cosmosDPO_CodeTest
This model is a fine-tuned version of ytu-ce-cosmos/turkish-gpt2-large-750m-instruct-v0.1 on an unknown dataset. It achieves the following results on the evaluation set:
- Loss: 0.5271
- Rewards/chosen: -2.6242
- Rewards/rejected: -6.3552
- Rewards/accuracies: 0.2667
- Rewards/margins: 3.7309
- Logps/rejected: -749.125
- Logps/chosen: -350.9360
- Logits/rejected: -5.2606
- Logits/chosen: -4.5085
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: 5e-06
- train_batch_size: 32
- eval_batch_size: 8
- seed: 42
- gradient_accumulation_steps: 4
- total_train_batch_size: 128
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: cosine
- lr_scheduler_warmup_ratio: 0.1
- num_epochs: 5
- mixed_precision_training: Native AMP
Training results
Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen |
---|---|---|---|---|---|---|---|---|---|---|---|
0.5159 | 1.3072 | 100 | 0.5241 | -0.9182 | -3.2061 | 0.2676 | 2.2879 | -434.2115 | -180.3287 | -4.0572 | -3.5729 |
0.5227 | 2.6144 | 200 | 0.5217 | -2.1076 | -5.3791 | 0.2695 | 3.2715 | -651.5153 | -299.2687 | -4.8098 | -4.1931 |
0.4937 | 3.9216 | 300 | 0.5271 | -2.6242 | -6.3552 | 0.2667 | 3.7309 | -749.125 | -350.9360 | -5.2606 | -4.5085 |
Framework versions
- PEFT 0.10.0
- Transformers 4.40.1
- Pytorch 2.2.1+cu121
- Datasets 2.19.0
- Tokenizers 0.19.1
- Downloads last month
- 0
Unable to determine this model’s pipeline type. Check the
docs
.