smpanaro commited on
Commit
e548ff4
1 Parent(s): bc97927

Upload of AutoGPTQ quantized model

Browse files
config.json ADDED
@@ -0,0 +1,43 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_name_or_path": "EleutherAI/pythia-6.9b",
3
+ "architectures": [
4
+ "GPTNeoXForCausalLM"
5
+ ],
6
+ "attention_bias": true,
7
+ "attention_dropout": 0.0,
8
+ "bos_token_id": 0,
9
+ "classifier_dropout": 0.1,
10
+ "eos_token_id": 0,
11
+ "hidden_act": "gelu",
12
+ "hidden_dropout": 0.0,
13
+ "hidden_size": 4096,
14
+ "initializer_range": 0.02,
15
+ "intermediate_size": 16384,
16
+ "layer_norm_eps": 1e-05,
17
+ "max_position_embeddings": 2048,
18
+ "model_type": "gpt_neox",
19
+ "num_attention_heads": 32,
20
+ "num_hidden_layers": 32,
21
+ "quantization_config": {
22
+ "bits": 4,
23
+ "damp_percent": 0.01,
24
+ "desc_act": false,
25
+ "group_size": 128,
26
+ "is_marlin_format": false,
27
+ "model_file_base_name": "gptq_model-4bit-128g",
28
+ "model_name_or_path": "EleutherAI/pythia-6.9b-4bit-128g",
29
+ "quant_method": "gptq",
30
+ "static_groups": false,
31
+ "sym": true,
32
+ "true_sequential": true
33
+ },
34
+ "rope_scaling": null,
35
+ "rotary_emb_base": 10000,
36
+ "rotary_pct": 0.25,
37
+ "tie_word_embeddings": false,
38
+ "torch_dtype": "float16",
39
+ "transformers_version": "4.38.2",
40
+ "use_cache": true,
41
+ "use_parallel_residual": true,
42
+ "vocab_size": 50432
43
+ }
gptq_model-4bit-128g.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0b732062ecf374ddecbd7decef9a9b1bd405711376b244bafbf8842aff446737
3
+ size 4180518352
quantize_config.json ADDED
@@ -0,0 +1,13 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "bits": 4,
3
+ "group_size": 128,
4
+ "damp_percent": 0.01,
5
+ "desc_act": false,
6
+ "static_groups": false,
7
+ "sym": true,
8
+ "true_sequential": true,
9
+ "model_name_or_path": "EleutherAI/pythia-6.9b-4bit-128g",
10
+ "model_file_base_name": "gptq_model-4bit-128g",
11
+ "is_marlin_format": false,
12
+ "quant_method": "gptq"
13
+ }