metadata
license: llama3
base_model: tsavage68/UTI_L3_1000steps_1e5rate_SFT
tags:
- trl
- dpo
- generated_from_trainer
model-index:
- name: UTI_L3_100steps_1e5rate_01beta_CSFTDPO
results: []
UTI_L3_100steps_1e5rate_01beta_CSFTDPO
This model is a fine-tuned version of tsavage68/UTI_L3_1000steps_1e5rate_SFT on an unknown dataset. It achieves the following results on the evaluation set:
- Loss: 0.0086
- Rewards/chosen: -1.2523
- Rewards/rejected: -14.7999
- Rewards/accuracies: 0.9900
- Rewards/margins: 13.5476
- Logps/rejected: -211.1941
- Logps/chosen: -45.0020
- Logits/rejected: -1.7169
- Logits/chosen: -1.6585
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-05
- 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: 100
Training results
| Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen |
|---|---|---|---|---|---|---|---|---|---|---|---|
| 0.0002 | 0.6667 | 50 | 0.0072 | 0.7766 | -13.0912 | 0.9900 | 13.8678 | -194.1069 | -24.7133 | -1.5529 | -1.4952 |
| 0.0173 | 1.3333 | 100 | 0.0086 | -1.2523 | -14.7999 | 0.9900 | 13.5476 | -211.1941 | -45.0020 | -1.7169 | -1.6585 |
Framework versions
- Transformers 4.41.2
- Pytorch 2.0.0+cu117
- Datasets 2.19.2
- Tokenizers 0.19.1