Update README.md
Browse files
README.md
CHANGED
|
@@ -6,6 +6,22 @@ Run:
|
|
| 6 |
1. Use PR https://github.com/AutoGPTQ/AutoGPTQ/pull/625
|
| 7 |
2. Need ~68GB of VRAM (1xA100 80G will do)
|
| 8 |
3. Use combine_sensors.sh script to combine the two split files into one. HF has max 50GB file size limit.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 9 |
---
|
| 10 |
inference: false
|
| 11 |
license: other
|
|
|
|
| 6 |
1. Use PR https://github.com/AutoGPTQ/AutoGPTQ/pull/625
|
| 7 |
2. Need ~68GB of VRAM (1xA100 80G will do)
|
| 8 |
3. Use combine_sensors.sh script to combine the two split files into one. HF has max 50GB file size limit.
|
| 9 |
+
|
| 10 |
+
```json
|
| 11 |
+
{
|
| 12 |
+
"bits": 4,
|
| 13 |
+
"group_size": 128,
|
| 14 |
+
"damp_percent": 0.005,
|
| 15 |
+
"desc_act": false,
|
| 16 |
+
"static_groups": false,
|
| 17 |
+
"sym": true,
|
| 18 |
+
"true_sequential": true,
|
| 19 |
+
"model_name_or_path": null,
|
| 20 |
+
"model_file_base_name": null,
|
| 21 |
+
"quant_method": "gptq",
|
| 22 |
+
"checkpoint_format": "gptq"
|
| 23 |
+
}
|
| 24 |
+
```
|
| 25 |
---
|
| 26 |
inference: false
|
| 27 |
license: other
|