Na_M2_1000steps_1e7rate_03beta_cSFTDPO
This model is a fine-tuned version of tsavage68/Na_M2_1000steps_1e7_SFT on an unknown dataset. It achieves the following results on the evaluation set:
- Loss: 0.0000
- Rewards/chosen: 3.5522
- Rewards/rejected: -11.1817
- Rewards/accuracies: 1.0
- Rewards/margins: 14.7339
- Logps/rejected: -117.1956
- Logps/chosen: -36.2917
- Logits/rejected: -2.4993
- Logits/chosen: -2.5149
Model description
More information needed
Intended uses & limitations
More information needed
Training and evaluation data
More information needed
Training procedure
Training hyperparameters
The following hyperparameters were used during training:
- learning_rate: 1e-07
- train_batch_size: 2
- eval_batch_size: 1
- seed: 42
- gradient_accumulation_steps: 2
- total_train_batch_size: 4
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: cosine
- lr_scheduler_warmup_steps: 100
- training_steps: 1000
Training results
| Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen |
|---|---|---|---|---|---|---|---|---|---|---|---|
| 0.0004 | 0.2667 | 50 | 0.0000 | 2.3360 | -8.0685 | 1.0 | 10.4045 | -106.8185 | -40.3458 | -2.5169 | -2.5309 |
| 0.0 | 0.5333 | 100 | 0.0000 | 2.8383 | -9.2541 | 1.0 | 12.0924 | -110.7703 | -38.6713 | -2.5103 | -2.5250 |
| 0.0 | 0.8 | 150 | 0.0000 | 3.0920 | -9.7555 | 1.0 | 12.8475 | -112.4418 | -37.8257 | -2.5067 | -2.5217 |
| 0.0 | 1.0667 | 200 | 0.0000 | 3.2037 | -10.0769 | 1.0 | 13.2806 | -113.5130 | -37.4533 | -2.5051 | -2.5202 |
| 0.0 | 1.3333 | 250 | 0.0000 | 3.2784 | -10.3241 | 1.0 | 13.6025 | -114.3372 | -37.2044 | -2.5046 | -2.5198 |
| 0.0 | 1.6 | 300 | 0.0000 | 3.3562 | -10.5498 | 1.0 | 13.9060 | -115.0894 | -36.9450 | -2.5033 | -2.5186 |
| 0.0 | 1.8667 | 350 | 0.0000 | 3.4141 | -10.7123 | 1.0 | 14.1265 | -115.6312 | -36.7520 | -2.5019 | -2.5173 |
| 0.0 | 2.1333 | 400 | 0.0000 | 3.4694 | -10.8608 | 1.0 | 14.3302 | -116.1261 | -36.5679 | -2.5020 | -2.5174 |
| 0.0 | 2.4 | 450 | 0.0000 | 3.4912 | -10.9759 | 1.0 | 14.4671 | -116.5096 | -36.4950 | -2.5011 | -2.5165 |
| 0.0 | 2.6667 | 500 | 0.0000 | 3.5172 | -11.0415 | 1.0 | 14.5587 | -116.7282 | -36.4083 | -2.5010 | -2.5165 |
| 0.0 | 2.9333 | 550 | 0.0000 | 3.5281 | -11.1219 | 1.0 | 14.6500 | -116.9964 | -36.3719 | -2.4999 | -2.5154 |
| 0.0 | 3.2 | 600 | 0.0000 | 3.5544 | -11.1376 | 1.0 | 14.6920 | -117.0486 | -36.2843 | -2.4985 | -2.5140 |
| 0.0 | 3.4667 | 650 | 0.0000 | 3.5412 | -11.1686 | 1.0 | 14.7098 | -117.1519 | -36.3284 | -2.4993 | -2.5149 |
| 0.0 | 3.7333 | 700 | 0.0000 | 3.5592 | -11.1405 | 1.0 | 14.6997 | -117.0585 | -36.2685 | -2.4988 | -2.5143 |
| 0.0 | 4.0 | 750 | 0.0000 | 3.5602 | -11.1575 | 1.0 | 14.7177 | -117.1151 | -36.2652 | -2.4993 | -2.5148 |
| 0.0 | 4.2667 | 800 | 0.0000 | 3.5534 | -11.1617 | 1.0 | 14.7151 | -117.1290 | -36.2877 | -2.4996 | -2.5152 |
| 0.0 | 4.5333 | 850 | 0.0000 | 3.5623 | -11.1612 | 1.0 | 14.7234 | -117.1272 | -36.2582 | -2.4994 | -2.5150 |
| 0.0 | 4.8 | 900 | 0.0000 | 3.5522 | -11.1817 | 1.0 | 14.7339 | -117.1956 | -36.2917 | -2.4993 | -2.5149 |
| 0.0 | 5.0667 | 950 | 0.0000 | 3.5522 | -11.1817 | 1.0 | 14.7339 | -117.1956 | -36.2917 | -2.4993 | -2.5149 |
| 0.0 | 5.3333 | 1000 | 0.0000 | 3.5522 | -11.1817 | 1.0 | 14.7339 | -117.1956 | -36.2917 | -2.4993 | -2.5149 |
Framework versions
- Transformers 4.44.2
- Pytorch 2.4.0+cu121
- Datasets 2.21.0
- Tokenizers 0.19.1
- Downloads last month
- 16
This model does not have enough activity to be deployed to Inference API (serverless) yet. Increase its social
visibility and check back later, or deploy to Inference Endpoints (dedicated)
instead.
Model tree for tsavage68/Na_M2_1000steps_1e7rate_03beta_cSFTDPO
Base model
mistralai/Mistral-7B-Instruct-v0.2
Finetuned
tsavage68/Na_M2_1000steps_1e7_SFT