Commit
·
ebd1fe6
1
Parent(s):
63e63e9
Upload model
Browse files
README.md
CHANGED
|
@@ -273,4 +273,42 @@ The following `bitsandbytes` quantization config was used during training:
|
|
| 273 |
### Framework versions
|
| 274 |
|
| 275 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 276 |
- PEFT 0.6.0.dev0
|
|
|
|
| 273 |
### Framework versions
|
| 274 |
|
| 275 |
|
| 276 |
+
- PEFT 0.6.0.dev0
|
| 277 |
+
## Training procedure
|
| 278 |
+
|
| 279 |
+
|
| 280 |
+
The following `bitsandbytes` quantization config was used during training:
|
| 281 |
+
- quant_method: bitsandbytes
|
| 282 |
+
- load_in_8bit: False
|
| 283 |
+
- load_in_4bit: True
|
| 284 |
+
- llm_int8_threshold: 6.0
|
| 285 |
+
- llm_int8_skip_modules: None
|
| 286 |
+
- llm_int8_enable_fp32_cpu_offload: False
|
| 287 |
+
- llm_int8_has_fp16_weight: False
|
| 288 |
+
- bnb_4bit_quant_type: nf4
|
| 289 |
+
- bnb_4bit_use_double_quant: True
|
| 290 |
+
- bnb_4bit_compute_dtype: bfloat16
|
| 291 |
+
|
| 292 |
+
### Framework versions
|
| 293 |
+
|
| 294 |
+
|
| 295 |
+
- PEFT 0.6.0.dev0
|
| 296 |
+
## Training procedure
|
| 297 |
+
|
| 298 |
+
|
| 299 |
+
The following `bitsandbytes` quantization config was used during training:
|
| 300 |
+
- quant_method: bitsandbytes
|
| 301 |
+
- load_in_8bit: False
|
| 302 |
+
- load_in_4bit: True
|
| 303 |
+
- llm_int8_threshold: 6.0
|
| 304 |
+
- llm_int8_skip_modules: None
|
| 305 |
+
- llm_int8_enable_fp32_cpu_offload: False
|
| 306 |
+
- llm_int8_has_fp16_weight: False
|
| 307 |
+
- bnb_4bit_quant_type: nf4
|
| 308 |
+
- bnb_4bit_use_double_quant: True
|
| 309 |
+
- bnb_4bit_compute_dtype: bfloat16
|
| 310 |
+
|
| 311 |
+
### Framework versions
|
| 312 |
+
|
| 313 |
+
|
| 314 |
- PEFT 0.6.0.dev0
|