Model save
Browse files
README.md
CHANGED
@@ -19,7 +19,7 @@ should probably proofread and complete it, then remove this comment. -->
|
|
19 |
|
20 |
This model is a fine-tuned version of [ytcheng/llama-3-8b-hf-sm-lora-merged](https://huggingface.co/ytcheng/llama-3-8b-hf-sm-lora-merged) on the generator dataset.
|
21 |
It achieves the following results on the evaluation set:
|
22 |
-
- Loss:
|
23 |
|
24 |
## Model description
|
25 |
|
@@ -39,29 +39,25 @@ More information needed
|
|
39 |
|
40 |
The following hyperparameters were used during training:
|
41 |
- learning_rate: 0.0001
|
42 |
-
- train_batch_size:
|
43 |
-
- eval_batch_size:
|
44 |
- seed: 42
|
45 |
- distributed_type: multi-GPU
|
46 |
- gradient_accumulation_steps: 2
|
47 |
-
- total_train_batch_size:
|
48 |
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
49 |
- lr_scheduler_type: cosine
|
50 |
- lr_scheduler_warmup_ratio: 0.3
|
51 |
-
- num_epochs:
|
52 |
|
53 |
### Training results
|
54 |
|
55 |
| Training Loss | Epoch | Step | Validation Loss |
|
56 |
|:-------------:|:------:|:----:|:---------------:|
|
57 |
-
|
|
58 |
-
|
|
59 |
-
|
|
60 |
-
|
|
61 |
-
| 1.9633 | 4.9851 | 167 | 3.0894 |
|
62 |
-
| 1.8968 | 6.0 | 201 | 3.1514 |
|
63 |
-
| 1.8985 | 6.9851 | 234 | 3.1113 |
|
64 |
-
| 1.8886 | 7.8806 | 264 | 3.0952 |
|
65 |
|
66 |
|
67 |
### Framework versions
|
|
|
19 |
|
20 |
This model is a fine-tuned version of [ytcheng/llama-3-8b-hf-sm-lora-merged](https://huggingface.co/ytcheng/llama-3-8b-hf-sm-lora-merged) on the generator dataset.
|
21 |
It achieves the following results on the evaluation set:
|
22 |
+
- Loss: 4.8363
|
23 |
|
24 |
## Model description
|
25 |
|
|
|
39 |
|
40 |
The following hyperparameters were used during training:
|
41 |
- learning_rate: 0.0001
|
42 |
+
- train_batch_size: 8
|
43 |
+
- eval_batch_size: 8
|
44 |
- seed: 42
|
45 |
- distributed_type: multi-GPU
|
46 |
- gradient_accumulation_steps: 2
|
47 |
+
- total_train_batch_size: 16
|
48 |
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
49 |
- lr_scheduler_type: cosine
|
50 |
- lr_scheduler_warmup_ratio: 0.3
|
51 |
+
- num_epochs: 4
|
52 |
|
53 |
### Training results
|
54 |
|
55 |
| Training Loss | Epoch | Step | Validation Loss |
|
56 |
|:-------------:|:------:|:----:|:---------------:|
|
57 |
+
| 2.3345 | 0.9963 | 133 | 2.4116 |
|
58 |
+
| 1.9916 | 2.0 | 267 | 3.9850 |
|
59 |
+
| 1.8552 | 2.9963 | 400 | 4.8651 |
|
60 |
+
| 1.8627 | 3.9850 | 532 | 4.8363 |
|
|
|
|
|
|
|
|
|
61 |
|
62 |
|
63 |
### Framework versions
|