"matrix_instr_nonkdim": 16,
|
132 |
+
"kpack": 2
|
133 |
+
},
|
134 |
+
"512": {
|
135 |
+
"BLOCK_SIZE_M": 64,
|
136 |
+
"BLOCK_SIZE_N": 64,
|
137 |
+
"BLOCK_SIZE_K": 128,
|
138 |
+
"GROUP_SIZE_M": 4,
|
139 |
+
"num_warps": 8,
|
140 |
+
"num_stages": 2,
|
141 |
+
"waves_per_eu": 0,
|
142 |
+
"matrix_instr_nonkdim": 16,
|
143 |
+
"kpack": 2
|
144 |
+
},
|
145 |
+
"1024": {
|
146 |
+
"BLOCK_SIZE_M": 64,
|
147 |
+
"BLOCK_SIZE_N": 64,
|
148 |
+
"BLOCK_SIZE_K": 128,
|
149 |
+
"GROUP_SIZE_M": 1,
|
150 |
+
"num_warps": 4,
|
151 |
+
"num_stages": 2,
|
152 |
+
"waves_per_eu": 0,
|
153 |
+
"matrix_instr_nonkdim": 16,
|
154 |
+
"kpack": 2
|
155 |
+
},
|
156 |
+
"1536": {
|
157 |
+
"BLOCK_SIZE_M": 128,
|
158 |
+
"BLOCK_SIZE_N": 128,
|
159 |
+
"BLOCK_SIZE_K": 64,
|
160 |
+
"GROUP_SIZE_M": 32,
|
161 |
+
"num_warps": 8,
|
162 |
+
"num_stages": 2,
|
163 |
+
"waves_per_eu": 0,
|
164 |
+
"matrix_instr_nonkdim": 16,
|
165 |
+
"kpack": 2
|
166 |
+
},
|
167 |
+
"2048": {
|
168 |
+
"BLOCK_SIZE_M": 128,
|
169 |
+
"BLOCK_SIZE_N": 128,
|
170 |
+
"BLOCK_SIZE_K": 64,
|
171 |
+
"GROUP_SIZE_M": 1,
|
172 |
+
"num_warps": 8,
|
173 |
+
"num_stages": 2,
|
174 |
+
"waves_per_eu": 0,
|
175 |
+
"matrix_instr_nonkdim": 16,
|
176 |
+
"kpack": 2
|
177 |
+
},
|
178 |
+
"3072": {
|
179 |
+
"BLOCK_SIZE_M": 128,
|
180 |
+
"BLOCK_SIZE_N": 128,
|
181 |
+
"BLOCK_SIZE_K": 64,
|
182 |
+
"GROUP_SIZE_M": 8,
|
183 |
+
"num_warps": 8,
|
184 |
+
"num_stages": 2,
|
185 |
+
"waves_per_eu": 0,
|
186 |
+
"matrix_instr_nonkdim": 16,
|
187 |
+
"kpack": 2
|
188 |
+
},
|
189 |
+
"4096": {
|
190 |
+
"BLOCK_SIZE_M": 128,
|
191 |
+
"BLOCK_SIZE_N": 128,
|
192 |
+
"BLOCK_SIZE_K": 64,
|
193 |
+
"GROUP_SIZE_M": 8,
|
194 |
+
"num_warps": 8,
|
195 |
+
"num_stages": 2,
|
196 |
+
"waves_per_eu": 0,
|
197 |
+
"matrix_instr_nonkdim": 16,
|
198 |
+
"kpack": 2
|
199 |
+
}
|
200 |
+
}
|
build/torch25-cxx11-cu118-x86_64-linux/moe/configs/E=64,N=1280,device_name=NVIDIA_A800-SXM4-80GB.json
ADDED
@@ -0,0 +1,146 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"1": {
|
3 |
+
"BLOCK_SIZE_M": 16,
|
4 |
+
"BLOCK_SIZE_N": 32,
|
5 |
+
"BLOCK_SIZE_K": 64,
|
6 |
+
"GROUP_SIZE_M": 16,
|
7 |
+
"num_warps": 4,
|
8 |
+
"num_stages": 5
|
9 |
+
},
|
10 |
+
"2": {
|
11 |
+
"BLOCK_SIZE_M": 16,
|
12 |
+
"BLOCK_SIZE_N": 128,
|
13 |
+
"BLOCK_SIZE_K": 64,
|
14 |
+
"GROUP_SIZE_M": 64,
|
15 |
+
"num_warps": 8,
|
16 |
+
"num_stages": 5
|
17 |
+
},
|
18 |
+
"4": {
|
19 |
+
"BLOCK_SIZE_M": 16,
|
20 |
+
"BLOCK_SIZE_N": 64,
|
21 |
+
"BLOCK_SIZE_K": 64,
|
22 |
+
"GROUP_SIZE_M": 16,
|
23 |
+
"num_warps": 4,
|
24 |
+
"num_stages": 4
|
25 |
+
},
|
26 |
+
"8": {
|
27 |
+
"BLOCK_SIZE_M": 16,
|
28 |
+
"BLOCK_SIZE_N": 128,
|
29 |
+
"BLOCK_SIZE_K": 128,
|
30 |
+
"GROUP_SIZE_M": 32,
|
31 |
+
"num_warps": 4,
|
32 |
+
"num_stages": 3
|
33 |
+
},
|
34 |
+
"16": {
|
35 |
+
"BLOCK_SIZE_M": 16,
|
36 |
+
"BLOCK_SIZE_N": 64,
|
37 |
+
"BLOCK_SIZE_K": 128,
|
38 |
+
"GROUP_SIZE_M": 64,
|
39 |
+
"num_warps": 4,
|
40 |
+
"num_stages": 3
|
41 |
+
},
|
42 |
+
"24": {
|
43 |
+
"BLOCK_SIZE_M": 32,
|
44 |
+
"BLOCK_SIZE_N": 128,
|
45 |
+
"BLOCK_SIZE_K": 128,
|
46 |
+
"GROUP_SIZE_M": 1,
|
47 |
+
"num_warps": 4,
|
48 |
+
"num_stages": 3
|
49 |
+
},
|
50 |
+
"32": {
|
51 |
+
"BLOCK_SIZE_M": 16,
|
52 |
+
"BLOCK_SIZE_N": 128,
|
53 |
+
"BLOCK_SIZE_K": 128,
|
54 |
+
"GROUP_SIZE_M": 1,
|
55 |
+
"num_warps": 4,
|
56 |
+
"num_stages": 3
|
57 |
+
},
|
58 |
+
"48": {
|
59 |
+
"BLOCK_SIZE_M": 16,
|
60 |
+
"BLOCK_SIZE_N": 128,
|
61 |
+
"BLOCK_SIZE_K": 128,
|
62 |
+
"GROUP_SIZE_M": 1,
|
63 |
+
"num_warps": 4,
|
64 |
+
"num_stages": 3
|
65 |
+
},
|
66 |
+
"64": {
|
67 |
+
"BLOCK_SIZE_M": 16,
|
68 |
+
"BLOCK_SIZE_N": 128,
|
69 |
+
"BLOCK_SIZE_K": 256,
|
70 |
+
"GROUP_SIZE_M": 1,
|
71 |
+
"num_warps": 4,
|
72 |
+
"num_stages": 2
|
73 |
+
},
|
74 |
+
"96": {
|
75 |
+
"BLOCK_SIZE_M": 32,
|
76 |
+
"BLOCK_SIZE_N": 128,
|
77 |
+
"BLOCK_SIZE_K": 128,
|
78 |
+
"GROUP_SIZE_M": 1,
|
79 |
+
"num_warps": 4,
|
80 |
+
"num_stages": 3
|
81 |
+
},
|
82 |
+
"128": {
|
83 |
+
"BLOCK_SIZE_M": 32,
|
84 |
+
"BLOCK_SIZE_N": 128,
|
85 |
+
"BLOCK_SIZE_K": 128,
|
86 |
+
"GROUP_SIZE_M": 1,
|
87 |
+
"num_warps": 4,
|
88 |
+
"num_stages": 3
|
89 |
+
},
|
90 |
+
"256": {
|
91 |
+
"BLOCK_SIZE_M": 64,
|
92 |
+
"BLOCK_SIZE_N": 256,
|
93 |
+
"BLOCK_SIZE_K": 128,
|
94 |
+
"GROUP_SIZE_M": 1,
|
95 |
+
"num_warps": 8,
|
96 |
+
"num_stages": 3
|
97 |
+
},
|
98 |
+
"512": {
|
99 |
+
"BLOCK_SIZE_M": 64,
|
100 |
+
"BLOCK_SIZE_N": 256,
|
101 |
+
"BLOCK_SIZE_K": 128,
|
102 |
+
"GROUP_SIZE_M": 1,
|
103 |
+
"num_warps": 8,
|
104 |
+
"num_stages": 3
|
105 |
+
},
|
106 |
+
"1024": {
|
107 |
+
"BLOCK_SIZE_M": 64,
|
108 |
+
"BLOCK_SIZE_N": 256,
|
109 |
+
"BLOCK_SIZE_K": 128,
|
110 |
+
"GROUP_SIZE_M": 1,
|
111 |
+
"num_warps": 8,
|
112 |
+
"num_stages": 3
|
113 |
+
},
|
114 |
+
"1536": {
|
115 |
+
"BLOCK_SIZE_M": 128,
|
116 |
+
"BLOCK_SIZE_N": 256,
|
117 |
+
"BLOCK_SIZE_K": 64,
|
118 |
+
"GROUP_SIZE_M": 1,
|
119 |
+
"num_warps": 8,
|
120 |
+
"num_stages": 4
|
121 |
+
},
|
122 |
+
"2048": {
|
123 |
+
"BLOCK_SIZE_M": 128,
|
124 |
+
"BLOCK_SIZE_N": 128,
|
125 |
+
"BLOCK_SIZE_K": 64,
|
126 |
+
"GROUP_SIZE_M": 1,
|
127 |
+
"num_warps": 4,
|
128 |
+
"num_stages": 3
|
129 |
+
},
|
130 |
+
"3072": {
|
131 |
+
"BLOCK_SIZE_M": 128,
|
132 |
+
"BLOCK_SIZE_N": 128,
|
133 |
+
"BLOCK_SIZE_K": 64,
|
134 |
+
"GROUP_SIZE_M": 1,
|
135 |
+
"num_warps": 4,
|
136 |
+
"num_stages": 3
|
137 |
+
},
|
138 |
+
"4096": {
|
139 |
+
"BLOCK_SIZE_M": 128,
|
140 |
+
"BLOCK_SIZE_N": 128,
|
141 |
+
"BLOCK_SIZE_K": 64,
|
142 |
+
"GROUP_SIZE_M": 1,
|
143 |
+
"num_warps": 4,
|
144 |
+
"num_stages": 3
|
145 |
+
}
|
146 |
+
}
|
build/torch25-cxx11-cu118-x86_64-linux/moe/configs/E=64,N=1280,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8.json
ADDED
@@ -0,0 +1,146 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"1": {
|
3 |
+
"BLOCK_SIZE_M": 64,
|
4 |
+
"BLOCK_SIZE_N": 128,
|
5 |
+
"BLOCK_SIZE_K": 128,
|
6 |
+
"GROUP_SIZE_M": 32,
|
7 |
+
"num_warps": 4,
|
8 |
+
"num_stages": 4
|
9 |
+
},
|
10 |
+
"2": {
|
11 |
+
"BLOCK_SIZE_M": 64,
|
12 |
+
"BLOCK_SIZE_N": 64,
|
13 |
+
"BLOCK_SIZE_K": 128,
|
14 |
+
"GROUP_SIZE_M": 1,
|
15 |
+
"num_warps": 4,
|
16 |
+
"num_stages": 4
|
17 |
+
},
|
18 |
+
"4": {
|
19 |
+
"BLOCK_SIZE_M": 64,
|
20 |
+
"BLOCK_SIZE_N": 64,
|
21 |
+
"BLOCK_SIZE_K": 128,
|
22 |
+
"GROUP_SIZE_M": 1,
|
23 |
+
"num_warps": 4,
|
24 |
+
"num_stages": 3
|
25 |
+
},
|
26 |
+
"8": {
|
27 |
+
"BLOCK_SIZE_M": 64,
|
28 |
+
"BLOCK_SIZE_N": 128,
|
29 |
+
"BLOCK_SIZE_K": 128,
|
30 |
+
"GROUP_SIZE_M": 32,
|
31 |
+
"num_warps": 8,
|
32 |
+
"num_stages": 4
|
33 |
+
},
|
34 |
+
"16": {
|
35 |
+
"BLOCK_SIZE_M": 64,
|
36 |
+
"BLOCK_SIZE_N": 128,
|
37 |
+
"BLOCK_SIZE_K": 128,
|
38 |
+
"GROUP_SIZE_M": 64,
|
39 |
+
"num_warps": 8,
|
40 |
+
"num_stages": 3
|
41 |
+
},
|
42 |
+
"24": {
|
43 |
+
"BLOCK_SIZE_M": 64,
|
44 |
+
"BLOCK_SIZE_N": 64,
|
45 |
+
"BLOCK_SIZE_K": 128,
|
46 |
+
"GROUP_SIZE_M": 1,
|
47 |
+
"num_warps": 4,
|
48 |
+
"num_stages": 3
|
49 |
+
},
|
50 |
+
"32": {
|
51 |
+
"BLOCK_SIZE_M": 64,
|
52 |
+
"BLOCK_SIZE_N": 64,
|
53 |
+
"BLOCK_SIZE_K": 128,
|
54 |
+
"GROUP_SIZE_M": 16,
|
55 |
+
"num_warps": 4,
|
56 |
+
"num_stages": 3
|
57 |
+
},
|
58 |
+
"48": {
|
59 |
+
"BLOCK_SIZE_M": 64,
|
60 |
+
"BLOCK_SIZE_N": 64,
|
61 |
+
"BLOCK_SIZE_K": 128,
|
62 |
+
"GROUP_SIZE_M": 16,
|
63 |
+
"num_warps": 4,
|
64 |
+
"num_stages": 3
|
65 |
+
},
|
66 |
+
"64": {
|
67 |
+
"BLOCK_SIZE_M": 64,
|
68 |
+
"BLOCK_SIZE_N": 64,
|
69 |
+
"BLOCK_SIZE_K": 128,
|
70 |
+
"GROUP_SIZE_M": 1,
|
71 |
+
"num_warps": 4,
|
72 |
+
"num_stages": 3
|
73 |
+
},
|
74 |
+
"96": {
|
75 |
+
"BLOCK_SIZE_M": 64,
|
76 |
+
"BLOCK_SIZE_N": 64,
|
77 |
+
"BLOCK_SIZE_K": 128,
|
78 |
+
"GROUP_SIZE_M": 1,
|
79 |
+
"num_warps": 4,
|
80 |
+
"num_stages": 3
|
81 |
+
},
|
82 |
+
"128": {
|
83 |
+
"BLOCK_SIZE_M": 64,
|
84 |
+
"BLOCK_SIZE_N": 64,
|
85 |
+
"BLOCK_SIZE_K": 128,
|
86 |
+
"GROUP_SIZE_M": 1,
|
87 |
+
"num_warps": 4,
|
88 |
+
"num_stages": 3
|
89 |
+
},
|
90 |
+
"256": {
|
91 |
+
"BLOCK_SIZE_M": 64,
|
92 |
+
"BLOCK_SIZE_N": 64,
|
93 |
+
"BLOCK_SIZE_K": 128,
|
94 |
+
"GROUP_SIZE_M": 1,
|
95 |
+
"num_warps": 4,
|
96 |
+
"num_stages": 3
|
97 |
+
},
|
98 |
+
"512": {
|
99 |
+
"BLOCK_SIZE_M": 64,
|
100 |
+
"BLOCK_SIZE_N": 128,
|
101 |
+
"BLOCK_SIZE_K": 128,
|
102 |
+
"GROUP_SIZE_M": 1,
|
103 |
+
"num_warps": 4,
|
104 |
+
"num_stages": 3
|
105 |
+
},
|
106 |
+
"1024": {
|
107 |
+
"BLOCK_SIZE_M": 128,
|
108 |
+
"BLOCK_SIZE_N": 256,
|
109 |
+
"BLOCK_SIZE_K": 128,
|
110 |
+
"GROUP_SIZE_M": 1,
|
111 |
+
"num_warps": 8,
|
112 |
+
"num_stages": 4
|
113 |
+
},
|
114 |
+
"1536": {
|
115 |
+
"BLOCK_SIZE_M": 128,
|
116 |
+
"BLOCK_SIZE_N": 256,
|
117 |
+
"BLOCK_SIZE_K": 128,
|
118 |
+
"GROUP_SIZE_M": 1,
|
119 |
+
"num_warps": 8,
|
120 |
+
"num_stages": 4
|
121 |
+
},
|
122 |
+
"2048": {
|
123 |
+
"BLOCK_SIZE_M": 128,
|
124 |
+
"BLOCK_SIZE_N": 256,
|
125 |
+
"BLOCK_SIZE_K": 128,
|
126 |
+
"GROUP_SIZE_M": 1,
|
127 |
+
"num_warps": 8,
|
128 |
+
"num_stages": 4
|
129 |
+
},
|
130 |
+
"3072": {
|
131 |
+
"BLOCK_SIZE_M": 128,
|
132 |
+
"BLOCK_SIZE_N": 256,
|
133 |
+
"BLOCK_SIZE_K": 128,
|
134 |
+
"GROUP_SIZE_M": 1,
|
135 |
+
"num_warps": 8,
|
136 |
+
"num_stages": 4
|
137 |
+
},
|
138 |
+
"4096": {
|
139 |
+
"BLOCK_SIZE_M": 128,
|
140 |
+
"BLOCK_SIZE_N": 256,
|
141 |
+
"BLOCK_SIZE_K": 128,
|
142 |
+
"GROUP_SIZE_M": 64,
|
143 |
+
"num_warps": 8,
|
144 |
+
"num_stages": 4
|
145 |
+
}
|
146 |
+
}
|
build/torch25-cxx11-cu118-x86_64-linux/moe/configs/E=64,N=1280,device_name=NVIDIA_H200,dtype=fp8_w8a8.json
ADDED
@@ -0,0 +1,146 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"1": {
|
3 |
+
"BLOCK_SIZE_M": 64,
|
4 |
+
"BLOCK_SIZE_N": 128,
|
5 |
+
"BLOCK_SIZE_K": 128,
|
6 |
+
"GROUP_SIZE_M": 1,
|
7 |
+
"num_warps": 4,
|
8 |
+
"num_stages": 4
|
9 |
+
},
|
10 |
+
"2": {
|
11 |
+
"BLOCK_SIZE_M": 64,
|
12 |
+
"BLOCK_SIZE_N": 128,
|
13 |
+
"BLOCK_SIZE_K": 128,
|
14 |
+
"GROUP_SIZE_M": 1,
|
15 |
+
"num_warps": 4,
|
16 |
+
"num_stages": 3
|
17 |
+
},
|
18 |
+
"4": {
|
19 |
+
"BLOCK_SIZE_M": 64,
|
20 |
+
"BLOCK_SIZE_N": 64,
|
21 |
+
"BLOCK_SIZE_K": 128,
|
22 |
+
"GROUP_SIZE_M": 64,
|
23 |
+
"num_warps": 4,
|
24 |
+
"num_stages": 3
|
25 |
+
},
|
26 |
+
"8": {
|
27 |
+
"BLOCK_SIZE_M": 64,
|
28 |
+
"BLOCK_SIZE_N": 128,
|
29 |
+
"BLOCK_SIZE_K": 128,
|
30 |
+
"GROUP_SIZE_M": 64,
|
31 |
+
"num_warps": 4,
|
32 |
+
"num_stages": 3
|
33 |
+
},
|
34 |
+
"16": {
|
35 |
+
"BLOCK_SIZE_M": 64,
|
36 |
+
"BLOCK_SIZE_N": 128,
|
37 |
+
"BLOCK_SIZE_K": 128,
|
38 |
+
"GROUP_SIZE_M": 1,
|
39 |
+
"num_warps": 4,
|
40 |
+
"num_stages": 3
|
41 |
+
},
|
42 |
+
"24": {
|
43 |
+
"BLOCK_SIZE_M": 64,
|
44 |
+
"BLOCK_SIZE_N": 128,
|
45 |
+
"BLOCK_SIZE_K": 128,
|
46 |
+
"GROUP_SIZE_M": 16,
|
47 |
+
"num_warps": 4,
|
48 |
+
"num_stages": 4
|
49 |
+
},
|
50 |
+
"32": {
|
51 |
+
"BLOCK_SIZE_M": 64,
|
52 |
+
"BLOCK_SIZE_N": 256,
|
53 |
+
"BLOCK_SIZE_K": 128,
|
54 |
+
"GROUP_SIZE_M": 1,
|
55 |
+
"num_warps": 4,
|
56 |
+
"num_stages": 5
|
57 |
+
},
|
58 |
+
"48": {
|
59 |
+
"BLOCK_SIZE_M": 64,
|
60 |
+
"BLOCK_SIZE_N": 128,
|
61 |
+
"BLOCK_SIZE_K": 128,
|
62 |
+
"GROUP_SIZE_M": 1,
|
63 |
+
"num_warps": 4,
|
64 |
+
"num_stages": 4
|
65 |
+
},
|
66 |
+
"64": {
|
67 |
+
"BLOCK_SIZE_M": 64,
|
68 |
+
"BLOCK_SIZE_N": 128,
|
69 |
+
"BLOCK_SIZE_K": 128,
|
70 |
+
"GROUP_SIZE_M": 1,
|
71 |
+
"num_warps": 4,
|
72 |
+
"num_stages": 4
|
73 |
+
},
|
74 |
+
"96": {
|
75 |
+
"BLOCK_SIZE_M": 64,
|
76 |
+
"BLOCK_SIZE_N": 128,
|
77 |
+
"BLOCK_SIZE_K": 128,
|
78 |
+
"GROUP_SIZE_M": 1,
|
79 |
+
"num_warps": 4,
|
80 |
+
"num_stages": 4
|
81 |
+
},
|
82 |
+
"128": {
|
83 |
+
"BLOCK_SIZE_M": 64,
|
84 |
+
"BLOCK_SIZE_N": 128,
|
85 |
+
"BLOCK_SIZE_K": 128,
|
86 |
+
"GROUP_SIZE_M": 1,
|
87 |
+
"num_warps": 4,
|
88 |
+
"num_stages": 4
|
89 |
+
},
|
90 |
+
"256": {
|
91 |
+
"BLOCK_SIZE_M": 64,
|
92 |
+
"BLOCK_SIZE_N": 128,
|
93 |
+
"BLOCK_SIZE_K": 128,
|
94 |
+
"GROUP_SIZE_M": 1,
|
95 |
+
"num_warps": 4,
|
96 |
+
"num_stages": 4
|
97 |
+
},
|
98 |
+
"512": {
|
99 |
+
"BLOCK_SIZE_M": 128,
|
100 |
+
"BLOCK_SIZE_N": 256,
|
101 |
+
"BLOCK_SIZE_K": 128,
|
102 |
+
"GROUP_SIZE_M": 1,
|
103 |
+
"num_warps": 8,
|
104 |
+
"num_stages": 4
|
105 |
+
},
|
106 |
+
"1024": {
|
107 |
+
"BLOCK_SIZE_M": 128,
|
108 |
+
"BLOCK_SIZE_N": 256,
|
109 |
+
"BLOCK_SIZE_K": 128,
|
110 |
+
"GROUP_SIZE_M": 1,
|
111 |
+
"num_warps": 8,
|
112 |
+
"num_stages": 4
|
113 |
+
},
|
114 |
+
"1536": {
|
115 |
+
"BLOCK_SIZE_M": 128,
|
116 |
+
"BLOCK_SIZE_N": 256,
|
117 |
+
"BLOCK_SIZE_K": 128,
|
118 |
+
"GROUP_SIZE_M": 16,
|
119 |
+
"num_warps": 8,
|
120 |
+
"num_stages": 4
|
121 |
+
},
|
122 |
+
"2048": {
|
123 |
+
"BLOCK_SIZE_M": 128,
|
124 |
+
"BLOCK_SIZE_N": 256,
|
125 |
+
"BLOCK_SIZE_K": 128,
|
126 |
+
"GROUP_SIZE_M": 16,
|
127 |
+
"num_warps": 8,
|
128 |
+
"num_stages": 4
|
129 |
+
},
|
130 |
+
"3072": {
|
131 |
+
"BLOCK_SIZE_M": 128,
|
132 |
+
"BLOCK_SIZE_N": 256,
|
133 |
+
"BLOCK_SIZE_K": 128,
|
134 |
+
"GROUP_SIZE_M": 1,
|
135 |
+
"num_warps": 8,
|
136 |
+
"num_stages": 4
|
137 |
+
},
|
138 |
+
"4096": {
|
139 |
+
"BLOCK_SIZE_M": 128,
|
140 |
+
"BLOCK_SIZE_N": 256,
|
141 |
+
"BLOCK_SIZE_K": 128,
|
142 |
+
"GROUP_SIZE_M": 16,
|
143 |
+
"num_warps": 8,
|
144 |
+
"num_stages": 4
|
145 |
+
}
|
146 |
+
}
|
build/torch25-cxx11-cu118-x86_64-linux/moe/configs/E=64,N=1280,device_name=NVIDIA_H200.json
ADDED
@@ -0,0 +1,146 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"1": {
