Commit
•
d8772a6
1
Parent(s):
4d4454a
quantized commit
Browse files- config.json +50 -0
- gptq_model-4bit-128g.safetensors +3 -0
- quantize_config.json +13 -0
config.json
ADDED
|
@@ -0,0 +1,50 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_attn_implementation_autoset": true,
|
| 3 |
+
"_name_or_path": "meta-llama/Llama-3.2-1B",
|
| 4 |
+
"architectures": [
|
| 5 |
+
"LlamaForCausalLM"
|
| 6 |
+
],
|
| 7 |
+
"attention_bias": false,
|
| 8 |
+
"attention_dropout": 0.0,
|
| 9 |
+
"bos_token_id": 128000,
|
| 10 |
+
"eos_token_id": 128001,
|
| 11 |
+
"head_dim": 64,
|
| 12 |
+
"hidden_act": "silu",
|
| 13 |
+
"hidden_size": 2048,
|
| 14 |
+
"initializer_range": 0.02,
|
| 15 |
+
"intermediate_size": 8192,
|
| 16 |
+
"max_position_embeddings": 131072,
|
| 17 |
+
"mlp_bias": false,
|
| 18 |
+
"model_type": "llama",
|
| 19 |
+
"num_attention_heads": 32,
|
| 20 |
+
"num_hidden_layers": 16,
|
| 21 |
+
"num_key_value_heads": 8,
|
| 22 |
+
"pretraining_tp": 1,
|
| 23 |
+
"quantization_config": {
|
| 24 |
+
"bits": 4,
|
| 25 |
+
"damp_percent": 0.01,
|
| 26 |
+
"desc_act": false,
|
| 27 |
+
"group_size": 128,
|
| 28 |
+
"is_marlin_format": false,
|
| 29 |
+
"model_file_base_name": null,
|
| 30 |
+
"model_name_or_path": null,
|
| 31 |
+
"quant_method": "gptq",
|
| 32 |
+
"static_groups": false,
|
| 33 |
+
"sym": true,
|
| 34 |
+
"true_sequential": true
|
| 35 |
+
},
|
| 36 |
+
"rms_norm_eps": 1e-05,
|
| 37 |
+
"rope_scaling": {
|
| 38 |
+
"factor": 32.0,
|
| 39 |
+
"high_freq_factor": 4.0,
|
| 40 |
+
"low_freq_factor": 1.0,
|
| 41 |
+
"original_max_position_embeddings": 8192,
|
| 42 |
+
"rope_type": "llama3"
|
| 43 |
+
},
|
| 44 |
+
"rope_theta": 500000.0,
|
| 45 |
+
"tie_word_embeddings": true,
|
| 46 |
+
"torch_dtype": "float16",
|
| 47 |
+
"transformers_version": "4.47.0",
|
| 48 |
+
"use_cache": true,
|
| 49 |
+
"vocab_size": 128256
|
| 50 |
+
}
|
gptq_model-4bit-128g.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:20d26ca865f4cac245b5a6ffdb38d5691aa6ab84c16ce1ee0278023ce8f9e52b
|
| 3 |
+
size 1558482944
|
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": null,
|
| 10 |
+
"model_file_base_name": null,
|
| 11 |
+
"is_marlin_format": false,
|
| 12 |
+
"quant_method": "gptq"
|
| 13 |
+
}
|