n1ck-guo commited on
Commit
9ef095a
·
1 Parent(s): 8190bfa

upload autoround format

Browse files

Signed-off-by: n1ck-guo <[email protected]>

config.json CHANGED
@@ -22,10 +22,10 @@
22
  "quantization_config": {
23
  "amp": true,
24
  "autoround_version": "0.3.1.dev",
 
25
  "bits": 4,
26
- "damp_percent": 0.01,
27
  "data_type": "int",
28
- "desc_act": false,
29
  "enable_minmax_tuning": true,
30
  "enable_norm_bias_tuning": false,
31
  "enable_quanted_input": true,
@@ -37,12 +37,11 @@
37
  "minmax_lr": 0.001,
38
  "nsamples": 512,
39
  "quant_block_list": null,
40
- "quant_method": "gptq",
41
  "scale_dtype": "torch.float16",
42
  "seqlen": 2048,
43
  "sym": true,
44
- "train_bs": 8,
45
- "true_sequential": false
46
  },
47
  "resid_pdrop": 0.1,
48
  "rope_scaling": null,
 
22
  "quantization_config": {
23
  "amp": true,
24
  "autoround_version": "0.3.1.dev",
25
+ "backend": "auto_round:gptq:exllamav2",
26
  "bits": 4,
 
27
  "data_type": "int",
28
+ "dataset": "NeelNanda/pile-10k",
29
  "enable_minmax_tuning": true,
30
  "enable_norm_bias_tuning": false,
31
  "enable_quanted_input": true,
 
37
  "minmax_lr": 0.001,
38
  "nsamples": 512,
39
  "quant_block_list": null,
40
+ "quant_method": "intel/auto-round",
41
  "scale_dtype": "torch.float16",
42
  "seqlen": 2048,
43
  "sym": true,
44
+ "train_bs": 8
 
45
  },
46
  "resid_pdrop": 0.1,
47
  "rope_scaling": null,
quantize_config.json → quantization_config.json RENAMED
@@ -17,9 +17,8 @@
17
  "low_gpu_mem_usage": false,
18
  "quant_block_list": null,
19
  "enable_norm_bias_tuning": false,
 
20
  "autoround_version": "0.3.1.dev",
21
- "quant_method": "gptq",
22
- "desc_act": false,
23
- "true_sequential": false,
24
- "damp_percent": 0.01
25
  }
 
17
  "low_gpu_mem_usage": false,
18
  "quant_block_list": null,
19
  "enable_norm_bias_tuning": false,
20
+ "dataset": "NeelNanda/pile-10k",
21
  "autoround_version": "0.3.1.dev",
22
+ "quant_method": "intel/auto-round",
23
+ "backend": "auto_round:gptq:exllamav2"
 
 
24
  }