|
3 |
+
"BLOCK_SIZE_M": 16,
|
4 |
+
"BLOCK_SIZE_N": 64,
|
5 |
+
"BLOCK_SIZE_K": 64,
|
6 |
+
"GROUP_SIZE_M": 1,
|
7 |
+
"num_warps": 4,
|
8 |
+
"num_stages": 4
|
9 |
+
},
|
10 |
+
"2": {
|
11 |
+
"BLOCK_SIZE_M": 16,
|
12 |
+
"BLOCK_SIZE_N": 128,
|
13 |
+
"BLOCK_SIZE_K": 128,
|
14 |
+
"GROUP_SIZE_M": 16,
|
15 |
+
"num_warps": 4,
|
16 |
+
"num_stages": 2
|
17 |
+
},
|
18 |
+
"4": {
|
19 |
+
"BLOCK_SIZE_M": 16,
|
20 |
+
"BLOCK_SIZE_N": 64,
|
21 |
+
"BLOCK_SIZE_K": 128,
|
22 |
+
"GROUP_SIZE_M": 32,
|
23 |
+
"num_warps": 4,
|
24 |
+
"num_stages": 5
|
25 |
+
},
|
26 |
+
"8": {
|
27 |
+
"BLOCK_SIZE_M": 16,
|
28 |
+
"BLOCK_SIZE_N": 128,
|
29 |
+
"BLOCK_SIZE_K": 128,
|
30 |
+
"GROUP_SIZE_M": 32,
|
31 |
+
"num_warps": 4,
|
32 |
+
"num_stages": 3
|
33 |
+
},
|
34 |
+
"16": {
|
35 |
+
"BLOCK_SIZE_M": 16,
|
36 |
+
"BLOCK_SIZE_N": 64,
|
37 |
+
"BLOCK_SIZE_K": 128,
|
38 |
+
"GROUP_SIZE_M": 16,
|
39 |
+
"num_warps": 4,
|
40 |
+
"num_stages": 2
|
41 |
+
},
|
42 |
+
"24": {
|
43 |
+
"BLOCK_SIZE_M": 16,
|
44 |
+
"BLOCK_SIZE_N": 128,
|
45 |
+
"BLOCK_SIZE_K": 256,
|
46 |
+
"GROUP_SIZE_M": 32,
|
47 |
+
"num_warps": 4,
|
48 |
+
"num_stages": 2
|
49 |
+
},
|
50 |
+
"32": {
|
51 |
+
"BLOCK_SIZE_M": 16,
|
52 |
+
"BLOCK_SIZE_N": 128,
|
53 |
+
"BLOCK_SIZE_K": 128,
|
54 |
+
"GROUP_SIZE_M": 1,
|
55 |
+
"num_warps": 4,
|
56 |
+
"num_stages": 4
|
57 |
+
},
|
58 |
+
"48": {
|
59 |
+
"BLOCK_SIZE_M": 16,
|
60 |
+
"BLOCK_SIZE_N": 128,
|
61 |
+
"BLOCK_SIZE_K": 128,
|
62 |
+
"GROUP_SIZE_M": 32,
|
63 |
+
"num_warps": 8,
|
64 |
+
"num_stages": 3
|
65 |
+
},
|
66 |
+
"64": {
|
67 |
+
"BLOCK_SIZE_M": 16,
|
68 |
+
"BLOCK_SIZE_N": 128,
|
69 |
+
"BLOCK_SIZE_K": 128,
|
70 |
+
"GROUP_SIZE_M": 1,
|
71 |
+
"num_warps": 8,
|
72 |
+
"num_stages": 3
|
73 |
+
},
|
74 |
+
"96": {
|
75 |
+
"BLOCK_SIZE_M": 32,
|
76 |
+
"BLOCK_SIZE_N": 128,
|
77 |
+
"BLOCK_SIZE_K": 128,
|
78 |
+
"GROUP_SIZE_M": 1,
|
79 |
+
"num_warps": 4,
|
80 |
+
"num_stages": 3
|
81 |
+
},
|
82 |
+
"128": {
|
83 |
+
"BLOCK_SIZE_M": 32,
|
84 |
+
"BLOCK_SIZE_N": 128,
|
85 |
+
"BLOCK_SIZE_K": 128,
|
86 |
+
"GROUP_SIZE_M": 1,
|
87 |
+
"num_warps": 4,
|
88 |
+
"num_stages": 3
|
89 |
+
},
|
90 |
+
"256": {
|
91 |
+
"BLOCK_SIZE_M": 64,
|
92 |
+
"BLOCK_SIZE_N": 64,
|
93 |
+
"BLOCK_SIZE_K": 64,
|
94 |
+
"GROUP_SIZE_M": 1,
|
95 |
+
"num_warps": 4,
|
96 |
+
"num_stages": 3
|
97 |
+
},
|
98 |
+
"512": {
|
99 |
+
"BLOCK_SIZE_M": 128,
|
100 |
+
"BLOCK_SIZE_N": 256,
|
101 |
+
"BLOCK_SIZE_K": 64,
|
102 |
+
"GROUP_SIZE_M": 1,
|
103 |
+
"num_warps": 8,
|
104 |
+
"num_stages": 4
|
105 |
+
},
|
106 |
+
"1024": {
|
107 |
+
"BLOCK_SIZE_M": 128,
|
108 |
+
"BLOCK_SIZE_N": 256,
|
109 |
+
"BLOCK_SIZE_K": 64,
|
110 |
+
"GROUP_SIZE_M": 1,
|
111 |
+
"num_warps": 8,
|
112 |
+
"num_stages": 4
|
113 |
+
},
|
114 |
+
"1536": {
|
115 |
+
"BLOCK_SIZE_M": 128,
|
116 |
+
"BLOCK_SIZE_N": 256,
|
117 |
+
"BLOCK_SIZE_K": 64,
|
118 |
+
"GROUP_SIZE_M": 1,
|
119 |
+
"num_warps": 8,
|
120 |
+
"num_stages": 4
|
121 |
+
},
|
122 |
+
"2048": {
|
123 |
+
"BLOCK_SIZE_M": 128,
|
124 |
+
"BLOCK_SIZE_N": 256,
|
125 |
+
"BLOCK_SIZE_K": 64,
|
126 |
+
"GROUP_SIZE_M": 1,
|
127 |
+
"num_warps": 8,
|
128 |
+
"num_stages": 4
|
129 |
+
},
|
130 |
+
"3072": {
|
131 |
+
"BLOCK_SIZE_M": 128,
|
132 |
+
"BLOCK_SIZE_N": 256,
|
133 |
+
"BLOCK_SIZE_K": 64,
|
134 |
+
"GROUP_SIZE_M": 16,
|
135 |
+
"num_warps": 8,
|
136 |
+
"num_stages": 4
|
137 |
+
},
|
138 |
+
"4096": {
|
139 |
+
"BLOCK_SIZE_M": 128,
|
140 |
+
"BLOCK_SIZE_N": 256,
|
141 |
+
"BLOCK_SIZE_K": 64,
|
142 |
+
"GROUP_SIZE_M": 16,
|
143 |
+
"num_warps": 8,
|
144 |
+
"num_stages": 4
|
145 |
+
}
|
146 |
+
}
|
build/torch25-cxx11-cu118-x86_64-linux/moe/configs/E=64,N=2560,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8.json
ADDED
@@ -0,0 +1,146 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"1": {
|
3 |
+
"BLOCK_SIZE_M": 64,
|
4 |
+
"BLOCK_SIZE_N": 128,
|
5 |
+
"BLOCK_SIZE_K": 128,
|
6 |
+
"GROUP_SIZE_M": 64,
|
7 |
+
"num_warps": 4,
|
8 |
+
"num_stages": 3
|
9 |
+
},
|
10 |
+
"2": {
|
11 |
+
"BLOCK_SIZE_M": 64,
|
12 |
+
"BLOCK_SIZE_N": 128,
|
13 |
+
"BLOCK_SIZE_K": 256,
|
14 |
+
"GROUP_SIZE_M": 64,
|
15 |
+
"num_warps": 8,
|
16 |
+
"num_stages": 4
|
17 |
+
},
|
18 |
+
"4": {
|
19 |
+
"BLOCK_SIZE_M": 64,
|
20 |
+
"BLOCK_SIZE_N": 128,
|
21 |
+
"BLOCK_SIZE_K": 128,
|
22 |
+
"GROUP_SIZE_M": 16,
|
23 |
+
"num_warps": 4,
|
24 |
+
"num_stages": 5
|
25 |
+
},
|
26 |
+
"8": {
|
27 |
+
"BLOCK_SIZE_M": 64,
|
28 |
+
"BLOCK_SIZE_N": 128,
|
29 |
+
"BLOCK_SIZE_K": 128,
|
30 |
+
"GROUP_SIZE_M": 16,
|
31 |
+
"num_warps": 4,
|
32 |
+
"num_stages": 4
|
33 |
+
},
|
34 |
+
"16": {
|
35 |
+
"BLOCK_SIZE_M": 64,
|
36 |
+
"BLOCK_SIZE_N": 128,
|
37 |
+
"BLOCK_SIZE_K": 128,
|
38 |
+
"GROUP_SIZE_M": 16,
|
39 |
+
"num_warps": 4,
|
40 |
+
"num_stages": 3
|
41 |
+
},
|
42 |
+
"24": {
|
43 |
+
"BLOCK_SIZE_M": 64,
|
44 |
+
"BLOCK_SIZE_N": 128,
|
45 |
+
"BLOCK_SIZE_K": 256,
|
46 |
+
"GROUP_SIZE_M": 1,
|
47 |
+
"num_warps": 8,
|
48 |
+
"num_stages": 4
|
49 |
+
},
|
50 |
+
"32": {
|
51 |
+
"BLOCK_SIZE_M": 64,
|
52 |
+
"BLOCK_SIZE_N": 128,
|
53 |
+
"BLOCK_SIZE_K": 256,
|
54 |
+
"GROUP_SIZE_M": 16,
|
55 |
+
"num_warps": 4,
|
56 |
+
"num_stages": 4
|
57 |
+
},
|
58 |
+
"48": {
|
59 |
+
"BLOCK_SIZE_M": 64,
|
60 |
+
"BLOCK_SIZE_N": 128,
|
61 |
+
"BLOCK_SIZE_K": 256,
|
62 |
+
"GROUP_SIZE_M": 1,
|
63 |
+
"num_warps": 8,
|
64 |
+
"num_stages": 4
|
65 |
+
},
|
66 |
+
"64": {
|
67 |
+
"BLOCK_SIZE_M": 64,
|
68 |
+
"BLOCK_SIZE_N": 64,
|
69 |
+
"BLOCK_SIZE_K": 128,
|
70 |
+
"GROUP_SIZE_M": 1,
|
71 |
+
"num_warps": 4,
|
72 |
+
"num_stages": 3
|
73 |
+
},
|
74 |
+
"96": {
|
75 |
+
"BLOCK_SIZE_M": 64,
|
76 |
+
"BLOCK_SIZE_N": 64,
|
77 |
+
"BLOCK_SIZE_K": 128,
|
78 |
+
"GROUP_SIZE_M": 1,
|
79 |
+
"num_warps": 4,
|
80 |
+
"num_stages": 3
|
81 |
+
},
|
82 |
+
"128": {
|
83 |
+
"BLOCK_SIZE_M": 64,
|
84 |
+
"BLOCK_SIZE_N": 64,
|
85 |
+
"BLOCK_SIZE_K": 128,
|
86 |
+
"GROUP_SIZE_M": 1,
|
87 |
+
"num_warps": 4,
|
88 |
+
"num_stages": 3
|
89 |
+
},
|
90 |
+
"256": {
|
91 |
+
"BLOCK_SIZE_M": 64,
|
92 |
+
"BLOCK_SIZE_N": 64,
|
93 |
+
"BLOCK_SIZE_K": 128,
|
94 |
+
"GROUP_SIZE_M": 1,
|
95 |
+
"num_warps": 4,
|
96 |
+
"num_stages": 3
|
97 |
+
},
|
98 |
+
"512": {
|
99 |
+
"BLOCK_SIZE_M": 128,
|
100 |
+
"BLOCK_SIZE_N": 128,
|
101 |
+
"BLOCK_SIZE_K": 128,
|
102 |
+
"GROUP_SIZE_M": 1,
|
103 |
+
"num_warps": 8,
|
104 |
+
"num_stages": 3
|
105 |
+
},
|
106 |
+
"1024": {
|
107 |
+
"BLOCK_SIZE_M": 128,
|
108 |
+
"BLOCK_SIZE_N": 256,
|
109 |
+
"BLOCK_SIZE_K": 128,
|
110 |
+
"GROUP_SIZE_M": 1,
|
111 |
+
"num_warps": 8,
|
112 |
+
"num_stages": 4
|
113 |
+
},
|
114 |
+
"1536": {
|
115 |
+
"BLOCK_SIZE_M": 128,
|
116 |
+
"BLOCK_SIZE_N": 256,
|
117 |
+
"BLOCK_SIZE_K": 128,
|
118 |
+
"GROUP_SIZE_M": 16,
|
119 |
+
"num_warps": 8,
|
120 |
+
"num_stages": 4
|
121 |
+
},
|
122 |
+
"2048": {
|
123 |
+
"BLOCK_SIZE_M": 128,
|
124 |
+
"BLOCK_SIZE_N": 256,
|
125 |
+
"BLOCK_SIZE_K": 128,
|
126 |
+
"GROUP_SIZE_M": 1,
|
127 |
+
"num_warps": 8,
|
128 |
+
"num_stages": 4
|
129 |
+
},
|
130 |
+
"3072": {
|
131 |
+
"BLOCK_SIZE_M": 128,
|
132 |
+
"BLOCK_SIZE_N": 256,
|
133 |
+
"BLOCK_SIZE_K": 128,
|
134 |
+
"GROUP_SIZE_M": 16,
|
135 |
+
"num_warps": 8,
|
136 |
+
"num_stages": 4
|
137 |
+
},
|
138 |
+
"4096": {
|
139 |
+
"BLOCK_SIZE_M": 128,
|
140 |
+
"BLOCK_SIZE_N": 256,
|
141 |
+
"BLOCK_SIZE_K": 128,
|
142 |
+
"GROUP_SIZE_M": 16,
|
143 |
+
"num_warps": 8,
|
144 |
+
"num_stages": 4
|
145 |
+
}
|
146 |
+
}
|
build/torch25-cxx11-cu118-x86_64-linux/moe/configs/E=64,N=2560,device_name=NVIDIA_H200,dtype=fp8_w8a8.json
ADDED
@@ -0,0 +1,146 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"1": {
|
3 |
+
"BLOCK_SIZE_M": 64,
|
4 |
+
"BLOCK_SIZE_N": 128,
|
5 |
+
"BLOCK_SIZE_K": 128,
|
6 |
+
"GROUP_SIZE_M": 16,
|
7 |
+
"num_warps": 4,
|
8 |
+
"num_stages": 3
|
9 |
+
},
|
10 |
+
"2": {
|
11 |
+
"BLOCK_SIZE_M": 64,
|
12 |
+
"BLOCK_SIZE_N": 256,
|
13 |
+
"BLOCK_SIZE_K": 128,
|
14 |
+
"GROUP_SIZE_M": 64,
|
15 |
+
"num_warps": 4,
|
16 |
+
"num_stages": 5
|
17 |
+
},
|
18 |
+
"4": {
|
19 |
+
"BLOCK_SIZE_M": 64,
|
20 |
+
"BLOCK_SIZE_N": 256,
|
21 |
+
"BLOCK_SIZE_K": 128,
|
22 |
+
"GROUP_SIZE_M": 1,
|
23 |
+
"num_warps": 4,
|
24 |
+
"num_stages": 5
|
25 |
+
},
|
26 |
+
"8": {
|
27 |
+
"BLOCK_SIZE_M": 64,
|
28 |
+
"BLOCK_SIZE_N": 256,
|
29 |
+
"BLOCK_SIZE_K": 128,
|
30 |
+
"GROUP_SIZE_M": 16,
|
31 |
+
"num_warps": 4,
|
32 |
+
"num_stages": 5
|
33 |
+
},
|
34 |
+
"16": {
|
35 |
+
"BLOCK_SIZE_M": 64,
|
36 |
+
"BLOCK_SIZE_N": 128,
|
37 |
+
"BLOCK_SIZE_K": 128,
|
38 |
+
"GROUP_SIZE_M": 64,
|
39 |
+
"num_warps": 4,
|
40 |
+
"num_stages": 3
|
41 |
+
},
|
42 |
+
"24": {
|
43 |
+
"BLOCK_SIZE_M": 64,
|
44 |
+
"BLOCK_SIZE_N": 128,
|
45 |
+
"BLOCK_SIZE_K": 128,
|
46 |
+
"GROUP_SIZE_M": 16,
|
47 |
+
"num_warps": 4,
|
48 |
+
"num_stages": 4
|
49 |
+
},
|
50 |
+
"32": {
|
51 |
+
"BLOCK_SIZE_M": 64,
|
52 |
+
"BLOCK_SIZE_N": 256,
|
53 |
+
"BLOCK_SIZE_K": 128,
|
54 |
+
"GROUP_SIZE_M": 1,
|
55 |
+
"num_warps": 4,
|
56 |
+
"num_stages": 5
|
57 |
+
},
|
58 |
+
"48": {
|
59 |
+
"BLOCK_SIZE_M": 64,
|
60 |
+
"BLOCK_SIZE_N": 256,
|
61 |
+
"BLOCK_SIZE_K": 128,
|
62 |
+
"GROUP_SIZE_M": 1,
|
63 |
+
"num_warps": 4,
|
64 |
+
"num_stages": 5
|
65 |
+
},
|
66 |
+
"64": {
|
67 |
+
"BLOCK_SIZE_M": 64,
|
68 |
+
"BLOCK_SIZE_N": 128,
|
69 |
+
"BLOCK_SIZE_K": 128,
|
70 |
+
"GROUP_SIZE_M": 1,
|
71 |
+
"num_warps": 4,
|
72 |
+
"num_stages": 3
|
73 |
+
},
|
74 |
+
"96": {
|
75 |
+
"BLOCK_SIZE_M": 64,
|
76 |
+
"BLOCK_SIZE_N": 128,
|
77 |
+
"BLOCK_SIZE_K": 128,
|
78 |
+
"GROUP_SIZE_M": 1,
|
79 |
+
"num_warps": 4,
|
80 |
+
"num_stages": 3
|
81 |
+
},
|
82 |
+
"128": {
|
83 |
+
"BLOCK_SIZE_M": 64,
|
84 |
+
"BLOCK_SIZE_N": 128,
|
85 |
+
"BLOCK_SIZE_K": 128,
|
86 |
+
"GROUP_SIZE_M": 1,
|
87 |
+
"num_warps": 4,
|
88 |
+
"num_stages": 3
|
89 |
+
},
|
90 |
+
"256": {
|
91 |
+
"BLOCK_SIZE_M": 64,
|
92 |
+
"BLOCK_SIZE_N": 128,
|
93 |
+
"BLOCK_SIZE_K": 128,
|
94 |
+
"GROUP_SIZE_M": 1,
|
95 |
+
"num_warps": 4,
|
96 |
+
"num_stages": 3
|
97 |
+
},
|
98 |
+
"512": {
|
99 |
+
"BLOCK_SIZE_M": 128,
|
100 |
+
"BLOCK_SIZE_N": 256,
|
101 |
+
"BLOCK_SIZE_K": 128,
|
102 |
+
"GROUP_SIZE_M": 1,
|
103 |
+
"num_warps": 8,
|
104 |
+
"num_stages": 4
|
105 |
+
},
|
106 |
+
"1024": {
|
107 |
+
"BLOCK_SIZE_M": 128,
|
108 |
+
"BLOCK_SIZE_N": 256,
|
109 |
+
"BLOCK_SIZE_K": 128,
|
110 |
+
"GROUP_SIZE_M": 1,
|
111 |
+
"num_warps": 8,
|
112 |
+
"num_stages": 4
|
113 |
+
},
|
114 |
+
"1536": {
|
115 |
+
"BLOCK_SIZE_M": 128,
|
116 |
+
"BLOCK_SIZE_N": 256,
|
117 |
+
"BLOCK_SIZE_K": 128,
|
118 |
+
"GROUP_SIZE_M": 1,
|
119 |
+
"num_warps": 8,
|
120 |
+
"num_stages": 4
|
121 |
+
},
|
122 |
+
"2048": {
|
123 |
+
"BLOCK_SIZE_M": 128,
|
124 |
+
"BLOCK_SIZE_N": 256,
|
125 |
+
"BLOCK_SIZE_K": 128,
|
126 |
+
"GROUP_SIZE_M": 16,
|
127 |
+
"num_warps": 8,
|
128 |
+
"num_stages": 4
|
129 |
+
},
|
130 |
+
"3072": {
|
131 |
+
"BLOCK_SIZE_M": 128,
|
132 |
+
"BLOCK_SIZE_N": 256,
|
133 |
+
"BLOCK_SIZE_K": 128,
|
134 |
+
"GROUP_SIZE_M": 1,
|
135 |
+
"num_warps": 8,
|
136 |
+
"num_stages": 4
|
137 |
+
},
|
138 |
+
"4096": {
|
139 |
+
"BLOCK_SIZE_M": 128,
|
140 |
+
"BLOCK_SIZE_N": 256,
|
141 |
+
"BLOCK_SIZE_K": 128,
|
142 |
+
"GROUP_SIZE_M": 1,
|
143 |
+
"num_warps": 8,
|
144 |
+
"num_stages": 4
|
145 |
+
}
|
146 |
+
}
|
build/torch25-cxx11-cu118-x86_64-linux/moe/configs/E=64,N=2560,device_name=NVIDIA_H200.json
ADDED
@@ -0,0 +1,146 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"1": {
|
3 |
+
"BLOCK_SIZE_M": 16,
|
4 |
+
"BLOCK_SIZE_N": 64,
|
5 |
+
"BLOCK_SIZE_K": 128,
|
6 |
+
"GROUP_SIZE_M": 32,
|
7 |
+
"num_warps": 4,
|
8 |
+
"num_stages": 2
|
9 |
+
},
|
10 |
+
"2": {
|
11 |
+
"BLOCK_SIZE_M": 16,
|
12 |
+
"BLOCK_SIZE_N": 128,
|
13 |
+
"BLOCK_SIZE_K": 128,
|
14 |
+
"GROUP_SIZE_M": 32,
|
15 |
+
"num_warps": 4,
|
16 |
+
"num_stages": 5
|
17 |
+
},
|
18 |
+
"4": {
|
19 |
+
"BLOCK_SIZE_M": 16,
|
20 |
+
"BLOCK_SIZE_N": 64,
|
21 |
+
"BLOCK_SIZE_K": 64,
|
22 |
+
"GROUP_SIZE_M": 16,
|
23 |
+
"num_warps": 4,
|
24 |
+
"num_stages": 5
|
25 |
+
},
|
26 |
+
"8": {
|
27 |
+
"BLOCK_SIZE_M": 16,
|
28 |
+
"BLOCK_SIZE_N": 64,
|
29 |
+
"BLOCK_SIZE_K": 256,
|
30 |
+
"GROUP_SIZE_M": 1,
|
31 |
+
"num_warps": 4,
|
32 |
+
"num_stages": 3
|
33 |
+
},
|
34 |
+
"16": {
|
35 |
+
"BLOCK_SIZE_M": 16,
|
36 |
+
"BLOCK_SIZE_N": 128,
|
37 |
+
"BLOCK_SIZE_K": 128,
|
38 |
+
"GROUP_SIZE_M": 16,
|
39 |
+
"num_warps": 4,
|
40 |
+
"num_stages": 5
|
41 |
+
},
|
42 |
+
"24": {
|
43 |
+
"BLOCK_SIZE_M": 16,
|
44 |
+
"BLOCK_SIZE_N": 64,
|
45 |
+
"BLOCK_SIZE_K": 256,
|
46 |
+
"GROUP_SIZE_M": 16,
|
47 |
+
"num_warps": 4,
|
48 |
+
"num_stages": 3
|
49 |
+
},
|
50 |
+
"32": {
|
51 |
+
"BLOCK_SIZE_M": 16,
|
52 |
+
"BLOCK_SIZE_N": 64,
|
53 |
+
"BLOCK_SIZE_K": 128,
|
54 |
+
"GROUP_SIZE_M": 1,
|
55 |
+
"num_warps": 4,
|
56 |
+
"num_stages": 5
|
57 |
+
},
|
58 |
+
"48": {
|
59 |
+
"BLOCK_SIZE_M": 16,
|
60 |
+
"BLOCK_SIZE_N": 128,
|
61 |
+
"BLOCK_SIZE_K": 256,
|
62 |
+
"GROUP_SIZE_M": 1,
|
63 |
+
"num_warps": 4,
|
64 |
+
"num_stages": 2
|
65 |
+
},
|
66 |
+
"64": {
|
67 |
+
"BLOCK_SIZE_M": 16,
|
68 |
+
"BLOCK_SIZE_N": 128,
|
69 |
+
"BLOCK_SIZE_K": 128,
|
70 |
+
"GROUP_SIZE_M": 1,
|
71 |
+
"num_warps": 8,
|
72 |
+
"num_stages": 3
|
73 |
+
},
|
74 |
+
"96": {
|
75 |
+
"BLOCK_SIZE_M": 32,
|
76 |
+
"BLOCK_SIZE_N": 128,
|
77 |
+
"BLOCK_SIZE_K": 128,
|
78 |
+
"GROUP_SIZE_M": 1,
|
79 |
+
"num_warps": 4,
|
80 |
+
"num_stages": 3
|
81 |
+
},
|
82 |
+
"128": {
|
83 |
+
"BLOCK_SIZE_M": 32,
|
84 |
+
