Hugging Face's logo

nytopop
/
Qwen3-8B-abliterated.w8a8

Text Generation
Transformers
Safetensors
qwen3
conversational
8-bit precision
compressed-tensors
Model card Files Community
nytopop commited on
Commit
f9580de
·
verified ·
1 Parent(s): 05ccd2f

Upload folder using huggingface_hub

Browse files
Files changed (4) hide show
  1. README.md +1 -1
  2. model-00001-of-00002.safetensors +1 -1
  3. model-00002-of-00002.safetensors +1 -1
  4. recipe.yaml +1 -1
README.md CHANGED
@@ -66,7 +66,7 @@ recipe = [
66
  targets="Linear",
67
  scheme="W8A8",
68
  ignore=["lm_head", "re:.*mlp.gate$"],
69
- dampening_frac=0.01,
70
  ),
71
  ]
72
 
 
66
  targets="Linear",
67
  scheme="W8A8",
68
  ignore=["lm_head", "re:.*mlp.gate$"],
69
+ dampening_frac=0.05,
70
  ),
71
  ]
72
 
model-00001-of-00002.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:ef799d36a6c94c9a2f8c1c37df30797d5c6104fb60766f09047de6e4c3cf3240
3
  size 4954286136
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:165f92726eaa123db62c14b1f96229a16ebe75061f6f9d69c21748fbaf1edcbb
3
  size 4954286136
model-00002-of-00002.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:2d11cd20214d412b7116da8b441207c4c581eaae30dfc49f6e074c442585b57b
3
  size 4484294568
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0c650f13f4d30dd897c230c77451da4403ab03683c23b25adb526eafc6895b47
3
  size 4484294568
recipe.yaml CHANGED
@@ -2,7 +2,7 @@ default_stage:
2
  default_modifiers:
3
  SmoothQuantModifier: {smoothing_strength: 0.7}
4
  GPTQModifier:
5
- dampening_frac: 0.01
6
  scheme: W8A8
7
  targets: Linear
8
  ignore: [lm_head, 're:.*mlp.gate$']
 
2
  default_modifiers:
3
  SmoothQuantModifier: {smoothing_strength: 0.7}
4
  GPTQModifier:
5
+ dampening_frac: 0.05
6
  scheme: W8A8
7
  targets: Linear
8
  ignore: [lm_head, 're:.*mlp.gate$']