AutoGPTQ model for Qwen/Qwen-7B: 4bits, gr128, desc_act=False
Browse files- config.json +55 -0
- gptq_model-4bit-128g.bin +3 -0
- gptq_model-4bit-128g.safetensors +3 -0
- quantize_config.json +13 -0
config.json
ADDED
|
@@ -0,0 +1,55 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_name_or_path": "Qwen/Qwen-7B",
|
| 3 |
+
"architectures": [
|
| 4 |
+
"QWenLMHeadModel"
|
| 5 |
+
],
|
| 6 |
+
"attn_dropout_prob": 0.0,
|
| 7 |
+
"auto_map": {
|
| 8 |
+
"AutoConfig": "Qwen/Qwen-7B--configuration_qwen.QWenConfig",
|
| 9 |
+
"AutoModelForCausalLM": "Qwen/Qwen-7B--modeling_qwen.QWenLMHeadModel"
|
| 10 |
+
},
|
| 11 |
+
"bf16": true,
|
| 12 |
+
"emb_dropout_prob": 0.0,
|
| 13 |
+
"fp16": false,
|
| 14 |
+
"fp32": false,
|
| 15 |
+
"hidden_size": 4096,
|
| 16 |
+
"initializer_range": 0.02,
|
| 17 |
+
"intermediate_size": 22016,
|
| 18 |
+
"kv_channels": 128,
|
| 19 |
+
"layer_norm_epsilon": 1e-06,
|
| 20 |
+
"max_position_embeddings": 32768,
|
| 21 |
+
"model_type": "qwen",
|
| 22 |
+
"no_bias": true,
|
| 23 |
+
"num_attention_heads": 32,
|
| 24 |
+
"num_hidden_layers": 32,
|
| 25 |
+
"onnx_safe": null,
|
| 26 |
+
"quantization_config": {
|
| 27 |
+
"bits": 4,
|
| 28 |
+
"damp_percent": 0.01,
|
| 29 |
+
"desc_act": false,
|
| 30 |
+
"group_size": 128,
|
| 31 |
+
"is_marlin_format": false,
|
| 32 |
+
"model_file_base_name": "gptq_model-4bit-128g",
|
| 33 |
+
"model_name_or_path": "Qwen-7B-4bit",
|
| 34 |
+
"quant_method": "gptq",
|
| 35 |
+
"static_groups": false,
|
| 36 |
+
"sym": true,
|
| 37 |
+
"true_sequential": true
|
| 38 |
+
},
|
| 39 |
+
"rotary_emb_base": 10000,
|
| 40 |
+
"rotary_pct": 1.0,
|
| 41 |
+
"scale_attn_weights": true,
|
| 42 |
+
"seq_length": 8192,
|
| 43 |
+
"softmax_in_fp32": false,
|
| 44 |
+
"tie_word_embeddings": false,
|
| 45 |
+
"tokenizer_class": "QWenTokenizer",
|
| 46 |
+
"torch_dtype": "float16",
|
| 47 |
+
"transformers_version": "4.37.2",
|
| 48 |
+
"use_cache": true,
|
| 49 |
+
"use_cache_kernel": false,
|
| 50 |
+
"use_cache_quantization": false,
|
| 51 |
+
"use_dynamic_ntk": true,
|
| 52 |
+
"use_flash_attn": true,
|
| 53 |
+
"use_logn_attn": true,
|
| 54 |
+
"vocab_size": 151936
|
| 55 |
+
}
|
gptq_model-4bit-128g.bin
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:2393b2d40537456cd8cd8cd876999c6ef834e4edc9346e8b7be6e4fa0769a90f
|
| 3 |
+
size 5860862723
|
gptq_model-4bit-128g.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:87a7324c6f793caaec5cc1c2c65d8b749dcba244a1770b636b84cffce00fb0ee
|
| 3 |
+
size 5860657608
|
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": "Qwen-7B-4bit",
|
| 10 |
+
"model_file_base_name": "gptq_model-4bit-128g",
|
| 11 |
+
"is_marlin_format": false,
|
| 12 |
+
"quant_method": "gptq"
|
| 13 |
+
}
|