"BLOCK_SIZE_N": 128,
|
85 |
+
"BLOCK_SIZE_K": 128,
|
86 |
+
"GROUP_SIZE_M": 1,
|
87 |
+
"num_warps": 4,
|
88 |
+
"num_stages": 3
|
89 |
+
},
|
90 |
+
"256": {
|
91 |
+
"BLOCK_SIZE_M": 64,
|
92 |
+
"BLOCK_SIZE_N": 64,
|
93 |
+
"BLOCK_SIZE_K": 64,
|
94 |
+
"GROUP_SIZE_M": 1,
|
95 |
+
"num_warps": 4,
|
96 |
+
"num_stages": 3
|
97 |
+
},
|
98 |
+
"512": {
|
99 |
+
"BLOCK_SIZE_M": 128,
|
100 |
+
"BLOCK_SIZE_N": 256,
|
101 |
+
"BLOCK_SIZE_K": 64,
|
102 |
+
"GROUP_SIZE_M": 1,
|
103 |
+
"num_warps": 8,
|
104 |
+
"num_stages": 4
|
105 |
+
},
|
106 |
+
"1024": {
|
107 |
+
"BLOCK_SIZE_M": 128,
|
108 |
+
"BLOCK_SIZE_N": 256,
|
109 |
+
"BLOCK_SIZE_K": 64,
|
110 |
+
"GROUP_SIZE_M": 16,
|
111 |
+
"num_warps": 8,
|
112 |
+
"num_stages": 4
|
113 |
+
},
|
114 |
+
"1536": {
|
115 |
+
"BLOCK_SIZE_M": 128,
|
116 |
+
"BLOCK_SIZE_N": 256,
|
117 |
+
"BLOCK_SIZE_K": 64,
|
118 |
+
"GROUP_SIZE_M": 16,
|
119 |
+
"num_warps": 8,
|
120 |
+
"num_stages": 4
|
121 |
+
},
|
122 |
+
"2048": {
|
123 |
+
"BLOCK_SIZE_M": 128,
|
124 |
+
"BLOCK_SIZE_N": 256,
|
125 |
+
"BLOCK_SIZE_K": 64,
|
126 |
+
"GROUP_SIZE_M": 1,
|
127 |
+
"num_warps": 8,
|
128 |
+
"num_stages": 4
|
129 |
+
},
|
130 |
+
"3072": {
|
131 |
+
"BLOCK_SIZE_M": 128,
|
132 |
+
"BLOCK_SIZE_N": 256,
|
133 |
+
"BLOCK_SIZE_K": 64,
|
134 |
+
"GROUP_SIZE_M": 1,
|
135 |
+
"num_warps": 8,
|
136 |
+
"num_stages": 4
|
137 |
+
},
|
138 |
+
"4096": {
|
139 |
+
"BLOCK_SIZE_M": 128,
|
140 |
+
"BLOCK_SIZE_N": 256,
|
141 |
+
"BLOCK_SIZE_K": 64,
|
142 |
+
"GROUP_SIZE_M": 16,
|
143 |
+
"num_warps": 8,
|
144 |
+
"num_stages": 4
|
145 |
+
}
|
146 |
+
}
|
build/torch25-cxx11-cu118-x86_64-linux/moe/configs/E=64,N=320,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8.json
ADDED
@@ -0,0 +1,146 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"1": {
|
3 |
+
"BLOCK_SIZE_M": 64,
|
4 |
+
"BLOCK_SIZE_N": 128,
|
5 |
+
"BLOCK_SIZE_K": 128,
|
6 |
+
"GROUP_SIZE_M": 16,
|
7 |
+
"num_warps": 4,
|
8 |
+
"num_stages": 4
|
9 |
+
},
|
10 |
+
"2": {
|
11 |
+
"BLOCK_SIZE_M": 64,
|
12 |
+
"BLOCK_SIZE_N": 128,
|
13 |
+
"BLOCK_SIZE_K": 128,
|
14 |
+
"GROUP_SIZE_M": 16,
|
15 |
+
"num_warps": 4,
|
16 |
+
"num_stages": 4
|
17 |
+
},
|
18 |
+
"4": {
|
19 |
+
"BLOCK_SIZE_M": 64,
|
20 |
+
"BLOCK_SIZE_N": 128,
|
21 |
+
"BLOCK_SIZE_K": 128,
|
22 |
+
"GROUP_SIZE_M": 16,
|
23 |
+
"num_warps": 4,
|
24 |
+
"num_stages": 4
|
25 |
+
},
|
26 |
+
"8": {
|
27 |
+
"BLOCK_SIZE_M": 64,
|
28 |
+
"BLOCK_SIZE_N": 128,
|
29 |
+
"BLOCK_SIZE_K": 128,
|
30 |
+
"GROUP_SIZE_M": 1,
|
31 |
+
"num_warps": 8,
|
32 |
+
"num_stages": 3
|
33 |
+
},
|
34 |
+
"16": {
|
35 |
+
"BLOCK_SIZE_M": 64,
|
36 |
+
"BLOCK_SIZE_N": 128,
|
37 |
+
"BLOCK_SIZE_K": 128,
|
38 |
+
"GROUP_SIZE_M": 64,
|
39 |
+
"num_warps": 4,
|
40 |
+
"num_stages": 3
|
41 |
+
},
|
42 |
+
"24": {
|
43 |
+
"BLOCK_SIZE_M": 64,
|
44 |
+
"BLOCK_SIZE_N": 128,
|
45 |
+
"BLOCK_SIZE_K": 128,
|
46 |
+
"GROUP_SIZE_M": 64,
|
47 |
+
"num_warps": 4,
|
48 |
+
"num_stages": 3
|
49 |
+
},
|
50 |
+
"32": {
|
51 |
+
"BLOCK_SIZE_M": 64,
|
52 |
+
"BLOCK_SIZE_N": 128,
|
53 |
+
"BLOCK_SIZE_K": 128,
|
54 |
+
"GROUP_SIZE_M": 1,
|
55 |
+
"num_warps": 4,
|
56 |
+
"num_stages": 3
|
57 |
+
},
|
58 |
+
"48": {
|
59 |
+
"BLOCK_SIZE_M": 64,
|
60 |
+
"BLOCK_SIZE_N": 128,
|
61 |
+
"BLOCK_SIZE_K": 128,
|
62 |
+
"GROUP_SIZE_M": 1,
|
63 |
+
"num_warps": 4,
|
64 |
+
"num_stages": 3
|
65 |
+
},
|
66 |
+
"64": {
|
67 |
+
"BLOCK_SIZE_M": 64,
|
68 |
+
"BLOCK_SIZE_N": 128,
|
69 |
+
"BLOCK_SIZE_K": 128,
|
70 |
+
"GROUP_SIZE_M": 1,
|
71 |
+
"num_warps": 4,
|
72 |
+
"num_stages": 3
|
73 |
+
},
|
74 |
+
"96": {
|
75 |
+
"BLOCK_SIZE_M": 64,
|
76 |
+
"BLOCK_SIZE_N": 128,
|
77 |
+
"BLOCK_SIZE_K": 128,
|
78 |
+
"GROUP_SIZE_M": 16,
|
79 |
+
"num_warps": 4,
|
80 |
+
"num_stages": 3
|
81 |
+
},
|
82 |
+
"128": {
|
83 |
+
"BLOCK_SIZE_M": 64,
|
84 |
+
"BLOCK_SIZE_N": 128,
|
85 |
+
"BLOCK_SIZE_K": 128,
|
86 |
+
"GROUP_SIZE_M": 1,
|
87 |
+
"num_warps": 4,
|
88 |
+
"num_stages": 3
|
89 |
+
},
|
90 |
+
"256": {
|
91 |
+
"BLOCK_SIZE_M": 64,
|
92 |
+
"BLOCK_SIZE_N": 128,
|
93 |
+
"BLOCK_SIZE_K": 128,
|
94 |
+
"GROUP_SIZE_M": 32,
|
95 |
+
"num_warps": 4,
|
96 |
+
"num_stages": 3
|
97 |
+
},
|
98 |
+
"512": {
|
99 |
+
"BLOCK_SIZE_M": 64,
|
100 |
+
"BLOCK_SIZE_N": 128,
|
101 |
+
"BLOCK_SIZE_K": 128,
|
102 |
+
"GROUP_SIZE_M": 64,
|
103 |
+
"num_warps": 4,
|
104 |
+
"num_stages": 2
|
105 |
+
},
|
106 |
+
"1024": {
|
107 |
+
"BLOCK_SIZE_M": 128,
|
108 |
+
"BLOCK_SIZE_N": 128,
|
109 |
+
"BLOCK_SIZE_K": 128,
|
110 |
+
"GROUP_SIZE_M": 1,
|
111 |
+
"num_warps": 8,
|
112 |
+
"num_stages": 3
|
113 |
+
},
|
114 |
+
"1536": {
|
115 |
+
"BLOCK_SIZE_M": 128,
|
116 |
+
"BLOCK_SIZE_N": 256,
|
117 |
+
"BLOCK_SIZE_K": 128,
|
118 |
+
"GROUP_SIZE_M": 1,
|
119 |
+
"num_warps": 8,
|
120 |
+
"num_stages": 4
|
121 |
+
},
|
122 |
+
"2048": {
|
123 |
+
"BLOCK_SIZE_M": 128,
|
124 |
+
"BLOCK_SIZE_N": 128,
|
125 |
+
"BLOCK_SIZE_K": 128,
|
126 |
+
"GROUP_SIZE_M": 1,
|
127 |
+
"num_warps": 8,
|
128 |
+
"num_stages": 3
|
129 |
+
},
|
130 |
+
"3072": {
|
131 |
+
"BLOCK_SIZE_M": 128,
|
132 |
+
"BLOCK_SIZE_N": 128,
|
133 |
+
"BLOCK_SIZE_K": 128,
|
134 |
+
"GROUP_SIZE_M": 1,
|
135 |
+
"num_warps": 8,
|
136 |
+
"num_stages": 3
|
137 |
+
},
|
138 |
+
"4096": {
|
139 |
+
"BLOCK_SIZE_M": 128,
|
140 |
+
"BLOCK_SIZE_N": 128,
|
141 |
+
"BLOCK_SIZE_K": 128,
|
142 |
+
"GROUP_SIZE_M": 1,
|
143 |
+
"num_warps": 8,
|
144 |
+
"num_stages": 3
|
145 |
+
}
|
146 |
+
}
|
build/torch25-cxx11-cu118-x86_64-linux/moe/configs/E=64,N=320,device_name=NVIDIA_H100_80GB_HBM3.json
ADDED
@@ -0,0 +1,146 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"1": {
|
3 |
+
"BLOCK_SIZE_M": 16,
|
4 |
+
"BLOCK_SIZE_N": 64,
|
5 |
+
"BLOCK_SIZE_K": 128,
|
6 |
+
"GROUP_SIZE_M": 1,
|
7 |
+
"num_warps": 4,
|
8 |
+
"num_stages": 4
|
9 |
+
},
|
10 |
+
"2": {
|
11 |
+
"BLOCK_SIZE_M": 16,
|
12 |
+
"BLOCK_SIZE_N": 64,
|
13 |
+
"BLOCK_SIZE_K": 128,
|
14 |
+
"GROUP_SIZE_M": 16,
|
15 |
+
"num_warps": 4,
|
16 |
+
"num_stages": 4
|
17 |
+
},
|
18 |
+
"4": {
|
19 |
+
"BLOCK_SIZE_M": 16,
|
20 |
+
"BLOCK_SIZE_N": 64,
|
21 |
+
"BLOCK_SIZE_K": 64,
|
22 |
+
"GROUP_SIZE_M": 1,
|
23 |
+
"num_warps": 4,
|
24 |
+
"num_stages": 4
|
25 |
+
},
|
26 |
+
"8": {
|
27 |
+
"BLOCK_SIZE_M": 16,
|
28 |
+
"BLOCK_SIZE_N": 64,
|
29 |
+
"BLOCK_SIZE_K": 64,
|
30 |
+
"GROUP_SIZE_M": 32,
|
31 |
+
"num_warps": 4,
|
32 |
+
"num_stages": 3
|
33 |
+
},
|
34 |
+
"16": {
|
35 |
+
"BLOCK_SIZE_M": 32,
|
36 |
+
"BLOCK_SIZE_N": 64,
|
37 |
+
"BLOCK_SIZE_K": 64,
|
38 |
+
"GROUP_SIZE_M": 64,
|
39 |
+
"num_warps": 4,
|
40 |
+
"num_stages": 3
|
41 |
+
},
|
42 |
+
"24": {
|
43 |
+
"BLOCK_SIZE_M": 16,
|
44 |
+
"BLOCK_SIZE_N": 64,
|
45 |
+
"BLOCK_SIZE_K": 128,
|
46 |
+
"GROUP_SIZE_M": 1,
|
47 |
+
"num_warps": 4,
|
48 |
+
"num_stages": 2
|
49 |
+
},
|
50 |
+
"32": {
|
51 |
+
"BLOCK_SIZE_M": 16,
|
52 |
+
"BLOCK_SIZE_N": 64,
|
53 |
+
"BLOCK_SIZE_K": 64,
|
54 |
+
"GROUP_SIZE_M": 1,
|
55 |
+
"num_warps": 4,
|
56 |
+
"num_stages": 3
|
57 |
+
},
|
58 |
+
"48": {
|
59 |
+
"BLOCK_SIZE_M": 16,
|
60 |
+
"BLOCK_SIZE_N": 64,
|
61 |
+
"BLOCK_SIZE_K": 64,
|
62 |
+
"GROUP_SIZE_M": 1,
|
63 |
+
"num_warps": 4,
|
64 |
+
"num_stages": 3
|
65 |
+
},
|
66 |
+
"64": {
|
67 |
+
"BLOCK_SIZE_M": 16,
|
68 |
+
"BLOCK_SIZE_N": 64,
|
69 |
+
"BLOCK_SIZE_K": 64,
|
70 |
+
"GROUP_SIZE_M": 1,
|
71 |
+
"num_warps": 4,
|
72 |
+
"num_stages": 3
|
73 |
+
},
|
74 |
+
"96": {
|
75 |
+
"BLOCK_SIZE_M": 16,
|
76 |
+
"BLOCK_SIZE_N": 64,
|
77 |
+
"BLOCK_SIZE_K": 128,
|
78 |
+
"GROUP_SIZE_M": 1,
|
79 |
+
"num_warps": 4,
|
80 |
+
"num_stages": 2
|
81 |
+
},
|
82 |
+
"128": {
|
83 |
+
"BLOCK_SIZE_M": 32,
|
84 |
+
"BLOCK_SIZE_N": 64,
|
85 |
+
"BLOCK_SIZE_K": 64,
|
86 |
+
"GROUP_SIZE_M": 1,
|
87 |
+
"num_warps": 4,
|
88 |
+
"num_stages": 3
|
89 |
+
},
|
90 |
+
"256": {
|
91 |
+
"BLOCK_SIZE_M": 64,
|
92 |
+
"BLOCK_SIZE_N": 128,
|
93 |
+
"BLOCK_SIZE_K": 64,
|
94 |
+
"GROUP_SIZE_M": 16,
|
95 |
+
"num_warps": 4,
|
96 |
+
"num_stages": 3
|
97 |
+
},
|
98 |
+
"512": {
|
99 |
+
"BLOCK_SIZE_M": 64,
|
100 |
+
"BLOCK_SIZE_N": 128,
|
101 |
+
"BLOCK_SIZE_K": 64,
|
102 |
+
"GROUP_SIZE_M": 32,
|
103 |
+
"num_warps": 4,
|
104 |
+
"num_stages": 4
|
105 |
+
},
|
106 |
+
"1024": {
|
107 |
+
"BLOCK_SIZE_M": 128,
|
108 |
+
"BLOCK_SIZE_N": 128,
|
109 |
+
"BLOCK_SIZE_K": 64,
|
110 |
+
"GROUP_SIZE_M": 1,
|
111 |
+
"num_warps": 8,
|
112 |
+
"num_stages": 3
|
113 |
+
},
|
114 |
+
"1536": {
|
115 |
+
"BLOCK_SIZE_M": 128,
|
116 |
+
"BLOCK_SIZE_N": 256,
|
117 |
+
"BLOCK_SIZE_K": 64,
|
118 |
+
"GROUP_SIZE_M": 1,
|
119 |
+
"num_warps": 8,
|
120 |
+
"num_stages": 4
|
121 |
+
},
|
122 |
+
"2048": {
|
123 |
+
"BLOCK_SIZE_M": 128,
|
124 |
+
"BLOCK_SIZE_N": 256,
|
125 |
+
"BLOCK_SIZE_K": 64,
|
126 |
+
"GROUP_SIZE_M": 1,
|
127 |
+
"num_warps": 8,
|
128 |
+
"num_stages": 4
|
129 |
+
},
|
130 |
+
"3072": {
|
131 |
+
"BLOCK_SIZE_M": 128,
|
132 |
+
"BLOCK_SIZE_N": 256,
|
133 |
+
"BLOCK_SIZE_K": 64,
|
134 |
+
"GROUP_SIZE_M": 16,
|
135 |
+
"num_warps": 8,
|
136 |
+
"num_stages": 4
|
137 |
+
},
|
138 |
+
"4096": {
|
139 |
+
"BLOCK_SIZE_M": 128,
|
140 |
+
"BLOCK_SIZE_N": 256,
|
141 |
+
"BLOCK_SIZE_K": 64,
|
142 |
+
"GROUP_SIZE_M": 16,
|
143 |
+
"num_warps": 8,
|
144 |
+
"num_stages": 4
|
145 |
+
}
|
146 |
+
}
|
build/torch25-cxx11-cu118-x86_64-linux/moe/configs/E=64,N=320,device_name=NVIDIA_H200,dtype=fp8_w8a8.json
ADDED
@@ -0,0 +1,146 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"1": {
|
3 |
+
"BLOCK_SIZE_M": 64,
|
4 |
+
"BLOCK_SIZE_N": 128,
|
5 |
+
"BLOCK_SIZE_K": 128,
|
6 |
+
"GROUP_SIZE_M": 16,
|
7 |
+
"num_warps": 4,
|
8 |
+
"num_stages": 4
|
9 |
+
},
|
10 |
+
"2": {
|
11 |
+
"BLOCK_SIZE_M": 64,
|
12 |
+
"BLOCK_SIZE_N": 128,
|
13 |
+
"BLOCK_SIZE_K": 128,
|
14 |
+
"GROUP_SIZE_M": 16,
|
15 |
+
"num_warps": 4,
|
16 |
+
"num_stages": 4
|
17 |
+
},
|
18 |
+
"4": {
|
19 |
+
"BLOCK_SIZE_M": 64,
|
20 |
+
"BLOCK_SIZE_N": 128,
|
21 |
+
"BLOCK_SIZE_K": 128,
|
22 |
+
"GROUP_SIZE_M": 32,
|
23 |
+
"num_warps": 4,
|
24 |
+
"num_stages": 4
|
25 |
+
},
|
26 |
+
"8": {
|
27 |
+
"BLOCK_SIZE_M": 64,
|
28 |
+
"BLOCK_SIZE_N": 128,
|
29 |
+
"BLOCK_SIZE_K": 128,
|
30 |
+
"GROUP_SIZE_M": 64,
|
31 |
+
"num_warps": 4,
|
32 |
+
"num_stages": 3
|
33 |
+
},
|
34 |
+
"16": {
|
35 |
+
"BLOCK_SIZE_M": 64,
|
36 |
+
"BLOCK_SIZE_N": 128,
|
37 |
+
"BLOCK_SIZE_K": 128,
|
38 |
+
"GROUP_SIZE_M": 64,
|
39 |
+
"num_warps": 4,
|
40 |
+
"num_stages": 3
|
41 |
+
},
|
42 |
+
"24": {
|
43 |
+
"BLOCK_SIZE_M": 64,
|
44 |
+
"BLOCK_SIZE_N": 128,
|
45 |
+
"BLOCK_SIZE_K": 128,
|
46 |
+
"GROUP_SIZE_M": 1,
|
47 |
+
"num_warps": 4,
|
48 |
+
"num_stages": 3
|
49 |
+
},
|
50 |
+
"32": {
|
51 |
+
"BLOCK_SIZE_M": 64,
|
52 |
+
"BLOCK_SIZE_N": 128,
|
53 |
+
"BLOCK_SIZE_K": 128,
|
54 |
+
"GROUP_SIZE_M": 1,
|
55 |
+
"num_warps": 4,
|
56 |
+
"num_stages": 3
|
57 |
+
},
|
58 |
+
"48": {
|
59 |
+
"BLOCK_SIZE_M": 64,
|
60 |
+
"BLOCK_SIZE_N": 128,
|
61 |
+
"BLOCK_SIZE_K": 128,
|
62 |
+
"GROUP_SIZE_M": 1,
|
63 |
+
"num_warps": 4,
|
64 |
+
"num_stages": 3
|
65 |
+
},
|
66 |
+
"64": {
|
67 |
+
"BLOCK_SIZE_M": 64,
|
68 |
+
"BLOCK_SIZE_N": 128,
|
69 |
+
"BLOCK_SIZE_K": 128,
|
70 |
+
"GROUP_SIZE_M": 1,
|
71 |
+
"num_warps": 4,
|
72 |
+
"num_stages": 3
|
73 |
+
},
|
74 |
+
"96": {
|
75 |
+
"BLOCK_SIZE_M": 64,
|
76 |
+
"BLOCK_SIZE_N": 128,
|
77 |
+
"BLOCK_SIZE_K": 128,
|
78 |
+
"GROUP_SIZE_M": 16,
|
79 |
+
"num_warps": 4,
|
80 |
+
"num_stages": 3
|
81 |
+
},
|
82 |
+
"128": {
|
83 |
+
"BLOCK_SIZE_M": 64,
|
84 |
+
"BLOCK_SIZE_N": 128,
|
85 |
+
"BLOCK_SIZE_K": 128,
|
86 |
+
"GROUP_SIZE_M": 1,
|
87 |
+
"num_warps": 4,
|
88 |
+
"num_stages": 3
|
89 |
+
},
|
90 |
+
"256": {
|
91 |
+
"BLOCK_SIZE_M": 64,
|
92 |
+
"BLOCK_SIZE_N": 128,
|
93 |
+
"BLOCK_SIZE_K": 128,
|
94 |
+
"GROUP_SIZE_M": 32,
|
95 |
+
"num_warps": 4,
|
96 |
+
"num_stages": 3
|
97 |
+
},
|
98 |
+
"512": {
|
99 |
+
"BLOCK_SIZE_M": 64,
|
100 |
+
"BLOCK_SIZE_N": 128,
|
101 |
+
"BLOCK_SIZE_K": 128,
|
102 |
+
"GROUP_SIZE_M": 16,
|
103 |
+
"num_warps": 4,
|
104 |
+
"num_stages": 2
|
105 |
+
},
|
106 |
+
"1024": {
|
107 |
+
"BLOCK_SIZE_M": 128,
|
108 |
+
"BLOCK_SIZE_N": 128,
|
109 |
+
"BLOCK_SIZE_K": 128,
|
110 |
+
"GROUP_SIZE_M": 1,
|
111 |
+
"num_warps": 8,
|
112 |
+
"num_stages": 3
|
113 |
+
},
|
114 |
+
"1536": {
|
115 |
+
"BLOCK_SIZE_M": 128,
|
116 |
+
"BLOCK_SIZE_N": 256,
|
117 |
+
"BLOCK_SIZE_K": 128,
|
118 |
+
"GROUP_SIZE_M": 1,
|
119 |
+
"num_warps": 8,
|
120 |
+
"num_stages": 4
|
121 |
+
},
|
122 |
+
"2048": {
|
123 |
+
"BLOCK_SIZE_M": 128,
|
124 |
+
"BLOCK_SIZE_N": 128,
|
125 |
+
"BLOCK_SIZE_K": 128,
|
126 |
+
"GROUP_SIZE_M": 1,
|
127 |
+
"num_warps": 8,
|
128 |
+
"num_stages": 3
|
129 |
+
},
|
130 |
+
"3072": {
|
131 |
+
"BLOCK_SIZE_M": 128,
|
132 |
+
"BLOCK_SIZE_N": 128,
|
133 |
+
"BLOCK_SIZE_K": 128,
|
134 |
+
"GROUP_SIZE_M": 1,
|
135 |
+
"num_warps": 8,
|
136 |
+
"num_stages": 3
|
137 |
+
},
|
138 |
+
"4096": {
|
139 |
+
"BLOCK_SIZE_M": 128,
|
140 |
+
"BLOCK_SIZE_N": 128,
|
141 |
+
"BLOCK_SIZE_K": 128,
|
142 |
+
"GROUP_SIZE_M": 1,
|
143 |
+
"num_warps": 8,
|
144 |
+
"num_stages": 3
|
145 |
+
}
|
146 |
+
}
|
build/torch25-cxx11-cu118-x86_64-linux/moe/configs/E=64,N=320,device_name=NVIDIA_H200.json
ADDED
@@ -0,0 +1,146 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"1": {
|
3 |
+
"BLOCK_SIZE_M": 16,
|
4 |
+
"BLOCK_SIZE_N": 64,
|
5 |
+
"BLOCK_SIZE_K": 128,
|
6 |
+
"GROUP_SIZE_M": 1,
|
7 |
+
"num_warps": 4,
|
8 |
+
"num_stages": 4
|
9 |
+
},
|
10 |
+
"2": {
|
11 |
+
"BLOCK_SIZE_M": 16,
|
12 |
+
"BLOCK_SIZE_N": 64,
|
13 |
+
