IliyanGochev
commited on
Commit
•
212dbcc
1
Parent(s):
3b05fa1
End of training
Browse files- README.md +26 -0
- adapter_model.bin +1 -1
- all_results.json +1 -0
README.md
CHANGED
@@ -378,6 +378,30 @@ The following `bitsandbytes` quantization config was used during training:
|
|
378 |
- bnb_4bit_use_double_quant: False
|
379 |
- bnb_4bit_compute_dtype: float32
|
380 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
381 |
The following `bitsandbytes` quantization config was used during training:
|
382 |
- quant_method: bitsandbytes
|
383 |
- load_in_8bit: True
|
@@ -435,6 +459,8 @@ The following `bitsandbytes` quantization config was used during training:
|
|
435 |
- PEFT 0.5.0
|
436 |
- PEFT 0.5.0
|
437 |
- PEFT 0.5.0
|
|
|
|
|
438 |
|
439 |
- PEFT 0.5.0.dev0
|
440 |
`bitsandbytes` quantization config was used during training:
|
|
|
378 |
- bnb_4bit_use_double_quant: False
|
379 |
- bnb_4bit_compute_dtype: float32
|
380 |
|
381 |
+
The following `bitsandbytes` quantization config was used during training:
|
382 |
+
- quant_method: bitsandbytes
|
383 |
+
- load_in_8bit: True
|
384 |
+
- load_in_4bit: False
|
385 |
+
- llm_int8_threshold: 6.0
|
386 |
+
- llm_int8_skip_modules: None
|
387 |
+
- llm_int8_enable_fp32_cpu_offload: False
|
388 |
+
- llm_int8_has_fp16_weight: False
|
389 |
+
- bnb_4bit_quant_type: fp4
|
390 |
+
- bnb_4bit_use_double_quant: False
|
391 |
+
- bnb_4bit_compute_dtype: float32
|
392 |
+
|
393 |
+
The following `bitsandbytes` quantization config was used during training:
|
394 |
+
- quant_method: bitsandbytes
|
395 |
+
- load_in_8bit: True
|
396 |
+
- load_in_4bit: False
|
397 |
+
- llm_int8_threshold: 6.0
|
398 |
+
- llm_int8_skip_modules: None
|
399 |
+
- llm_int8_enable_fp32_cpu_offload: False
|
400 |
+
- llm_int8_has_fp16_weight: False
|
401 |
+
- bnb_4bit_quant_type: fp4
|
402 |
+
- bnb_4bit_use_double_quant: False
|
403 |
+
- bnb_4bit_compute_dtype: float32
|
404 |
+
|
405 |
The following `bitsandbytes` quantization config was used during training:
|
406 |
- quant_method: bitsandbytes
|
407 |
- load_in_8bit: True
|
|
|
459 |
- PEFT 0.5.0
|
460 |
- PEFT 0.5.0
|
461 |
- PEFT 0.5.0
|
462 |
+
- PEFT 0.5.0
|
463 |
+
- PEFT 0.5.0
|
464 |
|
465 |
- PEFT 0.5.0.dev0
|
466 |
`bitsandbytes` quantization config was used during training:
|
adapter_model.bin
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
size 38697637
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:3e5dd6174d8b5524ec2c4a590bbbae741941169a385eba7f6dc02e3eea74d898
|
3 |
size 38697637
|
all_results.json
ADDED
@@ -0,0 +1 @@
|
|
|
|
|
1 |
+
{"eval/wer": 19.44762340733959, "eval/normalized_wer": 13.298817711150607}
|