Commit
•
be46e57
1
Parent(s):
d85a7bc
End of training
Browse files- README.md +19 -0
- adapter_model.bin +1 -1
README.md
CHANGED
|
@@ -369,4 +369,23 @@ The following `bitsandbytes` quantization config was used during training:
|
|
| 369 |
### Framework versions
|
| 370 |
|
| 371 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 372 |
- PEFT 0.6.2
|
|
|
|
| 369 |
### Framework versions
|
| 370 |
|
| 371 |
|
| 372 |
+
- PEFT 0.6.2
|
| 373 |
+
## Training procedure
|
| 374 |
+
|
| 375 |
+
|
| 376 |
+
The following `bitsandbytes` quantization config was used during training:
|
| 377 |
+
- quant_method: bitsandbytes
|
| 378 |
+
- load_in_8bit: False
|
| 379 |
+
- load_in_4bit: True
|
| 380 |
+
- llm_int8_threshold: 6.0
|
| 381 |
+
- llm_int8_skip_modules: None
|
| 382 |
+
- llm_int8_enable_fp32_cpu_offload: False
|
| 383 |
+
- llm_int8_has_fp16_weight: False
|
| 384 |
+
- bnb_4bit_quant_type: nf4
|
| 385 |
+
- bnb_4bit_use_double_quant: True
|
| 386 |
+
- bnb_4bit_compute_dtype: bfloat16
|
| 387 |
+
|
| 388 |
+
### Framework versions
|
| 389 |
+
|
| 390 |
+
|
| 391 |
- PEFT 0.6.2
|
adapter_model.bin
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 67155338
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:6e2e296b68522d5e7d0f3e53df5ecb2e6a64b4586c4958cb2d5f9a411fa13720
|
| 3 |
size 67155338
|