"BLOCK_SIZE_K": 128,
|
14 |
+
"GROUP_SIZE_M": 1,
|
15 |
+
"num_warps": 4,
|
16 |
+
"num_stages": 4
|
17 |
+
},
|
18 |
+
"4": {
|
19 |
+
"BLOCK_SIZE_M": 16,
|
20 |
+
"BLOCK_SIZE_N": 64,
|
21 |
+
"BLOCK_SIZE_K": 64,
|
22 |
+
"GROUP_SIZE_M": 32,
|
23 |
+
"num_warps": 4,
|
24 |
+
"num_stages": 4
|
25 |
+
},
|
26 |
+
"8": {
|
27 |
+
"BLOCK_SIZE_M": 16,
|
28 |
+
"BLOCK_SIZE_N": 64,
|
29 |
+
"BLOCK_SIZE_K": 64,
|
30 |
+
"GROUP_SIZE_M": 16,
|
31 |
+
"num_warps": 4,
|
32 |
+
"num_stages": 3
|
33 |
+
},
|
34 |
+
"16": {
|
35 |
+
"BLOCK_SIZE_M": 16,
|
36 |
+
"BLOCK_SIZE_N": 64,
|
37 |
+
"BLOCK_SIZE_K": 64,
|
38 |
+
"GROUP_SIZE_M": 1,
|
39 |
+
"num_warps": 4,
|
40 |
+
"num_stages": 3
|
41 |
+
},
|
42 |
+
"24": {
|
43 |
+
"BLOCK_SIZE_M": 16,
|
44 |
+
"BLOCK_SIZE_N": 64,
|
45 |
+
"BLOCK_SIZE_K": 64,
|
46 |
+
"GROUP_SIZE_M": 16,
|
47 |
+
"num_warps": 4,
|
48 |
+
"num_stages": 3
|
49 |
+
},
|
50 |
+
"32": {
|
51 |
+
"BLOCK_SIZE_M": 16,
|
52 |
+
"BLOCK_SIZE_N": 64,
|
53 |
+
"BLOCK_SIZE_K": 64,
|
54 |
+
"GROUP_SIZE_M": 16,
|
55 |
+
"num_warps": 4,
|
56 |
+
"num_stages": 3
|
57 |
+
},
|
58 |
+
"48": {
|
59 |
+
"BLOCK_SIZE_M": 16,
|
60 |
+
"BLOCK_SIZE_N": 64,
|
61 |
+
"BLOCK_SIZE_K": 64,
|
62 |
+
"GROUP_SIZE_M": 32,
|
63 |
+
"num_warps": 4,
|
64 |
+
"num_stages": 3
|
65 |
+
},
|
66 |
+
"64": {
|
67 |
+
"BLOCK_SIZE_M": 16,
|
68 |
+
"BLOCK_SIZE_N": 64,
|
69 |
+
"BLOCK_SIZE_K": 64,
|
70 |
+
"GROUP_SIZE_M": 16,
|
71 |
+
"num_warps": 4,
|
72 |
+
"num_stages": 3
|
73 |
+
},
|
74 |
+
"96": {
|
75 |
+
"BLOCK_SIZE_M": 64,
|
76 |
+
"BLOCK_SIZE_N": 128,
|
77 |
+
"BLOCK_SIZE_K": 64,
|
78 |
+
"GROUP_SIZE_M": 32,
|
79 |
+
"num_warps": 4,
|
80 |
+
"num_stages": 3
|
81 |
+
},
|
82 |
+
"128": {
|
83 |
+
"BLOCK_SIZE_M": 64,
|
84 |
+
"BLOCK_SIZE_N": 128,
|
85 |
+
"BLOCK_SIZE_K": 64,
|
86 |
+
"GROUP_SIZE_M": 64,
|
87 |
+
"num_warps": 4,
|
88 |
+
"num_stages": 3
|
89 |
+
},
|
90 |
+
"256": {
|
91 |
+
"BLOCK_SIZE_M": 64,
|
92 |
+
"BLOCK_SIZE_N": 128,
|
93 |
+
"BLOCK_SIZE_K": 64,
|
94 |
+
"GROUP_SIZE_M": 64,
|
95 |
+
"num_warps": 4,
|
96 |
+
"num_stages": 3
|
97 |
+
},
|
98 |
+
"512": {
|
99 |
+
"BLOCK_SIZE_M": 64,
|
100 |
+
"BLOCK_SIZE_N": 128,
|
101 |
+
"BLOCK_SIZE_K": 64,
|
102 |
+
"GROUP_SIZE_M": 16,
|
103 |
+
"num_warps": 4,
|
104 |
+
"num_stages": 2
|
105 |
+
},
|
106 |
+
"1024": {
|
107 |
+
"BLOCK_SIZE_M": 128,
|
108 |
+
"BLOCK_SIZE_N": 128,
|
109 |
+
"BLOCK_SIZE_K": 64,
|
110 |
+
"GROUP_SIZE_M": 1,
|
111 |
+
"num_warps": 8,
|
112 |
+
"num_stages": 3
|
113 |
+
},
|
114 |
+
"1536": {
|
115 |
+
"BLOCK_SIZE_M": 128,
|
116 |
+
"BLOCK_SIZE_N": 256,
|
117 |
+
"BLOCK_SIZE_K": 64,
|
118 |
+
"GROUP_SIZE_M": 16,
|
119 |
+
"num_warps": 8,
|
120 |
+
"num_stages": 4
|
121 |
+
},
|
122 |
+
"2048": {
|
123 |
+
"BLOCK_SIZE_M": 128,
|
124 |
+
"BLOCK_SIZE_N": 128,
|
125 |
+
"BLOCK_SIZE_K": 64,
|
126 |
+
"GROUP_SIZE_M": 1,
|
127 |
+
"num_warps": 8,
|
128 |
+
"num_stages": 3
|
129 |
+
},
|
130 |
+
"3072": {
|
131 |
+
"BLOCK_SIZE_M": 128,
|
132 |
+
"BLOCK_SIZE_N": 128,
|
133 |
+
"BLOCK_SIZE_K": 64,
|
134 |
+
"GROUP_SIZE_M": 1,
|
135 |
+
"num_warps": 8,
|
136 |
+
"num_stages": 3
|
137 |
+
},
|
138 |
+
"4096": {
|
139 |
+
"BLOCK_SIZE_M": 128,
|
140 |
+
"BLOCK_SIZE_N": 128,
|
141 |
+
"BLOCK_SIZE_K": 64,
|
142 |
+
"GROUP_SIZE_M": 16,
|
143 |
+
"num_warps": 8,
|
144 |
+
"num_stages": 3
|
145 |
+
}
|
146 |
+
}
|
build/torch25-cxx11-cu118-x86_64-linux/moe/configs/E=64,N=640,device_name=NVIDIA_A800-SXM4-80GB.json
ADDED
@@ -0,0 +1,146 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"1": {
|
3 |
+
"BLOCK_SIZE_M": 16,
|
4 |
+
"BLOCK_SIZE_N": 64,
|
5 |
+
"BLOCK_SIZE_K": 64,
|
6 |
+
"GROUP_SIZE_M": 16,
|
7 |
+
"num_warps": 4,
|
8 |
+
"num_stages": 4
|
9 |
+
},
|
10 |
+
"2": {
|
11 |
+
"BLOCK_SIZE_M": 32,
|
12 |
+
"BLOCK_SIZE_N": 128,
|
13 |
+
"BLOCK_SIZE_K": 64,
|
14 |
+
"GROUP_SIZE_M": 1,
|
15 |
+
"num_warps": 4,
|
16 |
+
"num_stages": 5
|
17 |
+
},
|
18 |
+
"4": {
|
19 |
+
"BLOCK_SIZE_M": 16,
|
20 |
+
"BLOCK_SIZE_N": 32,
|
21 |
+
"BLOCK_SIZE_K": 64,
|
22 |
+
"GROUP_SIZE_M": 16,
|
23 |
+
"num_warps": 4,
|
24 |
+
"num_stages": 4
|
25 |
+
},
|
26 |
+
"8": {
|
27 |
+
"BLOCK_SIZE_M": 16,
|
28 |
+
"BLOCK_SIZE_N": 128,
|
29 |
+
"BLOCK_SIZE_K": 64,
|
30 |
+
"GROUP_SIZE_M": 1,
|
31 |
+
"num_warps": 8,
|
32 |
+
"num_stages": 5
|
33 |
+
},
|
34 |
+
"16": {
|
35 |
+
"BLOCK_SIZE_M": 16,
|
36 |
+
"BLOCK_SIZE_N": 64,
|
37 |
+
"BLOCK_SIZE_K": 128,
|
38 |
+
"GROUP_SIZE_M": 16,
|
39 |
+
"num_warps": 4,
|
40 |
+
"num_stages": 2
|
41 |
+
},
|
42 |
+
"24": {
|
43 |
+
"BLOCK_SIZE_M": 16,
|
44 |
+
"BLOCK_SIZE_N": 64,
|
45 |
+
"BLOCK_SIZE_K": 128,
|
46 |
+
"GROUP_SIZE_M": 1,
|
47 |
+
"num_warps": 4,
|
48 |
+
"num_stages": 5
|
49 |
+
},
|
50 |
+
"32": {
|
51 |
+
"BLOCK_SIZE_M": 32,
|
52 |
+
"BLOCK_SIZE_N": 128,
|
53 |
+
"BLOCK_SIZE_K": 128,
|
54 |
+
"GROUP_SIZE_M": 1,
|
55 |
+
"num_warps": 4,
|
56 |
+
"num_stages": 3
|
57 |
+
},
|
58 |
+
"48": {
|
59 |
+
"BLOCK_SIZE_M": 32,
|
60 |
+
"BLOCK_SIZE_N": 128,
|
61 |
+
"BLOCK_SIZE_K": 128,
|
62 |
+
"GROUP_SIZE_M": 1,
|
63 |
+
"num_warps": 4,
|
64 |
+
"num_stages": 3
|
65 |
+
},
|
66 |
+
"64": {
|
67 |
+
"BLOCK_SIZE_M": 16,
|
68 |
+
"BLOCK_SIZE_N": 128,
|
69 |
+
"BLOCK_SIZE_K": 128,
|
70 |
+
"GROUP_SIZE_M": 1,
|
71 |
+
"num_warps": 4,
|
72 |
+
"num_stages": 3
|
73 |
+
},
|
74 |
+
"96": {
|
75 |
+
"BLOCK_SIZE_M": 32,
|
76 |
+
"BLOCK_SIZE_N": 128,
|
77 |
+
"BLOCK_SIZE_K": 128,
|
78 |
+
"GROUP_SIZE_M": 1,
|
79 |
+
"num_warps": 4,
|
80 |
+
"num_stages": 3
|
81 |
+
},
|
82 |
+
"128": {
|
83 |
+
"BLOCK_SIZE_M": 32,
|
84 |
+
"BLOCK_SIZE_N": 128,
|
85 |
+
"BLOCK_SIZE_K": 128,
|
86 |
+
"GROUP_SIZE_M": 1,
|
87 |
+
"num_warps": 4,
|
88 |
+
"num_stages": 3
|
89 |
+
},
|
90 |
+
"256": {
|
91 |
+
"BLOCK_SIZE_M": 64,
|
92 |
+
"BLOCK_SIZE_N": 256,
|
93 |
+
"BLOCK_SIZE_K": 128,
|
94 |
+
"GROUP_SIZE_M": 1,
|
95 |
+
"num_warps": 8,
|
96 |
+
"num_stages": 3
|
97 |
+
},
|
98 |
+
"512": {
|
99 |
+
"BLOCK_SIZE_M": 64,
|
100 |
+
"BLOCK_SIZE_N": 256,
|
101 |
+
"BLOCK_SIZE_K": 128,
|
102 |
+
"GROUP_SIZE_M": 1,
|
103 |
+
"num_warps": 8,
|
104 |
+
"num_stages": 3
|
105 |
+
},
|
106 |
+
"1024": {
|
107 |
+
"BLOCK_SIZE_M": 64,
|
108 |
+
"BLOCK_SIZE_N": 128,
|
109 |
+
"BLOCK_SIZE_K": 64,
|
110 |
+
"GROUP_SIZE_M": 1,
|
111 |
+
"num_warps": 4,
|
112 |
+
"num_stages": 4
|
113 |
+
},
|
114 |
+
"1536": {
|
115 |
+
"BLOCK_SIZE_M": 128,
|
116 |
+
"BLOCK_SIZE_N": 256,
|
117 |
+
"BLOCK_SIZE_K": 64,
|
118 |
+
"GROUP_SIZE_M": 32,
|
119 |
+
"num_warps": 8,
|
120 |
+
"num_stages": 3
|
121 |
+
},
|
122 |
+
"2048": {
|
123 |
+
"BLOCK_SIZE_M": 128,
|
124 |
+
"BLOCK_SIZE_N": 128,
|
125 |
+
"BLOCK_SIZE_K": 64,
|
126 |
+
"GROUP_SIZE_M": 1,
|
127 |
+
"num_warps": 4,
|
128 |
+
"num_stages": 3
|
129 |
+
},
|
130 |
+
"3072": {
|
131 |
+
"BLOCK_SIZE_M": 128,
|
132 |
+
"BLOCK_SIZE_N": 128,
|
133 |
+
"BLOCK_SIZE_K": 64,
|
134 |
+
"GROUP_SIZE_M": 32,
|
135 |
+
"num_warps": 4,
|
136 |
+
"num_stages": 3
|
137 |
+
},
|
138 |
+
"4096": {
|
139 |
+
"BLOCK_SIZE_M": 128,
|
140 |
+
"BLOCK_SIZE_N": 128,
|
141 |
+
"BLOCK_SIZE_K": 64,
|
142 |
+
"GROUP_SIZE_M": 1,
|
143 |
+
"num_warps": 4,
|
144 |
+
"num_stages": 3
|
145 |
+
}
|
146 |
+
}
|
build/torch25-cxx11-cu118-x86_64-linux/moe/configs/E=64,N=640,device_name=NVIDIA_GeForce_RTX_4090,dtype=fp8_w8a8.json
ADDED
@@ -0,0 +1,146 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"1": {
|
3 |
+
"BLOCK_SIZE_M": 16,
|
4 |
+
"BLOCK_SIZE_N": 32,
|
5 |
+
"BLOCK_SIZE_K": 128,
|
6 |
+
"GROUP_SIZE_M": 1,
|
7 |
+
"num_warps": 4,
|
8 |
+
"num_stages": 3
|
9 |
+
},
|
10 |
+
"2": {
|
11 |
+
"BLOCK_SIZE_M": 16,
|
12 |
+
"BLOCK_SIZE_N": 32,
|
13 |
+
"BLOCK_SIZE_K": 128,
|
14 |
+
"GROUP_SIZE_M": 32,
|
15 |
+
"num_warps": 4,
|
16 |
+
"num_stages": 2
|
17 |
+
},
|
18 |
+
"4": {
|
19 |
+
"BLOCK_SIZE_M": 16,
|
20 |
+
"BLOCK_SIZE_N": 64,
|
21 |
+
"BLOCK_SIZE_K": 256,
|
22 |
+
"GROUP_SIZE_M": 16,
|
23 |
+
"num_warps": 4,
|
24 |
+
"num_stages": 3
|
25 |
+
},
|
26 |
+
"8": {
|
27 |
+
"BLOCK_SIZE_M": 32,
|
28 |
+
"BLOCK_SIZE_N": 64,
|
29 |
+
"BLOCK_SIZE_K": 256,
|
30 |
+
"GROUP_SIZE_M": 32,
|
31 |
+
"num_warps": 8,
|
32 |
+
"num_stages": 3
|
33 |
+
},
|
34 |
+
"16": {
|
35 |
+
"BLOCK_SIZE_M": 16,
|
36 |
+
"BLOCK_SIZE_N": 32,
|
37 |
+
"BLOCK_SIZE_K": 256,
|
38 |
+
"GROUP_SIZE_M": 16,
|
39 |
+
"num_warps": 4,
|
40 |
+
"num_stages": 5
|
41 |
+
},
|
42 |
+
"24": {
|
43 |
+
"BLOCK_SIZE_M": 16,
|
44 |
+
"BLOCK_SIZE_N": 64,
|
45 |
+
"BLOCK_SIZE_K": 256,
|
46 |
+
"GROUP_SIZE_M": 64,
|
47 |
+
"num_warps": 4,
|
48 |
+
"num_stages": 2
|
49 |
+
},
|
50 |
+
"32": {
|
51 |
+
"BLOCK_SIZE_M": 16,
|
52 |
+
"BLOCK_SIZE_N": 32,
|
53 |
+
"BLOCK_SIZE_K": 256,
|
54 |
+
"GROUP_SIZE_M": 1,
|
55 |
+
"num_warps": 8,
|
56 |
+
"num_stages": 3
|
57 |
+
},
|
58 |
+
"48": {
|
59 |
+
"BLOCK_SIZE_M": 16,
|
60 |
+
"BLOCK_SIZE_N": 64,
|
61 |
+
"BLOCK_SIZE_K": 256,
|
62 |
+
"GROUP_SIZE_M": 64,
|
63 |
+
"num_warps": 8,
|
64 |
+
"num_stages": 2
|
65 |
+
},
|
66 |
+
"64": {
|
67 |
+
"BLOCK_SIZE_M": 32,
|
68 |
+
"BLOCK_SIZE_N": 32,
|
69 |
+
"BLOCK_SIZE_K": 256,
|
70 |
+
"GROUP_SIZE_M": 1,
|
71 |
+
"num_warps": 4,
|
72 |
+
"num_stages": 2
|
73 |
+
},
|
74 |
+
"96": {
|
75 |
+
"BLOCK_SIZE_M": 16,
|
76 |
+
"BLOCK_SIZE_N": 32,
|
77 |
+
"BLOCK_SIZE_K": 128,
|
78 |
+
"GROUP_SIZE_M": 16,
|
79 |
+
"num_warps": 4,
|
80 |
+
"num_stages": 3
|
81 |
+
},
|
82 |
+
"128": {
|
83 |
+
"BLOCK_SIZE_M": 32,
|
84 |
+
"BLOCK_SIZE_N": 64,
|
85 |
+
"BLOCK_SIZE_K": 256,
|
86 |
+
"GROUP_SIZE_M": 16,
|
87 |
+
"num_warps": 8,
|
88 |
+
"num_stages": 2
|
89 |
+
},
|
90 |
+
"256": {
|
91 |
+
"BLOCK_SIZE_M": 32,
|
92 |
+
"BLOCK_SIZE_N": 64,
|
93 |
+
"BLOCK_SIZE_K": 128,
|
94 |
+
"GROUP_SIZE_M": 16,
|
95 |
+
"num_warps": 8,
|
96 |
+
"num_stages": 3
|
97 |
+
},
|
98 |
+
"512": {
|
99 |
+
"BLOCK_SIZE_M": 32,
|
100 |
+
"BLOCK_SIZE_N": 64,
|
101 |
+
"BLOCK_SIZE_K": 128,
|
102 |
+
"GROUP_SIZE_M": 32,
|
103 |
+
"num_warps": 4,
|
104 |
+
"num_stages": 4
|
105 |
+
},
|
106 |
+
"1024": {
|
107 |
+
"BLOCK_SIZE_M": 64,
|
108 |
+
"BLOCK_SIZE_N": 128,
|
109 |
+
"BLOCK_SIZE_K": 128,
|
110 |
+
"GROUP_SIZE_M": 16,
|
111 |
+
"num_warps": 8,
|
112 |
+
"num_stages": 2
|
113 |
+
},
|
114 |
+
"1536": {
|
115 |
+
"BLOCK_SIZE_M": 128,
|
116 |
+
"BLOCK_SIZE_N": 128,
|
117 |
+
"BLOCK_SIZE_K": 64,
|
118 |
+
"GROUP_SIZE_M": 32,
|
119 |
+
"num_warps": 8,
|
120 |
+
"num_stages": 3
|
121 |
+
},
|
122 |
+
"2048": {
|
123 |
+
"BLOCK_SIZE_M": 128,
|
124 |
+
"BLOCK_SIZE_N": 128,
|
125 |
+
"BLOCK_SIZE_K": 64,
|
126 |
+
"GROUP_SIZE_M": 32,
|
127 |
+
"num_warps": 8,
|
128 |
+
"num_stages": 3
|
129 |
+
},
|
130 |
+
"3072": {
|
131 |
+
"BLOCK_SIZE_M": 128,
|
132 |
+
"BLOCK_SIZE_N": 128,
|
133 |
+
"BLOCK_SIZE_K": 64,
|
134 |
+
"GROUP_SIZE_M": 64,
|
135 |
+
"num_warps": 8,
|
136 |
+
"num_stages": 3
|
137 |
+
},
|
138 |
+
"4096": {
|
139 |
+
"BLOCK_SIZE_M": 128,
|
140 |
+
"BLOCK_SIZE_N": 128,
|
141 |
+
"BLOCK_SIZE_K": 64,
|
142 |
+
"GROUP_SIZE_M": 1,
|
143 |
+
"num_warps": 8,
|
144 |
+
"num_stages": 3
|
145 |
+
}
|
146 |
+
}
|
build/torch25-cxx11-cu118-x86_64-linux/moe/configs/E=64,N=640,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8.json
ADDED
@@ -0,0 +1,146 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"1": {
|
3 |
+
"BLOCK_SIZE_M": 64,
|
4 |
+
"BLOCK_SIZE_N": 128,
|
5 |
+
"BLOCK_SIZE_K": 128,
|
6 |
+
"GROUP_SIZE_M": 1,
|
7 |
+
"num_warps": 4,
|
8 |
+
"num_stages": 4
|
9 |
+
},
|
10 |
+
"2": {
|
11 |
+
"BLOCK_SIZE_M": 64,
|
12 |
+
"BLOCK_SIZE_N": 128,
|
13 |
+
"BLOCK_SIZE_K": 128,
|
14 |
+
"GROUP_SIZE_M": 16,
|
15 |
+
"num_warps": 4,
|
16 |
+
"num_stages": 4
|
17 |
+
},
|
18 |
+
"4": {
|
19 |
+
"BLOCK_SIZE_M": 64,
|
20 |
+
"BLOCK_SIZE_N": 128,
|
21 |
+
"BLOCK_SIZE_K": 128,
|
22 |
+
"GROUP_SIZE_M": 1,
|
23 |
+
"num_warps": 8,
|
24 |
+
"num_stages": 3
|
25 |
+
},
|
26 |
+
"8": {
|
27 |
+
"BLOCK_SIZE_M": 64,
|
28 |
+
"BLOCK_SIZE_N": 128,
|
29 |
+
"BLOCK_SIZE_K": 128,
|
30 |
+
"GROUP_SIZE_M": 1,
|
31 |
+
"num_warps": 4,
|
32 |
+
"num_stages": 4
|
33 |
+
},
|
34 |
+
"16": {
|
35 |
+
"BLOCK_SIZE_M": 64,
|
36 |
+
"BLOCK_SIZE_N": 64,
|
37 |
+
"BLOCK_SIZE_K": 128,
|
38 |
+
"GROUP_SIZE_M": 32,
|
39 |
+
"num_warps": 4,
|
40 |
+
"num_stages": 4
|
41 |
+
},
|
42 |
+
"24": {
|
43 |
+
"BLOCK_SIZE_M": 64,
|
44 |
+
"BLOCK_SIZE_N": 128,
|
45 |
+
"BLOCK_SIZE_K": 128,
|
46 |
+
"GROUP_SIZE_M": 16,
|
47 |
+
"num_warps": 4,
|
48 |
+
"num_stages": 3
|
49 |
+
},
|
50 |
+
"32": {
|
51 |
+
"BLOCK_SIZE_M": 64,
|
52 |
+
"BLOCK_SIZE_N": 128,
|
53 |
+
"BLOCK_SIZE_K": 128,
|
54 |
+
"GROUP_SIZE_M": 1,
|
55 |
+
"num_warps": 4,
|
56 |
+
"num_stages": 3
|
57 |
+
},
|
58 |
+
"48": {
|
59 |
+
"BLOCK_SIZE_M": 64,
|
60 |
+
"BLOCK_SIZE_N": 128,
|
61 |
+
"BLOCK_SIZE_K": 128,
|
62 |
+
"GROUP_SIZE_M": 1,
|
63 |
+
"num_warps": 4,
|
64 |
+
"num_stages": 3
|
65 |
+
},
|
66 |
+
"64": {
|
67 |
+
"BLOCK_SIZE_M": 64,
|
68 |
+
"BLOCK_SIZE_N": 128,
|
69 |
+
"BLOCK_SIZE_K": 128,
|
70 |
+
"GROUP_SIZE_M": 1,
|
71 |
+
"num_warps": 4,
|
72 |
+
"num_stages": 3
|
73 |
+
},
|
74 |
+
"96": {
|
75 |
+
"BLOCK_SIZE_M": 64,
|
76 |
+
"BLOCK_SIZE_N": 128,
|
77 |
+
"BLOCK_SIZE_K": 128,
|
78 |
+
"GROUP_SIZE_M": 1,
|
79 |
+
"num_warps": 8,
|
80 |
+
"num_stages": 3
|
81 |
+
},
|
82 |
+
"128": {
|
83 |
+
"BLOCK_SIZE_M": 64,
|
84 |
+
"BLOCK_SIZE_N": 128,
|
85 |
+
"BLOCK_SIZE_K": 128,
|
86 |
+
"GROUP_SIZE_M": 16,
|
87 |
+
"num_warps": 4,
|
88 |
+
"num_stages": 3
|
89 |
+
},
|
90 |
+
"256": {
|
91 |
+
"BLOCK_SIZE_M": 64,
|
92 |
+
"BLOCK_SIZE_N": 128,
|
93 |
+
"BLOCK_SIZE_K": 128,
|
94 |
+
"GROUP_SIZE_M": 1,
|
95 |
+
"num_warps": 4,
