End of training
Browse files
README.md
CHANGED
|
@@ -5,6 +5,8 @@ tags:
|
|
| 5 |
- trl
|
| 6 |
- reward-trainer
|
| 7 |
- generated_from_trainer
|
|
|
|
|
|
|
| 8 |
base_model: AI-Sweden-Models/gpt-sw3-1.3b
|
| 9 |
model-index:
|
| 10 |
- name: gpt1B_reward_model2
|
|
@@ -17,6 +19,9 @@ should probably proofread and complete it, then remove this comment. -->
|
|
| 17 |
# gpt1B_reward_model2
|
| 18 |
|
| 19 |
This model is a fine-tuned version of [AI-Sweden-Models/gpt-sw3-1.3b](https://huggingface.co/AI-Sweden-Models/gpt-sw3-1.3b) on an unknown dataset.
|
|
|
|
|
|
|
|
|
|
| 20 |
|
| 21 |
## Model description
|
| 22 |
|
|
@@ -39,14 +44,32 @@ The following hyperparameters were used during training:
|
|
| 39 |
- train_batch_size: 1
|
| 40 |
- eval_batch_size: 1
|
| 41 |
- seed: 42
|
| 42 |
-
- gradient_accumulation_steps: 8
|
| 43 |
-
- total_train_batch_size: 8
|
| 44 |
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 45 |
- lr_scheduler_type: linear
|
| 46 |
-
- num_epochs:
|
| 47 |
|
| 48 |
### Training results
|
| 49 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 50 |
|
| 51 |
|
| 52 |
### Framework versions
|
|
|
|
| 5 |
- trl
|
| 6 |
- reward-trainer
|
| 7 |
- generated_from_trainer
|
| 8 |
+
metrics:
|
| 9 |
+
- accuracy
|
| 10 |
base_model: AI-Sweden-Models/gpt-sw3-1.3b
|
| 11 |
model-index:
|
| 12 |
- name: gpt1B_reward_model2
|
|
|
|
| 19 |
# gpt1B_reward_model2
|
| 20 |
|
| 21 |
This model is a fine-tuned version of [AI-Sweden-Models/gpt-sw3-1.3b](https://huggingface.co/AI-Sweden-Models/gpt-sw3-1.3b) on an unknown dataset.
|
| 22 |
+
It achieves the following results on the evaluation set:
|
| 23 |
+
- Loss: 0.0000
|
| 24 |
+
- Accuracy: 1.0
|
| 25 |
|
| 26 |
## Model description
|
| 27 |
|
|
|
|
| 44 |
- train_batch_size: 1
|
| 45 |
- eval_batch_size: 1
|
| 46 |
- seed: 42
|
|
|
|
|
|
|
| 47 |
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 48 |
- lr_scheduler_type: linear
|
| 49 |
+
- num_epochs: 2
|
| 50 |
|
| 51 |
### Training results
|
| 52 |
|
| 53 |
+
| Training Loss | Epoch | Step | Validation Loss | Accuracy |
|
| 54 |
+
|:-------------:|:-----:|:----:|:---------------:|:--------:|
|
| 55 |
+
| 0.0 | 0.11 | 200 | 0.0124 | 0.9930 |
|
| 56 |
+
| 0.0002 | 0.22 | 400 | 0.0034 | 0.9965 |
|
| 57 |
+
| 0.0 | 0.33 | 600 | 0.0003 | 1.0 |
|
| 58 |
+
| 0.0 | 0.44 | 800 | 0.0003 | 1.0 |
|
| 59 |
+
| 0.0 | 0.55 | 1000 | 0.0003 | 1.0 |
|
| 60 |
+
| 0.0 | 0.65 | 1200 | 0.0004 | 1.0 |
|
| 61 |
+
| 0.0 | 0.76 | 1400 | 0.0000 | 1.0 |
|
| 62 |
+
| 0.0 | 0.87 | 1600 | 0.0000 | 1.0 |
|
| 63 |
+
| 0.0 | 0.98 | 1800 | 0.0000 | 1.0 |
|
| 64 |
+
| 0.0 | 1.09 | 2000 | 0.0000 | 1.0 |
|
| 65 |
+
| 0.0 | 1.2 | 2200 | 0.0000 | 1.0 |
|
| 66 |
+
| 0.0001 | 1.31 | 2400 | 0.0000 | 1.0 |
|
| 67 |
+
| 0.0 | 1.42 | 2600 | 0.0000 | 1.0 |
|
| 68 |
+
| 0.0 | 1.53 | 2800 | 0.0000 | 1.0 |
|
| 69 |
+
| 0.0 | 1.64 | 3000 | 0.0000 | 1.0 |
|
| 70 |
+
| 0.0 | 1.75 | 3200 | 0.0000 | 1.0 |
|
| 71 |
+
| 0.0 | 1.85 | 3400 | 0.0000 | 1.0 |
|
| 72 |
+
| 0.0 | 1.96 | 3600 | 0.0000 | 1.0 |
|
| 73 |
|
| 74 |
|
| 75 |
### Framework versions
|