Commit
•
714fa24
1
Parent(s):
49e8900
ajustes
Browse files
README.md
CHANGED
|
@@ -29,7 +29,24 @@ Details:
|
|
| 29 |
```
|
| 30 |
3 epochs, all dataset samples (split=train), 939 steps
|
| 31 |
1 x GPU NVidia RTX 3060 12GB - max. GPU memory: 7.44 GB
|
| 32 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 33 |
```
|
| 34 |
|
| 35 |
## Inference
|
|
@@ -78,3 +95,6 @@ pip install -q -U scipy
|
|
| 78 |
## Scripts
|
| 79 |
[https://github.com/nlpulse-io/sample_codes/tree/main/fine-tuning/peft_quantization_4bits/gptj-6b](https://github.com/nlpulse-io/sample_codes/tree/main/fine-tuning/peft_quantization_4bits/gptj-6b)
|
| 80 |
|
|
|
|
|
|
|
|
|
|
|
|
| 29 |
```
|
| 30 |
3 epochs, all dataset samples (split=train), 939 steps
|
| 31 |
1 x GPU NVidia RTX 3060 12GB - max. GPU memory: 7.44 GB
|
| 32 |
+
Duration: 1h45min
|
| 33 |
+
|
| 34 |
+
$ nvidia-smi && free -h
|
| 35 |
+
+-----------------------------------------------------------------------------+
|
| 36 |
+
| NVIDIA-SMI 515.105.01 Driver Version: 515.105.01 CUDA Version: 11.7 |
|
| 37 |
+
|-------------------------------+----------------------+----------------------+
|
| 38 |
+
| GPU Name Persistence-M| Bus-Id Disp.A | Volatile Uncorr. ECC |
|
| 39 |
+
| Fan Temp Perf Pwr:Usage/Cap| Memory-Usage | GPU-Util Compute M. |
|
| 40 |
+
| | | MIG M. |
|
| 41 |
+
|===============================+======================+======================|
|
| 42 |
+
| 1 NVIDIA GeForce ... Off | 00000000:04:00.0 Off | N/A |
|
| 43 |
+
|100% 89C P2 166W / 170W | 7439MiB / 12288MiB | 93% Default |
|
| 44 |
+
| | | N/A |
|
| 45 |
+
+-------------------------------+----------------------+----------------------+
|
| 46 |
+
total used free shared buff/cache available
|
| 47 |
+
Mem: 77Gi 14Gi 23Gi 79Mi 39Gi 62Gi
|
| 48 |
+
Swap: 37Gi 0B 37Gi
|
| 49 |
+
|
| 50 |
```
|
| 51 |
|
| 52 |
## Inference
|
|
|
|
| 95 |
## Scripts
|
| 96 |
[https://github.com/nlpulse-io/sample_codes/tree/main/fine-tuning/peft_quantization_4bits/gptj-6b](https://github.com/nlpulse-io/sample_codes/tree/main/fine-tuning/peft_quantization_4bits/gptj-6b)
|
| 97 |
|
| 98 |
+
|
| 99 |
+
# References
|
| 100 |
+
[https://towardsdatascience.com/qlora-fine-tune-a-large-language-model-on-your-gpu-27bed5a03e2b](https://towardsdatascience.com/qlora-fine-tune-a-large-language-model-on-your-gpu-27bed5a03e2b)
|