|
96 |
+
"num_stages": 3
|
97 |
+
},
|
98 |
+
"512": {
|
99 |
+
"BLOCK_SIZE_M": 64,
|
100 |
+
"BLOCK_SIZE_N": 128,
|
101 |
+
"BLOCK_SIZE_K": 128,
|
102 |
+
"GROUP_SIZE_M": 16,
|
103 |
+
"num_warps": 4,
|
104 |
+
"num_stages": 4
|
105 |
+
},
|
106 |
+
"1024": {
|
107 |
+
"BLOCK_SIZE_M": 128,
|
108 |
+
"BLOCK_SIZE_N": 256,
|
109 |
+
"BLOCK_SIZE_K": 128,
|
110 |
+
"GROUP_SIZE_M": 16,
|
111 |
+
"num_warps": 8,
|
112 |
+
"num_stages": 4
|
113 |
+
},
|
114 |
+
"1536": {
|
115 |
+
"BLOCK_SIZE_M": 128,
|
116 |
+
"BLOCK_SIZE_N": 256,
|
117 |
+
"BLOCK_SIZE_K": 128,
|
118 |
+
"GROUP_SIZE_M": 16,
|
119 |
+
"num_warps": 8,
|
120 |
+
"num_stages": 4
|
121 |
+
},
|
122 |
+
"2048": {
|
123 |
+
"BLOCK_SIZE_M": 128,
|
124 |
+
"BLOCK_SIZE_N": 256,
|
125 |
+
"BLOCK_SIZE_K": 128,
|
126 |
+
"GROUP_SIZE_M": 16,
|
127 |
+
"num_warps": 8,
|
128 |
+
"num_stages": 4
|
129 |
+
},
|
130 |
+
"3072": {
|
131 |
+
"BLOCK_SIZE_M": 128,
|
132 |
+
"BLOCK_SIZE_N": 256,
|
133 |
+
"BLOCK_SIZE_K": 128,
|
134 |
+
"GROUP_SIZE_M": 16,
|
135 |
+
"num_warps": 8,
|
136 |
+
"num_stages": 4
|
137 |
+
},
|
138 |
+
"4096": {
|
139 |
+
"BLOCK_SIZE_M": 128,
|
140 |
+
"BLOCK_SIZE_N": 256,
|
141 |
+
"BLOCK_SIZE_K": 128,
|
142 |
+
"GROUP_SIZE_M": 1,
|
143 |
+
"num_warps": 8,
|
144 |
+
"num_stages": 4
|
145 |
+
}
|
146 |
+
}
|
build/torch25-cxx11-cu118-x86_64-linux/moe/configs/E=64,N=640,device_name=NVIDIA_H200,dtype=fp8_w8a8.json
ADDED
@@ -0,0 +1,146 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"1": {
|
3 |
+
"BLOCK_SIZE_M": 64,
|
4 |
+
"BLOCK_SIZE_N": 128,
|
5 |
+
"BLOCK_SIZE_K": 128,
|
6 |
+
"GROUP_SIZE_M": 1,
|
7 |
+
"num_warps": 4,
|
8 |
+
"num_stages": 4
|
9 |
+
},
|
10 |
+
"2": {
|
11 |
+
"BLOCK_SIZE_M": 64,
|
12 |
+
"BLOCK_SIZE_N": 128,
|
13 |
+
"BLOCK_SIZE_K": 128,
|
14 |
+
"GROUP_SIZE_M": 16,
|
15 |
+
"num_warps": 4,
|
16 |
+
"num_stages": 4
|
17 |
+
},
|
18 |
+
"4": {
|
19 |
+
"BLOCK_SIZE_M": 64,
|
20 |
+
"BLOCK_SIZE_N": 128,
|
21 |
+
"BLOCK_SIZE_K": 128,
|
22 |
+
"GROUP_SIZE_M": 1,
|
23 |
+
"num_warps": 4,
|
24 |
+
"num_stages": 3
|
25 |
+
},
|
26 |
+
"8": {
|
27 |
+
"BLOCK_SIZE_M": 64,
|
28 |
+
"BLOCK_SIZE_N": 128,
|
29 |
+
"BLOCK_SIZE_K": 128,
|
30 |
+
"GROUP_SIZE_M": 64,
|
31 |
+
"num_warps": 4,
|
32 |
+
"num_stages": 4
|
33 |
+
},
|
34 |
+
"16": {
|
35 |
+
"BLOCK_SIZE_M": 64,
|
36 |
+
"BLOCK_SIZE_N": 128,
|
37 |
+
"BLOCK_SIZE_K": 128,
|
38 |
+
"GROUP_SIZE_M": 64,
|
39 |
+
"num_warps": 4,
|
40 |
+
"num_stages": 3
|
41 |
+
},
|
42 |
+
"24": {
|
43 |
+
"BLOCK_SIZE_M": 64,
|
44 |
+
"BLOCK_SIZE_N": 128,
|
45 |
+
"BLOCK_SIZE_K": 128,
|
46 |
+
"GROUP_SIZE_M": 16,
|
47 |
+
"num_warps": 4,
|
48 |
+
"num_stages": 3
|
49 |
+
},
|
50 |
+
"32": {
|
51 |
+
"BLOCK_SIZE_M": 64,
|
52 |
+
"BLOCK_SIZE_N": 128,
|
53 |
+
"BLOCK_SIZE_K": 128,
|
54 |
+
"GROUP_SIZE_M": 1,
|
55 |
+
"num_warps": 4,
|
56 |
+
"num_stages": 3
|
57 |
+
},
|
58 |
+
"48": {
|
59 |
+
"BLOCK_SIZE_M": 64,
|
60 |
+
"BLOCK_SIZE_N": 128,
|
61 |
+
"BLOCK_SIZE_K": 128,
|
62 |
+
"GROUP_SIZE_M": 1,
|
63 |
+
"num_warps": 4,
|
64 |
+
"num_stages": 3
|
65 |
+
},
|
66 |
+
"64": {
|
67 |
+
"BLOCK_SIZE_M": 64,
|
68 |
+
"BLOCK_SIZE_N": 128,
|
69 |
+
"BLOCK_SIZE_K": 128,
|
70 |
+
"GROUP_SIZE_M": 1,
|
71 |
+
"num_warps": 4,
|
72 |
+
"num_stages": 3
|
73 |
+
},
|
74 |
+
"96": {
|
75 |
+
"BLOCK_SIZE_M": 64,
|
76 |
+
"BLOCK_SIZE_N": 128,
|
77 |
+
"BLOCK_SIZE_K": 128,
|
78 |
+
"GROUP_SIZE_M": 32,
|
79 |
+
"num_warps": 4,
|
80 |
+
"num_stages": 3
|
81 |
+
},
|
82 |
+
"128": {
|
83 |
+
"BLOCK_SIZE_M": 64,
|
84 |
+
"BLOCK_SIZE_N": 128,
|
85 |
+
"BLOCK_SIZE_K": 128,
|
86 |
+
"GROUP_SIZE_M": 1,
|
87 |
+
"num_warps": 4,
|
88 |
+
"num_stages": 3
|
89 |
+
},
|
90 |
+
"256": {
|
91 |
+
"BLOCK_SIZE_M": 64,
|
92 |
+
"BLOCK_SIZE_N": 128,
|
93 |
+
"BLOCK_SIZE_K": 128,
|
94 |
+
"GROUP_SIZE_M": 64,
|
95 |
+
"num_warps": 4,
|
96 |
+
"num_stages": 3
|
97 |
+
},
|
98 |
+
"512": {
|
99 |
+
"BLOCK_SIZE_M": 64,
|
100 |
+
"BLOCK_SIZE_N": 128,
|
101 |
+
"BLOCK_SIZE_K": 128,
|
102 |
+
"GROUP_SIZE_M": 16,
|
103 |
+
"num_warps": 4,
|
104 |
+
"num_stages": 2
|
105 |
+
},
|
106 |
+
"1024": {
|
107 |
+
"BLOCK_SIZE_M": 128,
|
108 |
+
"BLOCK_SIZE_N": 256,
|
109 |
+
"BLOCK_SIZE_K": 128,
|
110 |
+
"GROUP_SIZE_M": 16,
|
111 |
+
"num_warps": 8,
|
112 |
+
"num_stages": 4
|
113 |
+
},
|
114 |
+
"1536": {
|
115 |
+
"BLOCK_SIZE_M": 128,
|
116 |
+
"BLOCK_SIZE_N": 256,
|
117 |
+
"BLOCK_SIZE_K": 128,
|
118 |
+
"GROUP_SIZE_M": 16,
|
119 |
+
"num_warps": 8,
|
120 |
+
"num_stages": 4
|
121 |
+
},
|
122 |
+
"2048": {
|
123 |
+
"BLOCK_SIZE_M": 128,
|
124 |
+
"BLOCK_SIZE_N": 256,
|
125 |
+
"BLOCK_SIZE_K": 128,
|
126 |
+
"GROUP_SIZE_M": 1,
|
127 |
+
"num_warps": 8,
|
128 |
+
"num_stages": 4
|
129 |
+
},
|
130 |
+
"3072": {
|
131 |
+
"BLOCK_SIZE_M": 128,
|
132 |
+
"BLOCK_SIZE_N": 256,
|
133 |
+
"BLOCK_SIZE_K": 128,
|
134 |
+
"GROUP_SIZE_M": 16,
|
135 |
+
"num_warps": 8,
|
136 |
+
"num_stages": 4
|
137 |
+
},
|
138 |
+
"4096": {
|
139 |
+
"BLOCK_SIZE_M": 128,
|
140 |
+
"BLOCK_SIZE_N": 256,
|
141 |
+
"BLOCK_SIZE_K": 128,
|
142 |
+
"GROUP_SIZE_M": 16,
|
143 |
+
"num_warps": 8,
|
144 |
+
"num_stages": 4
|
145 |
+
}
|
146 |
+
}
|
build/torch25-cxx11-cu118-x86_64-linux/moe/configs/E=64,N=640,device_name=NVIDIA_H200.json
ADDED
@@ -0,0 +1,146 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"1": {
|
3 |
+
"BLOCK_SIZE_M": 16,
|
4 |
+
"BLOCK_SIZE_N": 128,
|
5 |
+
"BLOCK_SIZE_K": 128,
|
6 |
+
"GROUP_SIZE_M": 32,
|
7 |
+
"num_warps": 4,
|
8 |
+
"num_stages": 5
|
9 |
+
},
|
10 |
+
"2": {
|
11 |
+
"BLOCK_SIZE_M": 16,
|
12 |
+
"BLOCK_SIZE_N": 64,
|
13 |
+
"BLOCK_SIZE_K": 64,
|
14 |
+
"GROUP_SIZE_M": 1,
|
15 |
+
"num_warps": 4,
|
16 |
+
"num_stages": 4
|
17 |
+
},
|
18 |
+
"4": {
|
19 |
+
"BLOCK_SIZE_M": 64,
|
20 |
+
"BLOCK_SIZE_N": 128,
|
21 |
+
"BLOCK_SIZE_K": 64,
|
22 |
+
"GROUP_SIZE_M": 32,
|
23 |
+
"num_warps": 4,
|
24 |
+
"num_stages": 3
|
25 |
+
},
|
26 |
+
"8": {
|
27 |
+
"BLOCK_SIZE_M": 16,
|
28 |
+
"BLOCK_SIZE_N": 128,
|
29 |
+
"BLOCK_SIZE_K": 128,
|
30 |
+
"GROUP_SIZE_M": 32,
|
31 |
+
"num_warps": 4,
|
32 |
+
"num_stages": 4
|
33 |
+
},
|
34 |
+
"16": {
|
35 |
+
"BLOCK_SIZE_M": 16,
|
36 |
+
"BLOCK_SIZE_N": 128,
|
37 |
+
"BLOCK_SIZE_K": 128,
|
38 |
+
"GROUP_SIZE_M": 32,
|
39 |
+
"num_warps": 4,
|
40 |
+
"num_stages": 3
|
41 |
+
},
|
42 |
+
"24": {
|
43 |
+
"BLOCK_SIZE_M": 16,
|
44 |
+
"BLOCK_SIZE_N": 64,
|
45 |
+
"BLOCK_SIZE_K": 128,
|
46 |
+
"GROUP_SIZE_M": 16,
|
47 |
+
"num_warps": 4,
|
48 |
+
"num_stages": 3
|
49 |
+
},
|
50 |
+
"32": {
|
51 |
+
"BLOCK_SIZE_M": 16,
|
52 |
+
"BLOCK_SIZE_N": 64,
|
53 |
+
"BLOCK_SIZE_K": 128,
|
54 |
+
"GROUP_SIZE_M": 16,
|
55 |
+
"num_warps": 4,
|
56 |
+
"num_stages": 3
|
57 |
+
},
|
58 |
+
"48": {
|
59 |
+
"BLOCK_SIZE_M": 16,
|
60 |
+
"BLOCK_SIZE_N": 64,
|
61 |
+
"BLOCK_SIZE_K": 128,
|
62 |
+
"GROUP_SIZE_M": 16,
|
63 |
+
"num_warps": 4,
|
64 |
+
"num_stages": 3
|
65 |
+
},
|
66 |
+
"64": {
|
67 |
+
"BLOCK_SIZE_M": 16,
|
68 |
+
"BLOCK_SIZE_N": 64,
|
69 |
+
"BLOCK_SIZE_K": 128,
|
70 |
+
"GROUP_SIZE_M": 1,
|
71 |
+
"num_warps": 4,
|
72 |
+
"num_stages": 3
|
73 |
+
},
|
74 |
+
"96": {
|
75 |
+
"BLOCK_SIZE_M": 16,
|
76 |
+
"BLOCK_SIZE_N": 64,
|
77 |
+
"BLOCK_SIZE_K": 128,
|
78 |
+
"GROUP_SIZE_M": 32,
|
79 |
+
"num_warps": 4,
|
80 |
+
"num_stages": 2
|
81 |
+
},
|
82 |
+
"128": {
|
83 |
+
"BLOCK_SIZE_M": 64,
|
84 |
+
"BLOCK_SIZE_N": 128,
|
85 |
+
"BLOCK_SIZE_K": 64,
|
86 |
+
"GROUP_SIZE_M": 1,
|
87 |
+
"num_warps": 4,
|
88 |
+
"num_stages": 3
|
89 |
+
},
|
90 |
+
"256": {
|
91 |
+
"BLOCK_SIZE_M": 64,
|
92 |
+
"BLOCK_SIZE_N": 128,
|
93 |
+
"BLOCK_SIZE_K": 64,
|
94 |
+
"GROUP_SIZE_M": 1,
|
95 |
+
"num_warps": 4,
|
96 |
+
"num_stages": 3
|
97 |
+
},
|
98 |
+
"512": {
|
99 |
+
"BLOCK_SIZE_M": 128,
|
100 |
+
"BLOCK_SIZE_N": 256,
|
101 |
+
"BLOCK_SIZE_K": 64,
|
102 |
+
"GROUP_SIZE_M": 16,
|
103 |
+
"num_warps": 8,
|
104 |
+
"num_stages": 4
|
105 |
+
},
|
106 |
+
"1024": {
|
107 |
+
"BLOCK_SIZE_M": 128,
|
108 |
+
"BLOCK_SIZE_N": 256,
|
109 |
+
"BLOCK_SIZE_K": 64,
|
110 |
+
"GROUP_SIZE_M": 1,
|
111 |
+
"num_warps": 8,
|
112 |
+
"num_stages": 4
|
113 |
+
},
|
114 |
+
"1536": {
|
115 |
+
"BLOCK_SIZE_M": 128,
|
116 |
+
"BLOCK_SIZE_N": 256,
|
117 |
+
"BLOCK_SIZE_K": 64,
|
118 |
+
"GROUP_SIZE_M": 16,
|
119 |
+
"num_warps": 8,
|
120 |
+
"num_stages": 4
|
121 |
+
},
|
122 |
+
"2048": {
|
123 |
+
"BLOCK_SIZE_M": 128,
|
124 |
+
"BLOCK_SIZE_N": 256,
|
125 |
+
"BLOCK_SIZE_K": 64,
|
126 |
+
"GROUP_SIZE_M": 1,
|
127 |
+
"num_warps": 8,
|
128 |
+
"num_stages": 4
|
129 |
+
},
|
130 |
+
"3072": {
|
131 |
+
"BLOCK_SIZE_M": 128,
|
132 |
+
"BLOCK_SIZE_N": 256,
|
133 |
+
"BLOCK_SIZE_K": 64,
|
134 |
+
"GROUP_SIZE_M": 16,
|
135 |
+
"num_warps": 8,
|
136 |
+
"num_stages": 4
|
137 |
+
},
|
138 |
+
"4096": {
|
139 |
+
"BLOCK_SIZE_M": 128,
|
140 |
+
"BLOCK_SIZE_N": 256,
|
141 |
+
"BLOCK_SIZE_K": 64,
|
142 |
+
"GROUP_SIZE_M": 1,
|
143 |
+
"num_warps": 8,
|
144 |
+
"num_stages": 4
|
145 |
+
}
|
146 |
+
}
|
build/torch25-cxx11-cu118-x86_64-linux/moe/configs/E=8,N=14336,device_name=AMD_Instinct_MI300X,dtype=fp8_w8a8.json
ADDED
@@ -0,0 +1,164 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"1": {
|
3 |
+
"BLOCK_SIZE_M": 16,
|
4 |
+
"BLOCK_SIZE_N": 64,
|
5 |
+
"BLOCK_SIZE_K": 256,
|
6 |
+
"GROUP_SIZE_M": 1,
|
7 |
+
"num_warps": 4,
|
8 |
+
"num_stages": 2,
|
9 |
+
"waves_per_eu": 0
|
10 |
+
},
|
11 |
+
"2": {
|
12 |
+
"BLOCK_SIZE_M": 16,
|
13 |
+
"BLOCK_SIZE_N": 16,
|
14 |
+
"BLOCK_SIZE_K": 256,
|
15 |
+
"GROUP_SIZE_M": 1,
|
16 |
+
"num_warps": 2,
|
17 |
+
"num_stages": 2,
|
18 |
+
"waves_per_eu": 0
|
19 |
+
},
|
20 |
+
"4": {
|
21 |
+
"BLOCK_SIZE_M": 16,
|
22 |
+
"BLOCK_SIZE_N": 64,
|
23 |
+
"BLOCK_SIZE_K": 256,
|
24 |
+
"GROUP_SIZE_M": 1,
|
25 |
+
"num_warps": 4,
|
26 |
+
"num_stages": 2,
|
27 |
+
"waves_per_eu": 0
|
28 |
+
},
|
29 |
+
"8": {
|
30 |
+
"BLOCK_SIZE_M": 16,
|
31 |
+
"BLOCK_SIZE_N": 32,
|
32 |
+
"BLOCK_SIZE_K": 256,
|
33 |
+
"GROUP_SIZE_M": 1,
|
34 |
+
"num_warps": 2,
|
35 |
+
"num_stages": 2,
|
36 |
+
"waves_per_eu": 0
|
37 |
+
},
|
38 |
+
"16": {
|
39 |
+
"BLOCK_SIZE_M": 16,
|
40 |
+
"BLOCK_SIZE_N": 64,
|
41 |
+
"BLOCK_SIZE_K": 256,
|
42 |
+
"GROUP_SIZE_M": 1,
|
43 |
+
"num_warps": 2,
|
44 |
+
"num_stages": 2,
|
45 |
+
"waves_per_eu": 0
|
46 |
+
},
|
47 |
+
"24": {
|
48 |
+
"BLOCK_SIZE_M": 16,
|
49 |
+
"BLOCK_SIZE_N": 64,
|
50 |
+
"BLOCK_SIZE_K": 256,
|
51 |
+
"GROUP_SIZE_M": 1,
|
52 |
+
"num_warps": 2,
|
53 |
+
"num_stages": 2,
|
54 |
+
"waves_per_eu": 0
|
55 |
+
},
|
56 |
+
"32": {
|
57 |
+
"BLOCK_SIZE_M": 16,
|
58 |
+
"BLOCK_SIZE_N": 32,
|
59 |
+
"BLOCK_SIZE_K": 256,
|
60 |
+
"GROUP_SIZE_M": 4,
|
61 |
+
"num_warps": 2,
|
62 |
+
"num_stages": 2,
|
63 |
+
"waves_per_eu": 0
|
64 |
+
},
|
65 |
+
"48": {
|
66 |
+
"BLOCK_SIZE_M": 16,
|
67 |
+
"BLOCK_SIZE_N": 64,
|
68 |
+
"BLOCK_SIZE_K": 128,
|
69 |
+
"GROUP_SIZE_M": 4,
|
70 |
+
"num_warps": 4,
|
71 |
+
"num_stages": 2,
|
72 |
+
"waves_per_eu": 0
|
73 |
+
},
|
74 |
+
"64": {
|
75 |
+
"BLOCK_SIZE_M": 32,
|
76 |
+
"BLOCK_SIZE_N": 64,
|
77 |
+
"BLOCK_SIZE_K": 256,
|
78 |
+
"GROUP_SIZE_M": 4,
|
79 |
+
"num_warps": 2,
|
80 |
+
"num_stages": 2,
|
81 |
+
"waves_per_eu": 0
|
82 |
+
},
|
83 |
+
"96": {
|
84 |
+
"BLOCK_SIZE_M": 32,
|
85 |
+
"BLOCK_SIZE_N": 64,
|
86 |
+
"BLOCK_SIZE_K": 256,
|
87 |
+
"GROUP_SIZE_M": 1,
|
88 |
+
"num_warps": 2,
|
89 |
+
"num_stages": 2,
|
90 |
+
"waves_per_eu": 0
|
91 |
+
},
|
92 |
+
"128": {
|
93 |
+
"BLOCK_SIZE_M": 64,
|
94 |
+
"BLOCK_SIZE_N": 128,
|
95 |
+
"BLOCK_SIZE_K": 256,
|
96 |
+
"GROUP_SIZE_M": 4,
|
97 |
+
"num_warps": 8,
|
98 |
+
"num_stages": 2,
|
99 |
+
"waves_per_eu": 0
|
100 |
+
},
|
101 |
+
"256": {
|
102 |
+
"BLOCK_SIZE_M": 128,
|
103 |
+
"BLOCK_SIZE_N": 128,
|
104 |
+
"BLOCK_SIZE_K": 256,
|
105 |
+
"GROUP_SIZE_M": 4,
|
106 |
+
"num_warps": 8,
|
107 |
+
"num_stages": 2,
|
108 |
+
"waves_per_eu": 0
|
109 |
+
},
|
110 |
+
"512": {
|
111 |
+
"BLOCK_SIZE_M": 256,
|
112 |
+
"BLOCK_SIZE_N": 128,
|
113 |
+
"BLOCK_SIZE_K": 128,
|
114 |
+
"GROUP_SIZE_M": 4,
|
115 |
+
"num_warps": 8,
|
116 |
+
"num_stages": 2,
|
117 |
+
"waves_per_eu": 0
|
118 |
+
},
|
119 |
+
"1024": {
|
120 |
+
"BLOCK_SIZE_M": 128,
|
121 |
+
"BLOCK_SIZE_N": 128,
|
122 |
+
"BLOCK_SIZE_K": 64,
|
123 |
+
"GROUP_SIZE_M": 1,
|
124 |
+
"num_warps": 4,
|
125 |
+
"num_stages": 2,
|
126 |
+
"waves_per_eu": 0
|
127 |
+
},
|
128 |
+
"1536": {
|
129 |
+
"BLOCK_SIZE_M": 128,
|
130 |
+
"BLOCK_SIZE_N": 256,
|
131 |
+
"BLOCK_SIZE_K": 128,
|
132 |
+
"GROUP_SIZE_M": 1,
|
133 |
+
"num_warps": 8,
|
134 |
+
"num_stages": 2,
|
135 |
+
"waves_per_eu": 0
|
136 |
+
},
|
137 |
+
"2048": {
|
138 |
+
"BLOCK_SIZE_M": 128,
|
139 |
+
"BLOCK_SIZE_N": 256,
|
140 |
+
"BLOCK_SIZE_K": 128,
|
141 |
+
"GROUP_SIZE_M": 1,
|
142 |
+
"num_warps": 8,
|
143 |
+
"num_stages": 2,
|
144 |
+
"waves_per_eu": 0
|
145 |
+
},
|
146 |
+
"3072": {
|
147 |
+
"BLOCK_SIZE_M": 128,
|
148 |
+
"BLOCK_SIZE_N": 256,
|
149 |
+
"BLOCK_SIZE_K": 128,
|
150 |
+
"GROUP_SIZE_M": 1,
|
151 |
+
"num_warps": 8,
|
152 |
+
"num_stages": 2,
|
153 |
+
"waves_per_eu": 0
|
154 |
+
},
|
155 |
+
"4096": {
|
156 |
+
"BLOCK_SIZE_M": 256,
|
157 |
+
"BLOCK_SIZE_N": 256,
|
158 |
+
"BLOCK_SIZE_K": 64,
|
159 |
+
"GROUP_SIZE_M": 1,
|
160 |
+
"num_warps": 8,
|
161 |
+
"num_stages": 2,
|
162 |
+
"waves_per_eu": 0
|
163 |
+
}
|
164 |
+
}
|
build/torch25-cxx11-cu118-x86_64-linux/moe/configs/E=8,N=14336,device_name=AMD_Instinct_MI300X.json
CHANGED
@@ -1,123 +1,123 @@
|
|
1 |
{
|
2 |
"1": {
|
3 |
"BLOCK_SIZE_M": 16,
|
4 |
-
"BLOCK_SIZE_N":
|
5 |
