Upload quantization_config.json with huggingface_hub
Browse files- quantization_config.json +24 -0
quantization_config.json
ADDED
|
@@ -0,0 +1,24 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"quant_method": "gptq",
|
| 3 |
+
"bits": 4,
|
| 4 |
+
"tokenizer": null,
|
| 5 |
+
"dataset": "c4",
|
| 6 |
+
"group_size": 32,
|
| 7 |
+
"damp_percent": 0.1,
|
| 8 |
+
"desc_act": true,
|
| 9 |
+
"sym": true,
|
| 10 |
+
"true_sequential": true,
|
| 11 |
+
"use_cuda_fp16": false,
|
| 12 |
+
"model_seqlen": null,
|
| 13 |
+
"block_name_to_quantize": null,
|
| 14 |
+
"module_name_preceding_first_block": null,
|
| 15 |
+
"batch_size": 1,
|
| 16 |
+
"pad_token_id": null,
|
| 17 |
+
"use_exllama": true,
|
| 18 |
+
"max_input_length": null,
|
| 19 |
+
"exllama_config": {
|
| 20 |
+
"version": 1
|
| 21 |
+
},
|
| 22 |
+
"cache_block_outputs": true,
|
| 23 |
+
"modules_in_block_to_quantize": null
|
| 24 |
+
}
|