Hugging Face's logo

nvidia
/
Kimi-K2.5-NVFP4

Text Generation
Safetensors
Model Optimizer
kimi_k25
nvidia
ModelOpt
Kimi-K2
quantized
FP4
conversational
custom_code
modelopt
Model card Files
xet
Community
5
zhiyucheng commited on
Commit
8d5dc8b
·
verified ·
1 Parent(s): 3f7790f

Fix: add .model after language_model in quantization ignore/exclude_modules

Browse files

In both `config.json` (`quantization_config.ignore`) and `hf_quant_config.json` (`quantization.exclude_modules`), all entries starting with `language_model.` have been updated to `language_model.model.` to correctly reference the submodule path.

For example:
- `language_model.lm_head` → `language_model.model.lm_head`
- `language_model.layers.*.self_attn*` → `language_model.model.layers.*.self_attn*`

Files changed (2) hide show
  1. config.json +62 -62
  2. hf_quant_config.json +62 -62
config.json CHANGED
@@ -35,68 +35,68 @@
35
  }
36
  },
37
  "ignore": [
38
- "language_model.lm_head",
39
- "language_model.layers.0.self_attn*",
40
- "language_model.layers.1.self_attn*",
41
- "language_model.layers.10.self_attn*",
42
- "language_model.layers.11.self_attn*",
43
- "language_model.layers.12.self_attn*",
44
- "language_model.layers.13.self_attn*",
45
- "language_model.layers.14.self_attn*",
46
- "language_model.layers.15.self_attn*",
47
- "language_model.layers.16.self_attn*",
48
- "language_model.layers.17.self_attn*",
49
- "language_model.layers.18.self_attn*",
50
- "language_model.layers.19.self_attn*",
51
- "language_model.layers.2.self_attn*",
52
- "language_model.layers.20.self_attn*",
53
- "language_model.layers.21.self_attn*",
54
- "language_model.layers.22.self_attn*",
55
- "language_model.layers.23.self_attn*",
56
- "language_model.layers.24.self_attn*",
57
- "language_model.layers.25.self_attn*",
58
- "language_model.layers.26.self_attn*",
59
- "language_model.layers.27.self_attn*",
60
- "language_model.layers.28.self_attn*",
61
- "language_model.layers.29.self_attn*",
62
- "language_model.layers.3.self_attn*",
63
- "language_model.layers.30.self_attn*",
64
- "language_model.layers.31.self_attn*",
65
- "language_model.layers.32.self_attn*",
66
- "language_model.layers.33.self_attn*",
67
- "language_model.layers.34.self_attn*",
68
- "language_model.layers.35.self_attn*",
69
- "language_model.layers.36.self_attn*",
70
- "language_model.layers.37.self_attn*",
71
- "language_model.layers.38.self_attn*",
72
- "language_model.layers.39.self_attn*",
73
- "language_model.layers.4.self_attn*",
74
- "language_model.layers.40.self_attn*",
75
- "language_model.layers.41.self_attn*",
76
- "language_model.layers.42.self_attn*",
77
- "language_model.layers.43.self_attn*",
78
- "language_model.layers.44.self_attn*",
79
- "language_model.layers.45.self_attn*",
80
- "language_model.layers.46.self_attn*",
81
- "language_model.layers.47.self_attn*",
82
- "language_model.layers.48.self_attn*",
83
- "language_model.layers.49.self_attn*",
84
- "language_model.layers.5.self_attn*",
85
- "language_model.layers.50.self_attn*",
86
- "language_model.layers.51.self_attn*",
87
- "language_model.layers.52.self_attn*",
88
- "language_model.layers.53.self_attn*",
89
- "language_model.layers.54.self_attn*",
90
- "language_model.layers.55.self_attn*",
91
- "language_model.layers.56.self_attn*",
92
- "language_model.layers.57.self_attn*",
93
- "language_model.layers.58.self_attn*",
94
- "language_model.layers.59.self_attn*",
95
- "language_model.layers.6.self_attn*",
96
- "language_model.layers.60.self_attn*",
97
- "language_model.layers.7.self_attn*",
98
- "language_model.layers.8.self_attn*",
99
- "language_model.layers.9.self_attn*",
100
  "mm_projector*",
101
  "vision_tower*"
102
  ],
 