"BLOCK_SIZE_K": 256,
|
6 |
"GROUP_SIZE_M": 1,
|
7 |
"num_warps": 2,
|
8 |
-
"num_stages":
|
9 |
"waves_per_eu": 0,
|
10 |
"matrix_instr_nonkdim": 16,
|
11 |
-
"kpack":
|
12 |
},
|
13 |
"2": {
|
14 |
"BLOCK_SIZE_M": 16,
|
15 |
"BLOCK_SIZE_N": 16,
|
16 |
-
"BLOCK_SIZE_K":
|
17 |
"GROUP_SIZE_M": 1,
|
18 |
-
"num_warps":
|
19 |
-
"num_stages":
|
20 |
"waves_per_eu": 0,
|
21 |
"matrix_instr_nonkdim": 16,
|
22 |
"kpack": 2
|
23 |
},
|
24 |
"4": {
|
25 |
"BLOCK_SIZE_M": 16,
|
26 |
-
"BLOCK_SIZE_N":
|
27 |
-
"BLOCK_SIZE_K":
|
28 |
"GROUP_SIZE_M": 1,
|
29 |
-
"num_warps":
|
30 |
-
"num_stages":
|
31 |
"waves_per_eu": 0,
|
32 |
"matrix_instr_nonkdim": 16,
|
33 |
"kpack": 2
|
34 |
},
|
35 |
"8": {
|
36 |
"BLOCK_SIZE_M": 16,
|
37 |
-
"BLOCK_SIZE_N":
|
38 |
-
"BLOCK_SIZE_K":
|
39 |
"GROUP_SIZE_M": 1,
|
40 |
-
"num_warps":
|
41 |
-
"num_stages":
|
42 |
"waves_per_eu": 0,
|
43 |
"matrix_instr_nonkdim": 16,
|
44 |
"kpack": 2
|
45 |
},
|
46 |
"16": {
|
47 |
"BLOCK_SIZE_M": 16,
|
48 |
-
"BLOCK_SIZE_N":
|
49 |
-
"BLOCK_SIZE_K":
|
50 |
"GROUP_SIZE_M": 1,
|
51 |
-
"num_warps":
|
52 |
-
"num_stages":
|
53 |
"waves_per_eu": 0,
|
54 |
"matrix_instr_nonkdim": 16,
|
55 |
"kpack": 2
|
56 |
},
|
57 |
"24": {
|
58 |
"BLOCK_SIZE_M": 16,
|
59 |
-
"BLOCK_SIZE_N":
|
60 |
-
"BLOCK_SIZE_K":
|
61 |
"GROUP_SIZE_M": 1,
|
62 |
-
"num_warps":
|
63 |
-
"num_stages":
|
64 |
"waves_per_eu": 0,
|
65 |
"matrix_instr_nonkdim": 16,
|
66 |
"kpack": 2
|
67 |
},
|
68 |
"32": {
|
69 |
"BLOCK_SIZE_M": 16,
|
70 |
-
"BLOCK_SIZE_N":
|
71 |
-
"BLOCK_SIZE_K":
|
72 |
"GROUP_SIZE_M": 4,
|
73 |
"num_warps": 2,
|
74 |
-
"num_stages":
|
75 |
"waves_per_eu": 0,
|
76 |
"matrix_instr_nonkdim": 16,
|
77 |
-
"kpack":
|
78 |
},
|
79 |
"48": {
|
80 |
"BLOCK_SIZE_M": 16,
|
81 |
-
"BLOCK_SIZE_N":
|
82 |
"BLOCK_SIZE_K": 128,
|
83 |
"GROUP_SIZE_M": 4,
|
84 |
-
"num_warps":
|
85 |
-
"num_stages":
|
86 |
"waves_per_eu": 0,
|
87 |
"matrix_instr_nonkdim": 16,
|
88 |
-
"kpack":
|
89 |
},
|
90 |
"64": {
|
91 |
"BLOCK_SIZE_M": 32,
|
92 |
"BLOCK_SIZE_N": 64,
|
93 |
"BLOCK_SIZE_K": 128,
|
94 |
"GROUP_SIZE_M": 4,
|
95 |
-
"num_warps":
|
96 |
-
"num_stages":
|
97 |
"waves_per_eu": 0,
|
98 |
"matrix_instr_nonkdim": 16,
|
99 |
"kpack": 2
|
100 |
},
|
101 |
"96": {
|
102 |
"BLOCK_SIZE_M": 32,
|
103 |
-
"BLOCK_SIZE_N":
|
104 |
-
"BLOCK_SIZE_K":
|
105 |
"GROUP_SIZE_M": 4,
|
106 |
-
"num_warps":
|
107 |
-
"num_stages":
|
108 |
"waves_per_eu": 0,
|
109 |
"matrix_instr_nonkdim": 16,
|
110 |
-
"kpack":
|
111 |
},
|
112 |
"128": {
|
113 |
"BLOCK_SIZE_M": 64,
|
114 |
"BLOCK_SIZE_N": 64,
|
115 |
-
"BLOCK_SIZE_K":
|
116 |
"GROUP_SIZE_M": 4,
|
117 |
-
"num_warps":
|
118 |
-
"num_stages":
|
119 |
"waves_per_eu": 0,
|
120 |
-
"matrix_instr_nonkdim":
|
121 |
"kpack": 2
|
122 |
},
|
123 |
"256": {
|
@@ -126,10 +126,10 @@
|
|
126 |
"BLOCK_SIZE_K": 64,
|
127 |
"GROUP_SIZE_M": 4,
|
128 |
"num_warps": 8,
|
129 |
-
"num_stages":
|
130 |
"waves_per_eu": 0,
|
131 |
"matrix_instr_nonkdim": 16,
|
132 |
-
"kpack":
|
133 |
},
|
134 |
"512": {
|
135 |
"BLOCK_SIZE_M": 128,
|
@@ -137,7 +137,7 @@
|
|
137 |
"BLOCK_SIZE_K": 64,
|
138 |
"GROUP_SIZE_M": 4,
|
139 |
"num_warps": 8,
|
140 |
-
"num_stages":
|
141 |
"waves_per_eu": 0,
|
142 |
"matrix_instr_nonkdim": 16,
|
143 |
"kpack": 2
|
@@ -148,9 +148,9 @@
|
|
148 |
"BLOCK_SIZE_K": 64,
|
149 |
"GROUP_SIZE_M": 1,
|
150 |
"num_warps": 8,
|
151 |
-
"num_stages":
|
152 |
"waves_per_eu": 0,
|
153 |
-
"matrix_instr_nonkdim":
|
154 |
"kpack": 2
|
155 |
},
|
156 |
"1536": {
|
@@ -159,7 +159,7 @@
|
|
159 |
"BLOCK_SIZE_K": 64,
|
160 |
"GROUP_SIZE_M": 1,
|
161 |
"num_warps": 8,
|
162 |
-
"num_stages":
|
163 |
"waves_per_eu": 0,
|
164 |
"matrix_instr_nonkdim": 16,
|
165 |
"kpack": 2
|
@@ -170,7 +170,7 @@
|
|
170 |
"BLOCK_SIZE_K": 64,
|
171 |
"GROUP_SIZE_M": 1,
|
172 |
"num_warps": 8,
|
173 |
-
"num_stages":
|
174 |
"waves_per_eu": 0,
|
175 |
"matrix_instr_nonkdim": 16,
|
176 |
"kpack": 2
|
@@ -181,10 +181,10 @@
|
|
181 |
"BLOCK_SIZE_K": 64,
|
182 |
"GROUP_SIZE_M": 1,
|
183 |
"num_warps": 8,
|
184 |
-
"num_stages":
|
185 |
"waves_per_eu": 0,
|
186 |
"matrix_instr_nonkdim": 16,
|
187 |
-
"kpack":
|
188 |
},
|
189 |
"4096": {
|
190 |
"BLOCK_SIZE_M": 128,
|
@@ -192,9 +192,9 @@
|
|
192 |
"BLOCK_SIZE_K": 64,
|
193 |
"GROUP_SIZE_M": 1,
|
194 |
"num_warps": 8,
|
195 |
-
"num_stages":
|
196 |
"waves_per_eu": 0,
|
197 |
"matrix_instr_nonkdim": 16,
|
198 |
-
"kpack":
|
199 |
}
|
200 |
}
|
|
|
1 |
{
|
2 |
"1": {
|
3 |
"BLOCK_SIZE_M": 16,
|
4 |
+
"BLOCK_SIZE_N": 16,
|
5 |
"BLOCK_SIZE_K": 256,
|
6 |
"GROUP_SIZE_M": 1,
|
7 |
"num_warps": 2,
|
8 |
+
"num_stages": 2,
|
9 |
"waves_per_eu": 0,
|
10 |
"matrix_instr_nonkdim": 16,
|
11 |
+
"kpack": 2
|
12 |
},
|
13 |
"2": {
|
14 |
"BLOCK_SIZE_M": 16,
|
15 |
"BLOCK_SIZE_N": 16,
|
16 |
+
"BLOCK_SIZE_K": 256,
|
17 |
"GROUP_SIZE_M": 1,
|
18 |
+
"num_warps": 4,
|
19 |
+
"num_stages": 2,
|
20 |
"waves_per_eu": 0,
|
21 |
"matrix_instr_nonkdim": 16,
|
22 |
"kpack": 2
|
23 |
},
|
24 |
"4": {
|
25 |
"BLOCK_SIZE_M": 16,
|
26 |
+
"BLOCK_SIZE_N": 16,
|
27 |
+
"BLOCK_SIZE_K": 128,
|
28 |
"GROUP_SIZE_M": 1,
|
29 |
+
"num_warps": 1,
|
30 |
+
"num_stages": 2,
|
31 |
"waves_per_eu": 0,
|
32 |
"matrix_instr_nonkdim": 16,
|
33 |
"kpack": 2
|
34 |
},
|
35 |
"8": {
|
36 |
"BLOCK_SIZE_M": 16,
|
37 |
+
"BLOCK_SIZE_N": 64,
|
38 |
+
"BLOCK_SIZE_K": 64,
|
39 |
"GROUP_SIZE_M": 1,
|
40 |
+
"num_warps": 2,
|
41 |
+
"num_stages": 2,
|
42 |
"waves_per_eu": 0,
|
43 |
"matrix_instr_nonkdim": 16,
|
44 |
"kpack": 2
|
45 |
},
|
46 |
"16": {
|
47 |
"BLOCK_SIZE_M": 16,
|
48 |
+
"BLOCK_SIZE_N": 64,
|
49 |
+
"BLOCK_SIZE_K": 64,
|
50 |
"GROUP_SIZE_M": 1,
|
51 |
+
"num_warps": 2,
|
52 |
+
"num_stages": 2,
|
53 |
"waves_per_eu": 0,
|
54 |
"matrix_instr_nonkdim": 16,
|
55 |
"kpack": 2
|
56 |
},
|
57 |
"24": {
|
58 |
"BLOCK_SIZE_M": 16,
|
59 |
+
"BLOCK_SIZE_N": 16,
|
60 |
+
"BLOCK_SIZE_K": 256,
|
61 |
"GROUP_SIZE_M": 1,
|
62 |
+
"num_warps": 2,
|
63 |
+
"num_stages": 2,
|
64 |
"waves_per_eu": 0,
|
65 |
"matrix_instr_nonkdim": 16,
|
66 |
"kpack": 2
|
67 |
},
|
68 |
"32": {
|
69 |
"BLOCK_SIZE_M": 16,
|
70 |
+
"BLOCK_SIZE_N": 32,
|
71 |
+
"BLOCK_SIZE_K": 256,
|
72 |
"GROUP_SIZE_M": 4,
|
73 |
"num_warps": 2,
|
74 |
+
"num_stages": 2,
|
75 |
"waves_per_eu": 0,
|
76 |
"matrix_instr_nonkdim": 16,
|
77 |
+
"kpack": 2
|
78 |
},
|
79 |
"48": {
|
80 |
"BLOCK_SIZE_M": 16,
|
81 |
+
"BLOCK_SIZE_N": 64,
|
82 |
"BLOCK_SIZE_K": 128,
|
83 |
"GROUP_SIZE_M": 4,
|
84 |
+
"num_warps": 4,
|
85 |
+
"num_stages": 2,
|
86 |
"waves_per_eu": 0,
|
87 |
"matrix_instr_nonkdim": 16,
|
88 |
+
"kpack": 1
|
89 |
},
|
90 |
"64": {
|
91 |
"BLOCK_SIZE_M": 32,
|
92 |
"BLOCK_SIZE_N": 64,
|
93 |
"BLOCK_SIZE_K": 128,
|
94 |
"GROUP_SIZE_M": 4,
|
95 |
+
"num_warps": 4,
|
96 |
+
"num_stages": 2,
|
97 |
"waves_per_eu": 0,
|
98 |
"matrix_instr_nonkdim": 16,
|
99 |
"kpack": 2
|
100 |
},
|
101 |
"96": {
|
102 |
"BLOCK_SIZE_M": 32,
|
103 |
+
"BLOCK_SIZE_N": 64,
|
104 |
+
"BLOCK_SIZE_K": 256,
|
105 |
"GROUP_SIZE_M": 4,
|
106 |
+
"num_warps": 8,
|
107 |
+
"num_stages": 2,
|
108 |
"waves_per_eu": 0,
|
109 |
"matrix_instr_nonkdim": 16,
|
110 |
+
"kpack": 1
|
111 |
},
|
112 |
"128": {
|
113 |
"BLOCK_SIZE_M": 64,
|
114 |
"BLOCK_SIZE_N": 64,
|
115 |
+
"BLOCK_SIZE_K": 128,
|
116 |
"GROUP_SIZE_M": 4,
|
117 |
+
"num_warps": 4,
|
118 |
+
"num_stages": 2,
|
119 |
"waves_per_eu": 0,
|
120 |
+
"matrix_instr_nonkdim": 32,
|
121 |
"kpack": 2
|
122 |
},
|
123 |
"256": {
|
|
|
126 |
"BLOCK_SIZE_K": 64,
|
127 |
"GROUP_SIZE_M": 4,
|
128 |
"num_warps": 8,
|
129 |
+
"num_stages": 2,
|
130 |
"waves_per_eu": 0,
|
131 |
"matrix_instr_nonkdim": 16,
|
132 |
+
"kpack": 2
|
133 |
},
|
134 |
"512": {
|
135 |
"BLOCK_SIZE_M": 128,
|
|
|
137 |
"BLOCK_SIZE_K": 64,
|
138 |
"GROUP_SIZE_M": 4,
|
139 |
"num_warps": 8,
|
140 |
+
"num_stages": 2,
|
141 |
"waves_per_eu": 0,
|
142 |
"matrix_instr_nonkdim": 16,
|
143 |
"kpack": 2
|
|
|
148 |
"BLOCK_SIZE_K": 64,
|
149 |
"GROUP_SIZE_M": 1,
|
150 |
"num_warps": 8,
|
151 |
+
"num_stages": 2,
|
152 |
"waves_per_eu": 0,
|
153 |
+
"matrix_instr_nonkdim": 16,
|
154 |
"kpack": 2
|
155 |
},
|
156 |
"1536": {
|
|
|
159 |
"BLOCK_SIZE_K": 64,
|
160 |
"GROUP_SIZE_M": 1,
|
161 |
"num_warps": 8,
|
162 |
+
"num_stages": 2,
|
163 |
"waves_per_eu": 0,
|
164 |
"matrix_instr_nonkdim": 16,
|
165 |
"kpack": 2
|
|
|
170 |
"BLOCK_SIZE_K": 64,
|
171 |
"GROUP_SIZE_M": 1,
|
172 |
"num_warps": 8,
|
173 |
+
"num_stages": 2,
|
174 |
"waves_per_eu": 0,
|
175 |
"matrix_instr_nonkdim": 16,
|
176 |
"kpack": 2
|
|
|
181 |
"BLOCK_SIZE_K": 64,
|
182 |
"GROUP_SIZE_M": 1,
|
183 |
"num_warps": 8,
|
184 |
+
"num_stages": 2,
|
185 |
"waves_per_eu": 0,
|
186 |
"matrix_instr_nonkdim": 16,
|
187 |
+
"kpack": 2
|
188 |
},
|
189 |
"4096": {
|
190 |
"BLOCK_SIZE_M": 128,
|
|
|
192 |
"BLOCK_SIZE_K": 64,
|
193 |
"GROUP_SIZE_M": 1,
|
194 |
"num_warps": 8,
|
195 |
+
"num_stages": 2,
|
196 |
"waves_per_eu": 0,
|
197 |
"matrix_instr_nonkdim": 16,
|
198 |
+
"kpack": 2
|
199 |
}
|
200 |
}
|
build/torch25-cxx11-cu118-x86_64-linux/moe/configs/E=8,N=14336,device_name=AMD_Instinct_MI325X,dtype=fp8_w8a8.json
ADDED
@@ -0,0 +1,164 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"1": {
|
3 |
+
"BLOCK_SIZE_M": 16,
|
4 |
+
"BLOCK_SIZE_N": 16,
|
5 |
+
"BLOCK_SIZE_K": 256,
|
6 |
+
"GROUP_SIZE_M": 1,
|
7 |
+
"num_warps": 2,
|
8 |
+
"num_stages": 2,
|
9 |
+
"waves_per_eu": 0
|
10 |
+
},
|
11 |
+
"2": {
|
12 |
+
"BLOCK_SIZE_M": 16,
|
13 |
+
"BLOCK_SIZE_N": 32,
|
14 |
+
"BLOCK_SIZE_K": 256,
|
15 |
+
"GROUP_SIZE_M": 1,
|
16 |
+
"num_warps": 4,
|
17 |
+
"num_stages": 2,
|
18 |
+
"waves_per_eu": 0
|
19 |
+
},
|
20 |
+
"4": {
|
21 |
+
"BLOCK_SIZE_M": 16,
|
22 |
+
"BLOCK_SIZE_N": 16,
|
23 |
+
"BLOCK_SIZE_K": 256,
|
24 |
+
"GROUP_SIZE_M": 1,
|
25 |
+
"num_warps": 4,
|
26 |
+
"num_stages": 2,
|
27 |
+
"waves_per_eu": 0
|
28 |
+
},
|
29 |
+
"8": {
|
30 |
+
"BLOCK_SIZE_M": 16,
|
31 |
+
"BLOCK_SIZE_N": 64,
|
32 |
+
"BLOCK_SIZE_K": 256,
|
33 |
+
"GROUP_SIZE_M": 1,
|
34 |
+
"num_warps": 2,
|
35 |
+
"num_stages": 2,
|
36 |
+
"waves_per_eu": 0
|
37 |
+
},
|
38 |
+
"16": {
|
39 |
+
"BLOCK_SIZE_M": 64,
|
40 |
+
"BLOCK_SIZE_N": 64,
|
41 |
+
"BLOCK_SIZE_K": 256,
|
42 |
+
"GROUP_SIZE_M": 1,
|
43 |
+
"num_warps": 4,
|
44 |
+
"num_stages": 2,
|
45 |
+
"waves_per_eu": 0
|
46 |
+
},
|
47 |
+
"24": {
|
48 |
+
"BLOCK_SIZE_M": 32,
|
49 |
+
"BLOCK_SIZE_N": 64,
|
50 |
+
"BLOCK_SIZE_K": 256,
|
51 |
+
"GROUP_SIZE_M": 1,
|
52 |
+
"num_warps": 2,
|
53 |
+
"num_stages": 2,
|
54 |
+
"waves_per_eu": 0
|
55 |
+
},
|
56 |
+
"32": {
|
57 |
+
"BLOCK_SIZE_M": 16,
|
58 |
+
"BLOCK_SIZE_N": 32,
|
59 |
+
"BLOCK_SIZE_K": 256,
|
60 |
+
"GROUP_SIZE_M": 4,
|
61 |
+
"num_warps": 2,
|
62 |
+
"num_stages": 2,
|
63 |
+
"waves_per_eu": 0
|
64 |
+
},
|
65 |
+
"48": {
|
66 |
+
"BLOCK_SIZE_M": 16,
|
67 |
+
"BLOCK_SIZE_N": 64,
|
68 |
+
"BLOCK_SIZE_K": 256,
|
69 |
+
"GROUP_SIZE_M": 1,
|
70 |
+
"num_warps": 4,
|
71 |
+
"num_stages": 2,
|
72 |
+
"waves_per_eu": 0
|
73 |
+
},
|
74 |
+
"64": {
|
75 |
+
"BLOCK_SIZE_M": 32,
|
76 |
+
"BLOCK_SIZE_N": 16,
|
77 |
+
"BLOCK_SIZE_K": 128,
|
78 |
+
"GROUP_SIZE_M": 4,
|
79 |
+
"num_warps": 2,
|
80 |
+
"num_stages": 2,
|
81 |
+
"waves_per_eu": 0
|
82 |
+
},
|
83 |
+
"96": {
|
84 |
+
"BLOCK_SIZE_M": 32,
|
85 |
+
"BLOCK_SIZE_N": 64,
|
86 |
+
"BLOCK_SIZE_K": 256,
|
87 |
+
"GROUP_SIZE_M": 1,
|
88 |
+
"num_warps": 2,
|
89 |
+
"num_stages": 2,
|
90 |
+
"waves_per_eu": 0
|
91 |
+
},
|
92 |
+
"128": {
|
93 |
+
"BLOCK_SIZE_M": 64,
|
94 |
+
"BLOCK_SIZE_N": 64,
|
95 |
+
"BLOCK_SIZE_K": 256,
|
96 |
+
"GROUP_SIZE_M": 4,
|
97 |
+
"num_warps": 4,
|
98 |
+
"num_stages": 2,
|
99 |
+
"waves_per_eu": 0
|
100 |
+
},
|
101 |
+
"256": {
|
102 |
+
"BLOCK_SIZE_M": 128,
|
103 |
+
"BLOCK_SIZE_N": 128,
|
104 |
+
"BLOCK_SIZE_K": 256,
|
105 |
+
"GROUP_SIZE_M": 4,
|
106 |
+
"num_warps": 8,
|
107 |
+
"num_stages": 2,
|
108 |
+
"waves_per_eu": 0
|
109 |
+
},
|
110 |
+
"512": {
|
111 |
+
"BLOCK_SIZE_M": 256,
|
112 |
+
"BLOCK_SIZE_N": 128,
|
113 |
+
"BLOCK_SIZE_K": 128,
|
114 |
+
"GROUP_SIZE_M": 4,
|
115 |
+
"num_warps": 8,
|
116 |
+
"num_stages": 2,
|
117 |
+
"waves_per_eu": 0
|
118 |
+
},
|
119 |
+
"1024": {
|
120 |
+
"BLOCK_SIZE_M": 128,
|
121 |
+
"BLOCK_SIZE_N": 128,
|
122 |
+
"BLOCK_SIZE_K": 256,
|
123 |
+
"GROUP_SIZE_M": 1,
|
124 |
+
"num_warps": 8,
|
125 |
+
"num_stages": 2,
|
126 |
+
"waves_per_eu": 0
|
127 |
+
},
|
128 |
+
"1536": {
|
129 |
+
"BLOCK_SIZE_M": 256,
|
130 |
+
"BLOCK_SIZE_N": 256,
|
131 |
+
"BLOCK_SIZE_K": 64,
|
132 |
+
"GROUP_SIZE_M": 1,
|
133 |
+
"num_warps": 8,
|
134 |
+
"num_stages": 2,
|
135 |
+
"waves_per_eu": 0
|
136 |
+
},
|
137 |
+
"2048": {
|
138 |
+
"BLOCK_SIZE_M": 128,
|
139 |
+
"BLOCK_SIZE_N": 256,
|
140 |
+
"BLOCK_SIZE_K": 128,
|
141 |
+
"GROUP_SIZE_M": 1,
|
142 |
+
"num_warps": 8,
|
143 |
+
"num_stages": 2,
|
144 |
+
"waves_per_eu": 0
|
145 |
+
},
|
146 |
+
"3072": {
|
147 |
+
"BLOCK_SIZE_M": 128,
|
148 |
+
"BLOCK_SIZE_N": 256,
|
149 |
+
"BLOCK_SIZE_K": 128,
|
150 |
+
"GROUP_SIZE_M": 1,
|
151 |
+
"num_warps": 8,
|
152 |
+
"num_stages": 2,
|
153 |
+
"waves_per_eu": 0
|
154 |
+
},
|
155 |
+
"4096": {
|
156 |
+
"BLOCK_SIZE_M": 256,
|
157 |
+
"BLOCK_SIZE_N": 256,
|
158 |
+
"BLOCK_SIZE_K": 64,
|
159 |
+
"GROUP_SIZE_M": 1,
|
160 |
+
"num_warps": 8,
|
161 |
+
"num_stages": 2,
|
162 |
+
"waves_per_eu": 0
|
163 |
+
}
|
164 |
+
}
|
build/torch25-cxx11-cu118-x86_64-linux/moe/configs/E=8,N=14336,device_name=AMD_Instinct_MI325X.json
ADDED
@@ -0,0 +1,200 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"1": {
