Synchronizing local compiler cache.
Browse files
neuronxcc-2.17.194.0+d312836f/0_REGISTRY/0.3.0.dev0/llama/unsloth/Llama-3.2-1B-Instruct/7726b0c1841e33ac8fb4.json
ADDED
@@ -0,0 +1,56 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"_entry_class": "SingleModelCacheEntry",
|
3 |
+
"_model_id": "unsloth/Llama-3.2-1B-Instruct",
|
4 |
+
"_task": "text-generation",
|
5 |
+
"architectures": [
|
6 |
+
"LlamaForCausalLM"
|
7 |
+
],
|
8 |
+
"attention_bias": false,
|
9 |
+
"attention_dropout": 0.0,
|
10 |
+
"head_dim": 64,
|
11 |
+
"hidden_act": "silu",
|
12 |
+
"hidden_size": 2048,
|
13 |
+
"initializer_range": 0.02,
|
14 |
+
"intermediate_size": 8192,
|
15 |
+
"max_position_embeddings": 131072,
|
16 |
+
"mlp_bias": false,
|
17 |
+
"model_type": "llama",
|
18 |
+
"neuron": {
|
19 |
+
"_serialized_key": "HloNeuronConfig",
|
20 |
+
"all_reduce_dtype": null,
|
21 |
+
"allow_flash_attention": true,
|
22 |
+
"attention_layout": "BSH",
|
23 |
+
"attn_output_transposed": false,
|
24 |
+
"auto_cast_type": "fp16",
|
25 |
+
"batch_size": 4,
|
26 |
+
"checkpoint_id": "unsloth/Llama-3.2-1B-Instruct",
|
27 |
+
"checkpoint_revision": "5a8abab4a5d6f164389b1079fb721cfab8d7126c",
|
28 |
+
"collectives_layout": "HSB",
|
29 |
+
"continuous_batching": true,
|
30 |
+
"fuse_qkv": true,
|
31 |
+
"group_query_attention": "shard-over-heads",
|
32 |
+
"log_softmax_scores": false,
|
33 |
+
"neuronxcc_version": "2.17.194.0+d312836f",
|
34 |
+
"optimum_neuron_version": "0.3.0.dev0",
|
35 |
+
"output_all_logits": false,
|
36 |
+
"sequence_length": 4096,
|
37 |
+
"tp_degree": 2
|
38 |
+
},
|
39 |
+
"num_attention_heads": 32,
|
40 |
+
"num_hidden_layers": 16,
|
41 |
+
"num_key_value_heads": 8,
|
42 |
+
"pretraining_tp": 1,
|
43 |
+
"rms_norm_eps": 1e-05,
|
44 |
+
"rope_scaling": {
|
45 |
+
"factor": 32.0,
|
46 |
+
"high_freq_factor": 4.0,
|
47 |
+
"low_freq_factor": 1.0,
|
48 |
+
"original_max_position_embeddings": 8192,
|
49 |
+
"rope_type": "llama3"
|
50 |
+
},
|
51 |
+
"rope_theta": 500000.0,
|
52 |
+
"tie_word_embeddings": true,
|
53 |
+
"unsloth_fixed": true,
|
54 |
+
"use_cache": true,
|
55 |
+
"vocab_size": 128256
|
56 |
+
}
|