35
  }
36
  },
37
  "ignore": [
38
+ "language_model.model.lm_head",
39
+ "language_model.model.layers.0.self_attn*",
40
+ "language_model.model.layers.1.self_attn*",
41
+ "language_model.model.layers.10.self_attn*",
42
+ "language_model.model.layers.11.self_attn*",
43
+ "language_model.model.layers.12.self_attn*",
44
+ "language_model.model.layers.13.self_attn*",
45
+ "language_model.model.layers.14.self_attn*",
46
+ "language_model.model.layers.15.self_attn*",
47
+ "language_model.model.layers.16.self_attn*",
48
+ "language_model.model.layers.17.self_attn*",
49
+ "language_model.model.layers.18.self_attn*",
50
+ "language_model.model.layers.19.self_attn*",
51
+ "language_model.model.layers.2.self_attn*",
52
+ "language_model.model.layers.20.self_attn*",
53
+ "language_model.model.layers.21.self_attn*",
54
+ "language_model.model.layers.22.self_attn*",
55
+ "language_model.model.layers.23.self_attn*",
56
+ "language_model.model.layers.24.self_attn*",
57
+ "language_model.model.layers.25.self_attn*",
58
+ "language_model.model.layers.26.self_attn*",
59
+ "language_model.model.layers.27.self_attn*",
60
+ "language_model.model.layers.28.self_attn*",
61
+ "language_model.model.layers.29.self_attn*",
62
+ "language_model.model.layers.3.self_attn*",
63
+ "language_model.model.layers.30.self_attn*",
64
+ "language_model.model.layers.31.self_attn*",
65
+ "language_model.model.layers.32.self_attn*",
66
+ "language_model.model.layers.33.self_attn*",
67
+ "language_model.model.layers.34.self_attn*",
68
+ "language_model.model.layers.35.self_attn*",
69
+ "language_model.model.layers.36.self_attn*",
70
+ "language_model.model.layers.37.self_attn*",
71
+ "language_model.model.layers.38.self_attn*",
72
+ "language_model.model.layers.39.self_attn*",
73
+ "language_model.model.layers.4.self_attn*",
74
+ "language_model.model.layers.40.self_attn*",
75
+ "language_model.model.layers.41.self_attn*",
76
+ "language_model.model.layers.42.self_attn*",
77
+ "language_model.model.layers.43.self_attn*",
78
+ "language_model.model.layers.44.self_attn*",
79
+ "language_model.model.layers.45.self_attn*",
80
+ "language_model.model.layers.46.self_attn*",
81
+ "language_model.model.layers.47.self_attn*",
82
+ "language_model.model.layers.48.self_attn*",
83
+ "language_model.model.layers.49.self_attn*",
84
+ "language_model.model.layers.5.self_attn*",
85
+ "language_model.model.layers.50.self_attn*",
86
+ "language_model.model.layers.51.self_attn*",
87
+ "language_model.model.layers.52.self_attn*",
88
+ "language_model.model.layers.53.self_attn*",
89
+ "language_model.model.layers.54.self_attn*",
90
+ "language_model.model.layers.55.self_attn*",
91
+ "language_model.model.layers.56.self_attn*",
92
+ "language_model.model.layers.57.self_attn*",
93
+ "language_model.model.layers.58.self_attn*",
94
+ "language_model.model.layers.59.self_attn*",
95
+ "language_model.model.layers.6.self_attn*",
96
+ "language_model.model.layers.60.self_attn*",
97
+ "language_model.model.layers.7.self_attn*",
98
+ "language_model.model.layers.8.self_attn*",
99
+ "language_model.model.layers.9.self_attn*",
100
  "mm_projector*",
101
  "vision_tower*"
102
  ],
hf_quant_config.json CHANGED
@@ -8,68 +8,68 @@
8
  "kv_cache_quant_algo": "FP8",
9
  "group_size": 16,
10
  "exclude_modules": [
11
- "language_model.lm_head",
12
- "language_model.layers.0.self_attn*",
13
- "language_model.layers.1.self_attn*",
14
- "language_model.layers.10.self_attn*",
15
- "language_model.layers.11.self_attn*",
16
- "language_model.layers.12.self_attn*",
17
- "language_model.layers.13.self_attn*",
18
- "language_model.layers.14.self_attn*",
19
- "language_model.layers.15.self_attn*",
20
- "language_model.layers.16.self_attn*",
21
- "language_model.layers.17.self_attn*",
22
- "language_model.layers.18.self_attn*",
23
- "language_model.layers.19.self_attn*",
24
- "language_model.layers.2.self_attn*",
25
- "language_model.layers.20.self_attn*",
26
- "language_model.layers.21.self_attn*",
27
- "language_model.layers.22.self_attn*",
28
- "language_model.layers.23.self_attn*",
29
- "language_model.layers.24.self_attn*",
30
- "language_model.layers.25.self_attn*",
31
- "language_model.layers.26.self_attn*",
32
- "language_model.layers.27.self_attn*",
33
- "language_model.layers.28.self_attn*",
34
- "language_model.layers.29.self_attn*",
35
- "language_model.layers.3.self_attn*",
36
- "language_model.layers.30.self_attn*",
37
- "language_model.layers.31.self_attn*",
38
- "language_model.layers.32.self_attn*",
39
- "language_model.layers.33.self_attn*",
40
- "language_model.layers.34.self_attn*",
41
- "language_model.layers.35.self_attn*",
42
- "language_model.layers.36.self_attn*",
43
- "language_model.layers.37.self_attn*",
44
- "language_model.layers.38.self_attn*",
45
- "language_model.layers.39.self_attn*",
46
- "language_model.layers.4.self_attn*",
47
- "language_model.layers.40.self_attn*",
48
- "language_model.layers.41.self_attn*",
49
- "language_model.layers.42.self_attn*",
50
- "language_model.layers.43.self_attn*",
51
- "language_model.layers.44.self_attn*",
52
- "language_model.layers.45.self_attn*",
53
- "language_model.layers.46.self_attn*",
54
- "language_model.layers.47.self_attn*",
55
- "language_model.layers.48.self_attn*",
56
- "language_model.layers.49.self_attn*",
57
- "language_model.layers.5.self_attn*",
58
- "language_model.layers.50.self_attn*",
59
- "language_model.layers.51.self_attn*",
60
- "language_model.layers.52.self_attn*",
61
- "language_model.layers.53.self_attn*",
62
- "language_model.layers.54.self_attn*",
63
- "language_model.layers.55.self_attn*",
64
- "language_model.layers.56.self_attn*",
65
- "language_model.layers.57.self_attn*",
66
- "language_model.layers.58.self_attn*",
67
- "language_model.layers.59.self_attn*",
68
- "language_model.layers.6.self_attn*",
69
- "language_model.layers.60.self_attn*",
70
- "language_model.layers.7.self_attn*",
71
- "language_model.layers.8.self_attn*",
72
- "language_model.layers.9.self_attn*",
73
  "mm_projector*",
74
  "vision_tower*"
75
  ]
 