|
3 |
+
"BLOCK_SIZE_M": 16,
|
4 |
+
"BLOCK_SIZE_N": 16,
|
5 |
+
"BLOCK_SIZE_K": 256,
|
6 |
+
"GROUP_SIZE_M": 1,
|
7 |
+
"num_warps": 2,
|
8 |
+
"num_stages": 2,
|
9 |
+
"waves_per_eu": 0,
|
10 |
+
"matrix_instr_nonkdim": 16,
|
11 |
+
"kpack": 1
|
12 |
+
},
|
13 |
+
"2": {
|
14 |
+
"BLOCK_SIZE_M": 16,
|
15 |
+
"BLOCK_SIZE_N": 16,
|
16 |
+
"BLOCK_SIZE_K": 256,
|
17 |
+
"GROUP_SIZE_M": 1,
|
18 |
+
"num_warps": 4,
|
19 |
+
"num_stages": 2,
|
20 |
+
"waves_per_eu": 0,
|
21 |
+
"matrix_instr_nonkdim": 16,
|
22 |
+
"kpack": 2
|
23 |
+
},
|
24 |
+
"4": {
|
25 |
+
"BLOCK_SIZE_M": 16,
|
26 |
+
"BLOCK_SIZE_N": 16,
|
27 |
+
"BLOCK_SIZE_K": 128,
|
28 |
+
"GROUP_SIZE_M": 1,
|
29 |
+
"num_warps": 1,
|
30 |
+
"num_stages": 2,
|
31 |
+
"waves_per_eu": 0,
|
32 |
+
"matrix_instr_nonkdim": 16,
|
33 |
+
"kpack": 2
|
34 |
+
},
|
35 |
+
"8": {
|
36 |
+
"BLOCK_SIZE_M": 16,
|
37 |
+
"BLOCK_SIZE_N": 32,
|
38 |
+
"BLOCK_SIZE_K": 128,
|
39 |
+
"GROUP_SIZE_M": 1,
|
40 |
+
"num_warps": 2,
|
41 |
+
"num_stages": 2,
|
42 |
+
"waves_per_eu": 0,
|
43 |
+
"matrix_instr_nonkdim": 16,
|
44 |
+
"kpack": 2
|
45 |
+
},
|
46 |
+
"16": {
|
47 |
+
"BLOCK_SIZE_M": 16,
|
48 |
+
"BLOCK_SIZE_N": 64,
|
49 |
+
"BLOCK_SIZE_K": 64,
|
50 |
+
"GROUP_SIZE_M": 1,
|
51 |
+
"num_warps": 2,
|
52 |
+
"num_stages": 2,
|
53 |
+
"waves_per_eu": 0,
|
54 |
+
"matrix_instr_nonkdim": 16,
|
55 |
+
"kpack": 2
|
56 |
+
},
|
57 |
+
"24": {
|
58 |
+
"BLOCK_SIZE_M": 16,
|
59 |
+
"BLOCK_SIZE_N": 128,
|
60 |
+
"BLOCK_SIZE_K": 64,
|
61 |
+
"GROUP_SIZE_M": 1,
|
62 |
+
"num_warps": 4,
|
63 |
+
"num_stages": 2,
|
64 |
+
"waves_per_eu": 0,
|
65 |
+
"matrix_instr_nonkdim": 16,
|
66 |
+
"kpack": 2
|
67 |
+
},
|
68 |
+
"32": {
|
69 |
+
"BLOCK_SIZE_M": 16,
|
70 |
+
"BLOCK_SIZE_N": 64,
|
71 |
+
"BLOCK_SIZE_K": 128,
|
72 |
+
"GROUP_SIZE_M": 4,
|
73 |
+
"num_warps": 4,
|
74 |
+
"num_stages": 2,
|
75 |
+
"waves_per_eu": 0,
|
76 |
+
"matrix_instr_nonkdim": 16,
|
77 |
+
"kpack": 2
|
78 |
+
},
|
79 |
+
"48": {
|
80 |
+
"BLOCK_SIZE_M": 16,
|
81 |
+
"BLOCK_SIZE_N": 64,
|
82 |
+
"BLOCK_SIZE_K": 128,
|
83 |
+
"GROUP_SIZE_M": 4,
|
84 |
+
"num_warps": 1,
|
85 |
+
"num_stages": 2,
|
86 |
+
"waves_per_eu": 0,
|
87 |
+
"matrix_instr_nonkdim": 16,
|
88 |
+
"kpack": 1
|
89 |
+
},
|
90 |
+
"64": {
|
91 |
+
"BLOCK_SIZE_M": 32,
|
92 |
+
"BLOCK_SIZE_N": 64,
|
93 |
+
"BLOCK_SIZE_K": 128,
|
94 |
+
"GROUP_SIZE_M": 4,
|
95 |
+
"num_warps": 8,
|
96 |
+
"num_stages": 2,
|
97 |
+
"waves_per_eu": 0,
|
98 |
+
"matrix_instr_nonkdim": 16,
|
99 |
+
"kpack": 2
|
100 |
+
},
|
101 |
+
"96": {
|
102 |
+
"BLOCK_SIZE_M": 32,
|
103 |
+
"BLOCK_SIZE_N": 64,
|
104 |
+
"BLOCK_SIZE_K": 256,
|
105 |
+
"GROUP_SIZE_M": 4,
|
106 |
+
"num_warps": 8,
|
107 |
+
"num_stages": 2,
|
108 |
+
"waves_per_eu": 0,
|
109 |
+
"matrix_instr_nonkdim": 16,
|
110 |
+
"kpack": 1
|
111 |
+
},
|
112 |
+
"128": {
|
113 |
+
"BLOCK_SIZE_M": 64,
|
114 |
+
"BLOCK_SIZE_N": 64,
|
115 |
+
"BLOCK_SIZE_K": 128,
|
116 |
+
"GROUP_SIZE_M": 4,
|
117 |
+
"num_warps": 4,
|
118 |
+
"num_stages": 2,
|
119 |
+
"waves_per_eu": 0,
|
120 |
+
"matrix_instr_nonkdim": 32,
|
121 |
+
"kpack": 2
|
122 |
+
},
|
123 |
+
"256": {
|
124 |
+
"BLOCK_SIZE_M": 128,
|
125 |
+
"BLOCK_SIZE_N": 128,
|
126 |
+
"BLOCK_SIZE_K": 64,
|
127 |
+
"GROUP_SIZE_M": 4,
|
128 |
+
"num_warps": 8,
|
129 |
+
"num_stages": 2,
|
130 |
+
"waves_per_eu": 0,
|
131 |
+
"matrix_instr_nonkdim": 16,
|
132 |
+
"kpack": 2
|
133 |
+
},
|
134 |
+
"512": {
|
135 |
+
"BLOCK_SIZE_M": 128,
|
136 |
+
"BLOCK_SIZE_N": 128,
|
137 |
+
"BLOCK_SIZE_K": 64,
|
138 |
+
"GROUP_SIZE_M": 1,
|
139 |
+
"num_warps": 8,
|
140 |
+
"num_stages": 2,
|
141 |
+
"waves_per_eu": 0,
|
142 |
+
"matrix_instr_nonkdim": 16,
|
143 |
+
"kpack": 2
|
144 |
+
},
|
145 |
+
"1024": {
|
146 |
+
"BLOCK_SIZE_M": 128,
|
147 |
+
"BLOCK_SIZE_N": 128,
|
148 |
+
"BLOCK_SIZE_K": 64,
|
149 |
+
"GROUP_SIZE_M": 1,
|
150 |
+
"num_warps": 8,
|
151 |
+
"num_stages": 2,
|
152 |
+
"waves_per_eu": 0,
|
153 |
+
"matrix_instr_nonkdim": 16,
|
154 |
+
"kpack": 2
|
155 |
+
},
|
156 |
+
"1536": {
|
157 |
+
"BLOCK_SIZE_M": 128,
|
158 |
+
"BLOCK_SIZE_N": 128,
|
159 |
+
"BLOCK_SIZE_K": 64,
|
160 |
+
"GROUP_SIZE_M": 1,
|
161 |
+
"num_warps": 8,
|
162 |
+
"num_stages": 2,
|
163 |
+
"waves_per_eu": 0,
|
164 |
+
"matrix_instr_nonkdim": 16,
|
165 |
+
"kpack": 2
|
166 |
+
},
|
167 |
+
"2048": {
|
168 |
+
"BLOCK_SIZE_M": 128,
|
169 |
+
"BLOCK_SIZE_N": 128,
|
170 |
+
"BLOCK_SIZE_K": 64,
|
171 |
+
"GROUP_SIZE_M": 1,
|
172 |
+
"num_warps": 8,
|
173 |
+
"num_stages": 2,
|
174 |
+
"waves_per_eu": 0,
|
175 |
+
"matrix_instr_nonkdim": 16,
|
176 |
+
"kpack": 2
|
177 |
+
},
|
178 |
+
"3072": {
|
179 |
+
"BLOCK_SIZE_M": 128,
|
180 |
+
"BLOCK_SIZE_N": 128,
|
181 |
+
"BLOCK_SIZE_K": 64,
|
182 |
+
"GROUP_SIZE_M": 1,
|
183 |
+
"num_warps": 8,
|
184 |
+
"num_stages": 2,
|
185 |
+
"waves_per_eu": 0,
|
186 |
+
"matrix_instr_nonkdim": 16,
|
187 |
+
"kpack": 2
|
188 |
+
},
|
189 |
+
"4096": {
|
190 |
+
"BLOCK_SIZE_M": 128,
|
191 |
+
"BLOCK_SIZE_N": 128,
|
192 |
+
"BLOCK_SIZE_K": 64,
|
193 |
+
"GROUP_SIZE_M": 1,
|
194 |
+
"num_warps": 8,
|
195 |
+
"num_stages": 2,
|
196 |
+
"waves_per_eu": 0,
|
197 |
+
"matrix_instr_nonkdim": 16,
|
198 |
+
"kpack": 2
|
199 |
+
}
|
200 |
+
}
|
build/torch25-cxx11-cu118-x86_64-linux/moe/configs/E=8,N=14336,device_name=NVIDIA_H200,dtype=fp8_w8a8.json
ADDED
@@ -0,0 +1,146 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"1": {
|
3 |
+
"BLOCK_SIZE_M": 64,
|
4 |
+
"BLOCK_SIZE_N": 128,
|
5 |
+
"BLOCK_SIZE_K": 256,
|
6 |
+
"GROUP_SIZE_M": 64,
|
7 |
+
"num_warps": 4,
|
8 |
+
"num_stages": 4
|
9 |
+
},
|
10 |
+
"2": {
|
11 |
+
"BLOCK_SIZE_M": 64,
|
12 |
+
"BLOCK_SIZE_N": 128,
|
13 |
+
"BLOCK_SIZE_K": 256,
|
14 |
+
"GROUP_SIZE_M": 64,
|
15 |
+
"num_warps": 8,
|
16 |
+
"num_stages": 4
|
17 |
+
},
|
18 |
+
"4": {
|
19 |
+
"BLOCK_SIZE_M": 64,
|
20 |
+
"BLOCK_SIZE_N": 256,
|
21 |
+
"BLOCK_SIZE_K": 128,
|
22 |
+
"GROUP_SIZE_M": 32,
|
23 |
+
"num_warps": 4,
|
24 |
+
"num_stages": 5
|
25 |
+
},
|
26 |
+
"8": {
|
27 |
+
"BLOCK_SIZE_M": 64,
|
28 |
+
"BLOCK_SIZE_N": 128,
|
29 |
+
"BLOCK_SIZE_K": 256,
|
30 |
+
"GROUP_SIZE_M": 64,
|
31 |
+
"num_warps": 4,
|
32 |
+
"num_stages": 4
|
33 |
+
},
|
34 |
+
"16": {
|
35 |
+
"BLOCK_SIZE_M": 64,
|
36 |
+
"BLOCK_SIZE_N": 128,
|
37 |
+
"BLOCK_SIZE_K": 128,
|
38 |
+
"GROUP_SIZE_M": 16,
|
39 |
+
"num_warps": 4,
|
40 |
+
"num_stages": 3
|
41 |
+
},
|
42 |
+
"24": {
|
43 |
+
"BLOCK_SIZE_M": 64,
|
44 |
+
"BLOCK_SIZE_N": 128,
|
45 |
+
"BLOCK_SIZE_K": 128,
|
46 |
+
"GROUP_SIZE_M": 64,
|
47 |
+
"num_warps": 4,
|
48 |
+
"num_stages": 4
|
49 |
+
},
|
50 |
+
"32": {
|
51 |
+
"BLOCK_SIZE_M": 64,
|
52 |
+
"BLOCK_SIZE_N": 256,
|
53 |
+
"BLOCK_SIZE_K": 128,
|
54 |
+
"GROUP_SIZE_M": 32,
|
55 |
+
"num_warps": 4,
|
56 |
+
"num_stages": 5
|
57 |
+
},
|
58 |
+
"48": {
|
59 |
+
"BLOCK_SIZE_M": 64,
|
60 |
+
"BLOCK_SIZE_N": 64,
|
61 |
+
"BLOCK_SIZE_K": 128,
|
62 |
+
"GROUP_SIZE_M": 1,
|
63 |
+
"num_warps": 4,
|
64 |
+
"num_stages": 3
|
65 |
+
},
|
66 |
+
"64": {
|
67 |
+
"BLOCK_SIZE_M": 64,
|
68 |
+
"BLOCK_SIZE_N": 128,
|
69 |
+
"BLOCK_SIZE_K": 128,
|
70 |
+
"GROUP_SIZE_M": 1,
|
71 |
+
"num_warps": 4,
|
72 |
+
"num_stages": 3
|
73 |
+
},
|
74 |
+
"96": {
|
75 |
+
"BLOCK_SIZE_M": 64,
|
76 |
+
"BLOCK_SIZE_N": 64,
|
77 |
+
"BLOCK_SIZE_K": 128,
|
78 |
+
"GROUP_SIZE_M": 1,
|
79 |
+
"num_warps": 4,
|
80 |
+
"num_stages": 3
|
81 |
+
},
|
82 |
+
"128": {
|
83 |
+
"BLOCK_SIZE_M": 64,
|
84 |
+
"BLOCK_SIZE_N": 128,
|
85 |
+
"BLOCK_SIZE_K": 128,
|
86 |
+
"GROUP_SIZE_M": 1,
|
87 |
+
"num_warps": 4,
|
88 |
+
"num_stages": 3
|
89 |
+
},
|
90 |
+
"256": {
|
91 |
+
"BLOCK_SIZE_M": 128,
|
92 |
+
"BLOCK_SIZE_N": 256,
|
93 |
+
"BLOCK_SIZE_K": 128,
|
94 |
+
"GROUP_SIZE_M": 32,
|
95 |
+
"num_warps": 8,
|
96 |
+
"num_stages": 4
|
97 |
+
},
|
98 |
+
"512": {
|
99 |
+
"BLOCK_SIZE_M": 128,
|
100 |
+
"BLOCK_SIZE_N": 256,
|
101 |
+
"BLOCK_SIZE_K": 128,
|
102 |
+
"GROUP_SIZE_M": 32,
|
103 |
+
"num_warps": 8,
|
104 |
+
"num_stages": 4
|
105 |
+
},
|
106 |
+
"1024": {
|
107 |
+
"BLOCK_SIZE_M": 128,
|
108 |
+
"BLOCK_SIZE_N": 256,
|
109 |
+
"BLOCK_SIZE_K": 128,
|
110 |
+
"GROUP_SIZE_M": 16,
|
111 |
+
"num_warps": 8,
|
112 |
+
"num_stages": 4
|
113 |
+
},
|
114 |
+
"1536": {
|
115 |
+
"BLOCK_SIZE_M": 128,
|
116 |
+
"BLOCK_SIZE_N": 256,
|
117 |
+
"BLOCK_SIZE_K": 128,
|
118 |
+
"GROUP_SIZE_M": 16,
|
119 |
+
"num_warps": 8,
|
120 |
+
"num_stages": 4
|
121 |
+
},
|
122 |
+
"2048": {
|
123 |
+
"BLOCK_SIZE_M": 128,
|
124 |
+
"BLOCK_SIZE_N": 256,
|
125 |
+
"BLOCK_SIZE_K": 128,
|
126 |
+
"GROUP_SIZE_M": 32,
|
127 |
+
"num_warps": 8,
|
128 |
+
"num_stages": 3
|
129 |
+
},
|
130 |
+
"3072": {
|
131 |
+
"BLOCK_SIZE_M": 128,
|
132 |
+
"BLOCK_SIZE_N": 256,
|
133 |
+
"BLOCK_SIZE_K": 128,
|
134 |
+
"GROUP_SIZE_M": 32,
|
135 |
+
"num_warps": 8,
|
136 |
+
"num_stages": 3
|
137 |
+
},
|
138 |
+
"4096": {
|
139 |
+
"BLOCK_SIZE_M": 128,
|
140 |
+
"BLOCK_SIZE_N": 256,
|
141 |
+
"BLOCK_SIZE_K": 128,
|
142 |
+
"GROUP_SIZE_M": 16,
|
143 |
+
"num_warps": 8,
|
144 |
+
"num_stages": 3
|
145 |
+
}
|
146 |
+
}
|
build/torch25-cxx11-cu118-x86_64-linux/moe/configs/E=8,N=14336,device_name=NVIDIA_H200.json
ADDED
@@ -0,0 +1,146 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"1": {
|
3 |
+
"BLOCK_SIZE_M": 16,
|
4 |
+
"BLOCK_SIZE_N": 128,
|
5 |
+
"BLOCK_SIZE_K": 128,
|
6 |
+
"GROUP_SIZE_M": 16,
|
7 |
+
"num_warps": 4,
|
8 |
+
"num_stages": 5
|
9 |
+
},
|
10 |
+
"2": {
|
11 |
+
"BLOCK_SIZE_M": 32,
|
12 |
+
"BLOCK_SIZE_N": 64,
|
13 |
+
"BLOCK_SIZE_K": 128,
|
14 |
+
"GROUP_SIZE_M": 1,
|
15 |
+
"num_warps": 4,
|
16 |
+
"num_stages": 5
|
17 |
+
},
|
18 |
+
"4": {
|
19 |
+
"BLOCK_SIZE_M": 16,
|
20 |
+
"BLOCK_SIZE_N": 64,
|
21 |
+
"BLOCK_SIZE_K": 64,
|
22 |
+
"GROUP_SIZE_M": 1,
|
23 |
+
"num_warps": 4,
|
24 |
+
"num_stages": 5
|
25 |
+
},
|
26 |
+
"8": {
|
27 |
+
"BLOCK_SIZE_M": 16,
|
28 |
+
"BLOCK_SIZE_N": 128,
|
29 |
+
"BLOCK_SIZE_K": 128,
|
30 |
+
"GROUP_SIZE_M": 1,
|
31 |
+
"num_warps": 4,
|
32 |
+
"num_stages": 5
|
33 |
+
},
|
34 |
+
"16": {
|
35 |
+
"BLOCK_SIZE_M": 16,
|
36 |
+
"BLOCK_SIZE_N": 64,
|
37 |
+
"BLOCK_SIZE_K": 256,
|
38 |
+
"GROUP_SIZE_M": 16,
|
39 |
+
"num_warps": 4,
|
40 |
+
"num_stages": 3
|
41 |
+
},
|
42 |
+
"24": {
|
43 |
+
"BLOCK_SIZE_M": 16,
|
44 |
+
"BLOCK_SIZE_N": 128,
|
45 |
+
"BLOCK_SIZE_K": 256,
|
46 |
+
"GROUP_SIZE_M": 16,
|
47 |
+
"num_warps": 4,
|
48 |
+
"num_stages": 2
|
49 |
+
},
|
50 |
+
"32": {
|
51 |
+
"BLOCK_SIZE_M": 16,
|
52 |
+
"BLOCK_SIZE_N": 128,
|
53 |
+
"BLOCK_SIZE_K": 128,
|
54 |
+
"GROUP_SIZE_M": 16,
|
55 |
+
"num_warps": 8,
|
56 |
+
"num_stages": 3
|
57 |
+
},
|
58 |
+
"48": {
|
59 |
+
"BLOCK_SIZE_M": 32,
|
60 |
+
"BLOCK_SIZE_N": 128,
|
61 |
+
"BLOCK_SIZE_K": 128,
|
62 |
+
"GROUP_SIZE_M": 16,
|
63 |
+
"num_warps": 4,
|
64 |
+
"num_stages": 3
|
65 |
+
},
|
66 |
+
"64": {
|
67 |
+
"BLOCK_SIZE_M": 32,
|
68 |
+
"BLOCK_SIZE_N": 128,
|
69 |
+
"BLOCK_SIZE_K": 128,
|
70 |
+
"GROUP_SIZE_M": 64,
|
71 |
+
"num_warps": 4,
|
72 |
+
"num_stages": 3
|
73 |
+
},
|
74 |
+
"96": {
|
75 |
+
"BLOCK_SIZE_M": 32,
|
76 |
+
"BLOCK_SIZE_N": 128,
|
77 |
+
"BLOCK_SIZE_K": 128,
|
78 |
+
"GROUP_SIZE_M": 64,
|
79 |
+
"num_warps": 4,
|
80 |
+
"num_stages": 3
|
81 |
+
},
|
82 |
+
"128": {
|
83 |
+
"BLOCK_SIZE_M": 64,
|
84 |
+
"BLOCK_SIZE_N": 128,
|
85 |
+
"BLOCK_SIZE_K": 128,
|
86 |
+
"GROUP_SIZE_M": 1,
|
87 |
+
"num_warps": 8,
|
88 |
+
"num_stages": 4
|
89 |
+
},
|
90 |
+
"256": {
|
91 |
+
"BLOCK_SIZE_M": 128,
|
92 |
+
"BLOCK_SIZE_N": 256,
|
93 |
+
"BLOCK_SIZE_K": 64,
|
94 |
+
"GROUP_SIZE_M": 1,
|
95 |
+
"num_warps": 8,
|
96 |
+
"num_stages": 4
|
97 |
+
},
|
98 |
+
"512": {
|
99 |
+
"BLOCK_SIZE_M": 128,
|
100 |
+
"BLOCK_SIZE_N": 256,
|
101 |
+
"BLOCK_SIZE_K": 64,
|
102 |
+
"GROUP_SIZE_M": 16,
|
103 |
+
"num_warps": 8,
|
104 |
+
"num_stages": 4
|
105 |
+
},
|
106 |
+
"1024": {
|
107 |
+
"BLOCK_SIZE_M": 128,
|
108 |
+
"BLOCK_SIZE_N": 256,
|
109 |
+
"BLOCK_SIZE_K": 64,
|
110 |
+
"GROUP_SIZE_M": 16,
|
111 |
+
"num_warps": 8,
|
112 |
+
"num_stages": 4
|
113 |
+
},
|
114 |
+
"1536": {
|
115 |
+
"BLOCK_SIZE_M": 128,
|
116 |
+
"BLOCK_SIZE_N": 256,
|
117 |
+
"BLOCK_SIZE_K": 64,
|
118 |
+
"GROUP_SIZE_M": 16,
|
119 |
+
"num_warps": 8,
|
120 |
+
"num_stages": 4
|
121 |
+
},
|
122 |
+
"2048": {
|
123 |
+
"BLOCK_SIZE_M": 128,
|
124 |
+
"BLOCK_SIZE_N": 256,
|
125 |
+
"BLOCK_SIZE_K": 64,
|
126 |
+
"GROUP_SIZE_M": 32,
|
127 |
+
"num_warps": 8,
|
128 |
+
"num_stages": 4
|
129 |
+
},
|
130 |
+
"3072": {
|
131 |
+
"BLOCK_SIZE_M": 128,
|
132 |
+
"BLOCK_SIZE_N": 256,
|
133 |
+
"BLOCK_SIZE_K": 64,
|
134 |
+
"GROUP_SIZE_M": 16,
|
135 |
+
"num_warps": 8,
|
136 |
+
"num_stages": 4
|
137 |
+
},
|
138 |
+
"4096": {
|
139 |
+
"BLOCK_SIZE_M": 128,
|
140 |
+
"BLOCK_SIZE_N": 256,
|
141 |
+
"BLOCK_SIZE_K": 64,
|
142 |
+
"GROUP_SIZE_M": 16,
|
143 |
+
"num_warps": 8,
|
144 |
+
"num_stages": 4
|
145 |
+
}
|
146 |
+
}
|
build/torch25-cxx11-cu118-x86_64-linux/moe/configs/E=8,N=16384,device_name=AMD_Instinct_MI300X,dtype=fp8_w8a8.json
ADDED
@@ -0,0 +1,164 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"1": {
|
3 |
+
"BLOCK_SIZE_M": 16,
|
4 |
+
"BLOCK_SIZE_N": 16,
|
5 |
+
"BLOCK_SIZE_K": 256,
|
6 |
+
"GROUP_SIZE_M": 1,
|
7 |
+
"num_warps": 2,
|
8 |
+
"num_stages": 2,
|
9 |
+
"waves_per_eu": 0
|
10 |
+
},
|
11 |
+
"2": {
|
12 |
+
"BLOCK_SIZE_M": 16,
|
13 |
+
"BLOCK_SIZE_N": 32,
|
14 |
+
"BLOCK_SIZE_K": 256,
|
15 |
+
"GROUP_SIZE_M": 1,
|
16 |
+
"num_warps": 2,
|
17 |
+
"num_stages": 2,
|
18 |
+
"waves_per_eu": 0
