Shubhra Pandit commited on
Commit
183210a
·
1 Parent(s): 06c60d1

Update model files

Browse files
Files changed (3) hide show
  1. config.json +2 -2
  2. model.safetensors +1 -1
  3. recipe.yaml +5 -1
config.json CHANGED
@@ -26,7 +26,7 @@
26
  "Linear"
27
  ],
28
  "weights": {
29
- "actorder": null,
30
  "block_structure": null,
31
  "dynamic": false,
32
  "group_size": 128,
@@ -211,7 +211,7 @@
211
  "quantization_status": "compressed",
212
  "sparsity_config": {
213
  "format": "dense",
214
- "global_sparsity": 0.1291273641521281,
215
  "ignore": [
216
  "visual.blocks.0.attn.qkv",
217
  "visual.blocks.0.attn.proj",
 
26
  "Linear"
27
  ],
28
  "weights": {
29
+ "actorder": "weight",
30
  "block_structure": null,
31
  "dynamic": false,
32
  "group_size": 128,
 
211
  "quantization_status": "compressed",
212
  "sparsity_config": {
213
  "format": "dense",
214
+ "global_sparsity": 0.12852197810556548,
215
  "ignore": [
216
  "visual.blocks.0.attn.qkv",
217
  "visual.blocks.0.attn.proj",
model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:07853553261fc4237824096ec8bd93223ccea164df4fd4b47a483cc6c7837326
3
  size 4013282536
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1811c5fa738747144c60821401a200d4e56339d80377c0c497eb3b341d85ff23
3
  size 4013282536
recipe.yaml CHANGED
@@ -3,6 +3,10 @@ DEFAULT_stage:
3
  GPTQModifier:
4
  sequential_targets: [Qwen2_5_VLDecoderLayer]
5
  dampening_frac: 0.01
6
- scheme: W4A16
 
 
 
 
7
  targets: Linear
8
  ignore: [lm_head, 're:visual.*']
 
3
  GPTQModifier:
4
  sequential_targets: [Qwen2_5_VLDecoderLayer]
5
  dampening_frac: 0.01
6
+ config_groups:
7
+ config_group:
8
+ targets: [Linear]
9
+ weights: {num_bits: 4, type: int, symmetric: true, group_size: 128, strategy: group,
10
+ dynamic: false, actorder: weight, observer: minmax}
11
  targets: Linear
12
  ignore: [lm_head, 're:visual.*']