metadata
license: mit
library_name: peft
tags:
- trl
- kto
- KTO
- WeniGPT
- generated_from_trainer
base_model: HuggingFaceH4/zephyr-7b-beta
model-index:
- name: WeniGPT-Agents-Zephyr-1.0.1-KTO
results: []
WeniGPT-Agents-Zephyr-1.0.1-KTO
This model is a fine-tuned version of HuggingFaceH4/zephyr-7b-beta on the None dataset. It achieves the following results on the evaluation set:
- Loss: 0.3754
- Rewards/chosen: -1.3231
- Logps/chosen: -330.0658
- Kl: 1.3236
- Rewards/rejected: -6.3958
- Logps/rejected: -357.2245
- Rewards/margins: 5.1890
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: 0.0002
- train_batch_size: 4
- eval_batch_size: 4
- seed: 42
- gradient_accumulation_steps: 4
- total_train_batch_size: 16
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_ratio: 0.03
- training_steps: 145
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Logps/chosen | Kl | Rewards/rejected | Logps/rejected | Rewards/margins |
|---|---|---|---|---|---|---|---|---|---|
| 0.4315 | 0.34 | 50 | 0.4092 | 5.2045 | -264.7897 | 32.9343 | 2.5655 | -267.6116 | 2.7203 |
| 0.3678 | 0.68 | 100 | 0.3754 | -1.3231 | -330.0658 | 1.3236 | -6.3958 | -357.2245 | 5.1890 |
Framework versions
- PEFT 0.10.0
- Transformers 4.39.1
- Pytorch 2.1.0+cu118
- Datasets 2.18.0
- Tokenizers 0.15.2