8
  "kv_cache_quant_algo": "FP8",
9
  "group_size": 16,
10
  "exclude_modules": [
11
+ "language_model.model.lm_head",
12
+ "language_model.model.layers.0.self_attn*",
13
+ "language_model.model.layers.1.self_attn*",
14
+ "language_model.model.layers.10.self_attn*",
15
+ "language_model.model.layers.11.self_attn*",
16
+ "language_model.model.layers.12.self_attn*",
17
+ "language_model.model.layers.13.self_attn*",
18
+ "language_model.model.layers.14.self_attn*",
19
+ "language_model.model.layers.15.self_attn*",
20
+ "language_model.model.layers.16.self_attn*",
21
+ "language_model.model.layers.17.self_attn*",
22
+ "language_model.model.layers.18.self_attn*",
23
+ "language_model.model.layers.19.self_attn*",
24
+ "language_model.model.layers.2.self_attn*",
25
+ "language_model.model.layers.20.self_attn*",
26
+ "language_model.model.layers.21.self_attn*",
27
+ "language_model.model.layers.22.self_attn*",
28
+ "language_model.model.layers.23.self_attn*",
29
+ "language_model.model.layers.24.self_attn*",
30
+ "language_model.model.layers.25.self_attn*",
31
+ "language_model.model.layers.26.self_attn*",
32
+ "language_model.model.layers.27.self_attn*",
33
+ "language_model.model.layers.28.self_attn*",
34
+ "language_model.model.layers.29.self_attn*",
35
+ "language_model.model.layers.3.self_attn*",
36
+ "language_model.model.layers.30.self_attn*",
37
+ "language_model.model.layers.31.self_attn*",
38
+ "language_model.model.layers.32.self_attn*",
39
+ "language_model.model.layers.33.self_attn*",
40
+ "language_model.model.layers.34.self_attn*",
41
+ "language_model.model.layers.35.self_attn*",
42
+ "language_model.model.layers.36.self_attn*",
43
+ "language_model.model.layers.37.self_attn*",
44
+ "language_model.model.layers.38.self_attn*",
45
+ "language_model.model.layers.39.self_attn*",
46
+ "language_model.model.layers.4.self_attn*",
47
+ "language_model.model.layers.40.self_attn*",
48
+ "language_model.model.layers.41.self_attn*",
49
+ "language_model.model.layers.42.self_attn*",
50
+ "language_model.model.layers.43.self_attn*",
51
+ "language_model.model.layers.44.self_attn*",
52
+ "language_model.model.layers.45.self_attn*",
53
+ "language_model.model.layers.46.self_attn*",
54
+ "language_model.model.layers.47.self_attn*",
55
+ "language_model.model.layers.48.self_attn*",
56
+ "language_model.model.layers.49.self_attn*",
57
+ "language_model.model.layers.5.self_attn*",
58
+ "language_model.model.layers.50.self_attn*",
59
+ "language_model.model.layers.51.self_attn*",
60
+ "language_model.model.layers.52.self_attn*",
61
+ "language_model.model.layers.53.self_attn*",
62
+ "language_model.model.layers.54.self_attn*",
63
+ "language_model.model.layers.55.self_attn*",
64
+ "language_model.model.layers.56.self_attn*",
65
+ "language_model.model.layers.57.self_attn*",
66
+ "language_model.model.layers.58.self_attn*",
67
+ "language_model.model.layers.59.self_attn*",
68
+ "language_model.model.layers.6.self_attn*",
69
+ "language_model.model.layers.60.self_attn*",
70
+ "language_model.model.layers.7.self_attn*",
71
+ "language_model.model.layers.8.self_attn*",
72
+ "language_model.model.layers.9.self_attn*",
73
  "mm_projector*",
74
  "vision_tower*"
75
  ]