1default_stage:
2 default_modifiers:
3 GPTQModifier:
4 config_groups:
5 group_0:
6 targets: [Linear]
7 weights:
8 num_bits: 4
9 type: float
10 symmetric: true
11 group_size: 16
12 strategy: tensor_group
13 block_structure: null
14 dynamic: false
15 actorder: static
16 scale_dtype: torch.float8_e4m3fn
17 zp_dtype: null
18 observer: imatrix_mse
19 observer_kwargs: {strict: true, expand: 1.8, maxshrink: 1 - 0.8 / 1.8, grid: 200.0, norm: 2.4, patience: 1000}
20 input_activations:
21 num_bits: 4
22 type: float
23 symmetric: true
24 group_size: 16
25 strategy: tensor_group
26 block_structure: null
27 dynamic: local
28 actorder: null
29 scale_dtype: torch.float8_e4m3fn
30 zp_dtype: null
31 observer: static_minmax
32 observer_kwargs: {}
33 output_activations: null
34 format: null
35 targets: [Linear]
36 ignore: [lm_head, 're:.*embed.*', 're:.*vision.*', 're:.*audio.*', 're:.*router.*', 're:.*per_layer.*']
37 kv_cache_scheme:
38 num_bits: 8
39 type: float
40 symmetric: true
41 group_size: null
42 strategy: tensor
43 block_structure: null
44 dynamic: false
45 actorder: null
46 scale_dtype: null
47 zp_dtype: null
48 observer: static_minmax
49 observer_kwargs: {}
50 bypass_divisibility_checks: false
51 requires_calibration_data: true
52 block_size: 128
53 dampening_frac: 0.01
54 actorder: static
55 offload_hessians: false
1transformers: 5.14.1
2llmcompressor: 0.13.1.dev34+g040aa8cb3
3compressed-tensors: 0.18.1a20260818