|
19 |
+
},
|
20 |
+
"4": {
|
21 |
+
"BLOCK_SIZE_M": 16,
|
22 |
+
"BLOCK_SIZE_N": 64,
|
23 |
+
"BLOCK_SIZE_K": 256,
|
24 |
+
"GROUP_SIZE_M": 1,
|
25 |
+
"num_warps": 2,
|
26 |
+
"num_stages": 2,
|
27 |
+
"waves_per_eu": 0
|
28 |
+
},
|
29 |
+
"8": {
|
30 |
+
"BLOCK_SIZE_M": 16,
|
31 |
+
"BLOCK_SIZE_N": 64,
|
32 |
+
"BLOCK_SIZE_K": 256,
|
33 |
+
"GROUP_SIZE_M": 1,
|
34 |
+
"num_warps": 2,
|
35 |
+
"num_stages": 2,
|
36 |
+
"waves_per_eu": 0
|
37 |
+
},
|
38 |
+
"16": {
|
39 |
+
"BLOCK_SIZE_M": 16,
|
40 |
+
"BLOCK_SIZE_N": 64,
|
41 |
+
"BLOCK_SIZE_K": 256,
|
42 |
+
"GROUP_SIZE_M": 1,
|
43 |
+
"num_warps": 2,
|
44 |
+
"num_stages": 2,
|
45 |
+
"waves_per_eu": 0
|
46 |
+
},
|
47 |
+
"24": {
|
48 |
+
"BLOCK_SIZE_M": 16,
|
49 |
+
"BLOCK_SIZE_N": 64,
|
50 |
+
"BLOCK_SIZE_K": 256,
|
51 |
+
"GROUP_SIZE_M": 1,
|
52 |
+
"num_warps": 2,
|
53 |
+
"num_stages": 2,
|
54 |
+
"waves_per_eu": 0
|
55 |
+
},
|
56 |
+
"32": {
|
57 |
+
"BLOCK_SIZE_M": 16,
|
58 |
+
"BLOCK_SIZE_N": 64,
|
59 |
+
"BLOCK_SIZE_K": 256,
|
60 |
+
"GROUP_SIZE_M": 4,
|
61 |
+
"num_warps": 2,
|
62 |
+
"num_stages": 2,
|
63 |
+
"waves_per_eu": 0
|
64 |
+
},
|
65 |
+
"48": {
|
66 |
+
"BLOCK_SIZE_M": 16,
|
67 |
+
"BLOCK_SIZE_N": 64,
|
68 |
+
"BLOCK_SIZE_K": 256,
|
69 |
+
"GROUP_SIZE_M": 1,
|
70 |
+
"num_warps": 4,
|
71 |
+
"num_stages": 2,
|
72 |
+
"waves_per_eu": 0
|
73 |
+
},
|
74 |
+
"64": {
|
75 |
+
"BLOCK_SIZE_M": 32,
|
76 |
+
"BLOCK_SIZE_N": 64,
|
77 |
+
"BLOCK_SIZE_K": 128,
|
78 |
+
"GROUP_SIZE_M": 4,
|
79 |
+
"num_warps": 2,
|
80 |
+
"num_stages": 2,
|
81 |
+
"waves_per_eu": 0
|
82 |
+
},
|
83 |
+
"96": {
|
84 |
+
"BLOCK_SIZE_M": 32,
|
85 |
+
"BLOCK_SIZE_N": 64,
|
86 |
+
"BLOCK_SIZE_K": 128,
|
87 |
+
"GROUP_SIZE_M": 1,
|
88 |
+
"num_warps": 2,
|
89 |
+
"num_stages": 2,
|
90 |
+
"waves_per_eu": 0
|
91 |
+
},
|
92 |
+
"128": {
|
93 |
+
"BLOCK_SIZE_M": 64,
|
94 |
+
"BLOCK_SIZE_N": 64,
|
95 |
+
"BLOCK_SIZE_K": 128,
|
96 |
+
"GROUP_SIZE_M": 4,
|
97 |
+
"num_warps": 4,
|
98 |
+
"num_stages": 2,
|
99 |
+
"waves_per_eu": 0
|
100 |
+
},
|
101 |
+
"256": {
|
102 |
+
"BLOCK_SIZE_M": 64,
|
103 |
+
"BLOCK_SIZE_N": 128,
|
104 |
+
"BLOCK_SIZE_K": 128,
|
105 |
+
"GROUP_SIZE_M": 1,
|
106 |
+
"num_warps": 8,
|
107 |
+
"num_stages": 2,
|
108 |
+
"waves_per_eu": 0
|
109 |
+
},
|
110 |
+
"512": {
|
111 |
+
"BLOCK_SIZE_M": 128,
|
112 |
+
"BLOCK_SIZE_N": 128,
|
113 |
+
"BLOCK_SIZE_K": 256,
|
114 |
+
"GROUP_SIZE_M": 1,
|
115 |
+
"num_warps": 8,
|
116 |
+
"num_stages": 2,
|
117 |
+
"waves_per_eu": 0
|
118 |
+
},
|
119 |
+
"1024": {
|
120 |
+
"BLOCK_SIZE_M": 128,
|
121 |
+
"BLOCK_SIZE_N": 256,
|
122 |
+
"BLOCK_SIZE_K": 128,
|
123 |
+
"GROUP_SIZE_M": 1,
|
124 |
+
"num_warps": 8,
|
125 |
+
"num_stages": 2,
|
126 |
+
"waves_per_eu": 0
|
127 |
+
},
|
128 |
+
"1536": {
|
129 |
+
"BLOCK_SIZE_M": 128,
|
130 |
+
"BLOCK_SIZE_N": 256,
|
131 |
+
"BLOCK_SIZE_K": 128,
|
132 |
+
"GROUP_SIZE_M": 1,
|
133 |
+
"num_warps": 8,
|
134 |
+
"num_stages": 2,
|
135 |
+
"waves_per_eu": 0
|
136 |
+
},
|
137 |
+
"2048": {
|
138 |
+
"BLOCK_SIZE_M": 128,
|
139 |
+
"BLOCK_SIZE_N": 256,
|
140 |
+
"BLOCK_SIZE_K": 128,
|
141 |
+
"GROUP_SIZE_M": 1,
|
142 |
+
"num_warps": 8,
|
143 |
+
"num_stages": 2,
|
144 |
+
"waves_per_eu": 0
|
145 |
+
},
|
146 |
+
"3072": {
|
147 |
+
"BLOCK_SIZE_M": 128,
|
148 |
+
"BLOCK_SIZE_N": 256,
|
149 |
+
"BLOCK_SIZE_K": 128,
|
150 |
+
"GROUP_SIZE_M": 1,
|
151 |
+
"num_warps": 8,
|
152 |
+
"num_stages": 2,
|
153 |
+
"waves_per_eu": 0
|
154 |
+
},
|
155 |
+
"4096": {
|
156 |
+
"BLOCK_SIZE_M": 256,
|
157 |
+
"BLOCK_SIZE_N": 256,
|
158 |
+
"BLOCK_SIZE_K": 64,
|
159 |
+
"GROUP_SIZE_M": 1,
|
160 |
+
"num_warps": 8,
|
161 |
+
"num_stages": 2,
|
162 |
+
"waves_per_eu": 0
|
163 |
+
}
|
164 |
+
}
|
build/torch25-cxx11-cu118-x86_64-linux/moe/configs/E=8,N=16384,device_name=AMD_Instinct_MI300X.json
ADDED
@@ -0,0 +1,200 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"1": {
|
3 |
+
"BLOCK_SIZE_M": 16,
|
4 |
+
"BLOCK_SIZE_N": 16,
|
5 |
+
"BLOCK_SIZE_K": 256,
|
6 |
+
"GROUP_SIZE_M": 1,
|
7 |
+
"num_warps": 2,
|
8 |
+
"num_stages": 2,
|
9 |
+
"waves_per_eu": 0,
|
10 |
+
"matrix_instr_nonkdim": 16,
|
11 |
+
"kpack": 2
|
12 |
+
},
|
13 |
+
"2": {
|
14 |
+
"BLOCK_SIZE_M": 16,
|
15 |
+
"BLOCK_SIZE_N": 64,
|
16 |
+
"BLOCK_SIZE_K": 128,
|
17 |
+
"GROUP_SIZE_M": 1,
|
18 |
+
"num_warps": 2,
|
19 |
+
"num_stages": 2,
|
20 |
+
"waves_per_eu": 0,
|
21 |
+
"matrix_instr_nonkdim": 16,
|
22 |
+
"kpack": 2
|
23 |
+
},
|
24 |
+
"4": {
|
25 |
+
"BLOCK_SIZE_M": 16,
|
26 |
+
"BLOCK_SIZE_N": 64,
|
27 |
+
"BLOCK_SIZE_K": 64,
|
28 |
+
"GROUP_SIZE_M": 1,
|
29 |
+
"num_warps": 2,
|
30 |
+
"num_stages": 2,
|
31 |
+
"waves_per_eu": 0,
|
32 |
+
"matrix_instr_nonkdim": 16,
|
33 |
+
"kpack": 2
|
34 |
+
},
|
35 |
+
"8": {
|
36 |
+
"BLOCK_SIZE_M": 16,
|
37 |
+
"BLOCK_SIZE_N": 32,
|
38 |
+
"BLOCK_SIZE_K": 64,
|
39 |
+
"GROUP_SIZE_M": 1,
|
40 |
+
"num_warps": 2,
|
41 |
+
"num_stages": 2,
|
42 |
+
"waves_per_eu": 0,
|
43 |
+
"matrix_instr_nonkdim": 16,
|
44 |
+
"kpack": 2
|
45 |
+
},
|
46 |
+
"16": {
|
47 |
+
"BLOCK_SIZE_M": 16,
|
48 |
+
"BLOCK_SIZE_N": 16,
|
49 |
+
"BLOCK_SIZE_K": 256,
|
50 |
+
"GROUP_SIZE_M": 1,
|
51 |
+
"num_warps": 1,
|
52 |
+
"num_stages": 2,
|
53 |
+
"waves_per_eu": 0,
|
54 |
+
"matrix_instr_nonkdim": 16,
|
55 |
+
"kpack": 2
|
56 |
+
},
|
57 |
+
"24": {
|
58 |
+
"BLOCK_SIZE_M": 16,
|
59 |
+
"BLOCK_SIZE_N": 64,
|
60 |
+
"BLOCK_SIZE_K": 128,
|
61 |
+
"GROUP_SIZE_M": 1,
|
62 |
+
"num_warps": 2,
|
63 |
+
"num_stages": 2,
|
64 |
+
"waves_per_eu": 0,
|
65 |
+
"matrix_instr_nonkdim": 16,
|
66 |
+
"kpack": 2
|
67 |
+
},
|
68 |
+
"32": {
|
69 |
+
"BLOCK_SIZE_M": 16,
|
70 |
+
"BLOCK_SIZE_N": 64,
|
71 |
+
"BLOCK_SIZE_K": 128,
|
72 |
+
"GROUP_SIZE_M": 4,
|
73 |
+
"num_warps": 2,
|
74 |
+
"num_stages": 2,
|
75 |
+
"waves_per_eu": 0,
|
76 |
+
"matrix_instr_nonkdim": 16,
|
77 |
+
"kpack": 2
|
78 |
+
},
|
79 |
+
"48": {
|
80 |
+
"BLOCK_SIZE_M": 32,
|
81 |
+
"BLOCK_SIZE_N": 32,
|
82 |
+
"BLOCK_SIZE_K": 256,
|
83 |
+
"GROUP_SIZE_M": 1,
|
84 |
+
"num_warps": 4,
|
85 |
+
"num_stages": 2,
|
86 |
+
"waves_per_eu": 0,
|
87 |
+
"matrix_instr_nonkdim": 16,
|
88 |
+
"kpack": 2
|
89 |
+
},
|
90 |
+
"64": {
|
91 |
+
"BLOCK_SIZE_M": 32,
|
92 |
+
"BLOCK_SIZE_N": 32,
|
93 |
+
"BLOCK_SIZE_K": 256,
|
94 |
+
"GROUP_SIZE_M": 4,
|
95 |
+
"num_warps": 4,
|
96 |
+
"num_stages": 2,
|
97 |
+
"waves_per_eu": 0,
|
98 |
+
"matrix_instr_nonkdim": 16,
|
99 |
+
"kpack": 2
|
100 |
+
},
|
101 |
+
"96": {
|
102 |
+
"BLOCK_SIZE_M": 32,
|
103 |
+
"BLOCK_SIZE_N": 32,
|
104 |
+
"BLOCK_SIZE_K": 256,
|
105 |
+
"GROUP_SIZE_M": 4,
|
106 |
+
"num_warps": 4,
|
107 |
+
"num_stages": 2,
|
108 |
+
"waves_per_eu": 0,
|
109 |
+
"matrix_instr_nonkdim": 16,
|
110 |
+
"kpack": 1
|
111 |
+
},
|
112 |
+
"128": {
|
113 |
+
"BLOCK_SIZE_M": 64,
|
114 |
+
"BLOCK_SIZE_N": 64,
|
115 |
+
"BLOCK_SIZE_K": 128,
|
116 |
+
"GROUP_SIZE_M": 4,
|
117 |
+
"num_warps": 8,
|
118 |
+
"num_stages": 2,
|
119 |
+
"waves_per_eu": 0,
|
120 |
+
"matrix_instr_nonkdim": 16,
|
121 |
+
"kpack": 2
|
122 |
+
},
|
123 |
+
"256": {
|
124 |
+
"BLOCK_SIZE_M": 128,
|
125 |
+
"BLOCK_SIZE_N": 128,
|
126 |
+
"BLOCK_SIZE_K": 64,
|
127 |
+
"GROUP_SIZE_M": 4,
|
128 |
+
"num_warps": 8,
|
129 |
+
"num_stages": 2,
|
130 |
+
"waves_per_eu": 0,
|
131 |
+
"matrix_instr_nonkdim": 16,
|
132 |
+
"kpack": 2
|
133 |
+
},
|
134 |
+
"512": {
|
135 |
+
"BLOCK_SIZE_M": 128,
|
136 |
+
"BLOCK_SIZE_N": 128,
|
137 |
+
"BLOCK_SIZE_K": 64,
|
138 |
+
"GROUP_SIZE_M": 1,
|
139 |
+
"num_warps": 8,
|
140 |
+
"num_stages": 2,
|
141 |
+
"waves_per_eu": 0,
|
142 |
+
"matrix_instr_nonkdim": 16,
|
143 |
+
"kpack": 2
|
144 |
+
},
|
145 |
+
"1024": {
|
146 |
+
"BLOCK_SIZE_M": 128,
|
147 |
+
"BLOCK_SIZE_N": 128,
|
148 |
+
"BLOCK_SIZE_K": 64,
|
149 |
+
"GROUP_SIZE_M": 4,
|
150 |
+
"num_warps": 8,
|
151 |
+
"num_stages": 2,
|
152 |
+
"waves_per_eu": 0,
|
153 |
+
"matrix_instr_nonkdim": 16,
|
154 |
+
"kpack": 2
|
155 |
+
},
|
156 |
+
"1536": {
|
157 |
+
"BLOCK_SIZE_M": 128,
|
158 |
+
"BLOCK_SIZE_N": 128,
|
159 |
+
"BLOCK_SIZE_K": 64,
|
160 |
+
"GROUP_SIZE_M": 4,
|
161 |
+
"num_warps": 8,
|
162 |
+
"num_stages": 2,
|
163 |
+
"waves_per_eu": 0,
|
164 |
+
"matrix_instr_nonkdim": 16,
|
165 |
+
"kpack": 2
|
166 |
+
},
|
167 |
+
"2048": {
|
168 |
+
"BLOCK_SIZE_M": 128,
|
169 |
+
"BLOCK_SIZE_N": 128,
|
170 |
+
"BLOCK_SIZE_K": 64,
|
171 |
+
"GROUP_SIZE_M": 4,
|
172 |
+
"num_warps": 8,
|
173 |
+
"num_stages": 2,
|
174 |
+
"waves_per_eu": 0,
|
175 |
+
"matrix_instr_nonkdim": 16,
|
176 |
+
"kpack": 2
|
177 |
+
},
|
178 |
+
"3072": {
|
179 |
+
"BLOCK_SIZE_M": 128,
|
180 |
+
"BLOCK_SIZE_N": 128,
|
181 |
+
"BLOCK_SIZE_K": 64,
|
182 |
+
"GROUP_SIZE_M": 16,
|
183 |
+
"num_warps": 8,
|
184 |
+
"num_stages": 2,
|
185 |
+
"waves_per_eu": 0,
|
186 |
+
"matrix_instr_nonkdim": 16,
|
187 |
+
"kpack": 2
|
188 |
+
},
|
189 |
+
"4096": {
|
190 |
+
"BLOCK_SIZE_M": 128,
|
191 |
+
"BLOCK_SIZE_N": 128,
|
192 |
+
"BLOCK_SIZE_K": 64,
|
193 |
+
"GROUP_SIZE_M": 16,
|
194 |
+
"num_warps": 8,
|
195 |
+
"num_stages": 2,
|
196 |
+
"waves_per_eu": 0,
|
197 |
+
"matrix_instr_nonkdim": 16,
|
198 |
+
"kpack": 2
|
199 |
+
}
|
200 |
+
}
|
build/torch25-cxx11-cu118-x86_64-linux/moe/configs/E=8,N=16384,device_name=AMD_Instinct_MI325X,dtype=fp8_w8a8.json
ADDED
@@ -0,0 +1,164 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"1": {
|
3 |
+
"BLOCK_SIZE_M": 16,
|
4 |
+
"BLOCK_SIZE_N": 16,
|
5 |
+
"BLOCK_SIZE_K": 256,
|
6 |
+
"GROUP_SIZE_M": 1,
|
7 |
+
"num_warps": 2,
|
8 |
+
"num_stages": 2,
|
9 |
+
"waves_per_eu": 0
|
10 |
+
},
|
11 |
+
"2": {
|
12 |
+
"BLOCK_SIZE_M": 16,
|
13 |
+
"BLOCK_SIZE_N": 16,
|
14 |
+
"BLOCK_SIZE_K": 256,
|
15 |
+
"GROUP_SIZE_M": 1,
|
16 |
+
"num_warps": 4,
|
17 |
+
"num_stages": 2,
|
18 |
+
"waves_per_eu": 0
|
19 |
+
},
|
20 |
+
"4": {
|
21 |
+
"BLOCK_SIZE_M": 16,
|
22 |
+
"BLOCK_SIZE_N": 32,
|
23 |
+
"BLOCK_SIZE_K": 256,
|
24 |
+
"GROUP_SIZE_M": 1,
|
25 |
+
"num_warps": 2,
|
26 |
+
"num_stages": 2,
|
27 |
+
"waves_per_eu": 0
|
28 |
+
},
|
29 |
+
"8": {
|
30 |
+
"BLOCK_SIZE_M": 16,
|
31 |
+
"BLOCK_SIZE_N": 64,
|
32 |
+
"BLOCK_SIZE_K": 256,
|
33 |
+
"GROUP_SIZE_M": 1,
|
34 |
+
"num_warps": 2,
|
35 |
+
"num_stages": 2,
|
36 |
+
"waves_per_eu": 0
|
37 |
+
},
|
38 |
+
"16": {
|
39 |
+
"BLOCK_SIZE_M": 16,
|
40 |
+
"BLOCK_SIZE_N": 64,
|
41 |
+
"BLOCK_SIZE_K": 256,
|
42 |
+
"GROUP_SIZE_M": 1,
|
43 |
+
"num_warps": 2,
|
44 |
+
"num_stages": 2,
|
45 |
+
"waves_per_eu": 0
|
46 |
+
},
|
47 |
+
"24": {
|
48 |
+
"BLOCK_SIZE_M": 16,
|
49 |
+
"BLOCK_SIZE_N": 64,
|
50 |
+
"BLOCK_SIZE_K": 256,
|
51 |
+
"GROUP_SIZE_M": 1,
|
52 |
+
"num_warps": 4,
|
53 |
+
"num_stages": 2,
|
54 |
+
"waves_per_eu": 0
|
55 |
+
},
|
56 |
+
"32": {
|
57 |
+
"BLOCK_SIZE_M": 16,
|
58 |
+
"BLOCK_SIZE_N": 64,
|
59 |
+
"BLOCK_SIZE_K": 256,
|
60 |
+
"GROUP_SIZE_M": 4,
|
61 |
+
"num_warps": 4,
|
62 |
+
"num_stages": 2,
|
63 |
+
"waves_per_eu": 0
|
64 |
+
},
|
65 |
+
"48": {
|
66 |
+
"BLOCK_SIZE_M": 16,
|
67 |
+
"BLOCK_SIZE_N": 64,
|
68 |
+
"BLOCK_SIZE_K": 256,
|
69 |
+
"GROUP_SIZE_M": 1,
|
70 |
+
"num_warps": 4,
|
71 |
+
"num_stages": 2,
|
72 |
+
"waves_per_eu": 0
|
73 |
+
},
|
74 |
+
"64": {
|
75 |
+
"BLOCK_SIZE_M": 32,
|
76 |
+
"BLOCK_SIZE_N": 64,
|
77 |
+
"BLOCK_SIZE_K": 128,
|
78 |
+
"GROUP_SIZE_M": 4,
|
79 |
+
"num_warps": 2,
|
80 |
+
"num_stages": 2,
|
81 |
+
"waves_per_eu": 0
|
82 |
+
},
|
83 |
+
"96": {
|
84 |
+
"BLOCK_SIZE_M": 32,
|
85 |
+
"BLOCK_SIZE_N": 64,
|
86 |
+
"BLOCK_SIZE_K": 256,
|
87 |
+
"GROUP_SIZE_M": 4,
|
88 |
+
"num_warps": 2,
|
89 |
+
"num_stages": 2,
|
90 |
+
"waves_per_eu": 0
|
91 |
+
},
|
92 |
+
"128": {
|
93 |
+
"BLOCK_SIZE_M": 64,
|
94 |
+
"BLOCK_SIZE_N": 64,
|
95 |
+
"BLOCK_SIZE_K": 128,
|
96 |
+
"GROUP_SIZE_M": 4,
|
97 |
+
"num_warps": 4,
|
98 |
+
"num_stages": 2,
|
99 |
+
"waves_per_eu": 0
|
100 |
+
},
|
101 |
+
"256": {
|
102 |
+
"BLOCK_SIZE_M": 64,
|
103 |
+
"BLOCK_SIZE_N": 128,
|
104 |
+
"BLOCK_SIZE_K": 256,
|
105 |
+
"GROUP_SIZE_M": 1,
|
106 |
+
"num_warps": 8,
|
107 |
+
"num_stages": 2,
|
108 |
+
"waves_per_eu": 0
|
109 |
+
},
|
110 |
+
"512": {
|
111 |
+
"BLOCK_SIZE_M": 128,
|
112 |
+
"BLOCK_SIZE_N": 128,
|
113 |
+
"BLOCK_SIZE_K": 256,
|
114 |
+
"GROUP_SIZE_M": 1,
|
115 |
+
"num_warps": 8,
|
116 |
+
"num_stages": 2,
|
117 |
+
"waves_per_eu": 0
|
118 |
+
},
|
119 |
+
"1024": {
|
120 |
+
"BLOCK_SIZE_M": 128,
|
121 |
+
"BLOCK_SIZE_N": 256,
|
122 |
+
"BLOCK_SIZE_K": 128,
|
123 |
+
"GROUP_SIZE_M": 1,
|
124 |
+
"num_warps": 8,
|
125 |
+
"num_stages": 2,
|
126 |
+
"waves_per_eu": 0
|
127 |
+
},
|
128 |
+
"1536": {
|
129 |
+
"BLOCK_SIZE_M": 256,
|
130 |
+
"BLOCK_SIZE_N": 128,
|
131 |
+
"BLOCK_SIZE_K": 128,
|
132 |
+
"GROUP_SIZE_M": 1,
|
133 |
+
"num_warps": 8,
|
134 |
+
"num_stages": 2,
|
135 |
+
"waves_per_eu": 0
|
136 |
+
},
|
137 |
+
"2048": {
|
138 |
+
"BLOCK_SIZE_M": 128,
|
139 |
+
"BLOCK_SIZE_N": 256,
|
140 |
+
"BLOCK_SIZE_K": 128,
|
141 |
+
"GROUP_SIZE_M": 1,
|
142 |
+
"num_warps": 8,
|
143 |
+
"num_stages": 2,
|
144 |
+
"waves_per_eu": 0
|
145 |
+
},
|
146 |
+
"3072": {
|
147 |
+
"BLOCK_SIZE_M": 128,
|
148 |
+
"BLOCK_SIZE_N": 256,
|
149 |
+
"BLOCK_SIZE_K": 128,
|
150 |
+
"GROUP_SIZE_M": 1,
|
151 |
+
"num_warps": 8,
|
152 |
+
"num_stages": 2,
|
153 |
+
"waves_per_eu": 0
|
154 |
+
},
|
155 |
+
"4096": {
|
156 |
+
"BLOCK_SIZE_M": 256,
|
157 |
+
"BLOCK_SIZE_N": 256,
|
158 |
+
"BLOCK_SIZE_K": 64,
|
159 |
+
"GROUP_SIZE_M": 1,
|
160 |
+
"num_warps": 8,
|
161 |
+
"num_stages": 2,
|
162 |
+
"waves_per_eu": 0
|
163 |
+
}
|
164 |
+
}
|