metadata
base_model: meta-llama/Llama-2-7b-hf
tags:
- alignment-handbook
- trl
- dpo
- generated_from_trainer
- trl
- dpo
- generated_from_trainer
datasets:
- HuggingFaceH4/ultrafeedback_binarized
model-index:
- name: llama2-7b-dpo-full-wo-medication_qa-ep3
results: []
llama2-7b-dpo-full-wo-medication_qa-ep3
This model is a fine-tuned version of meta-llama/Llama-2-7b-hf on the HuggingFaceH4/ultrafeedback_binarized dataset. It achieves the following results on the evaluation set:
- Loss: 0.6486
- Rewards/chosen: 0.0485
- Rewards/rejected: -0.0716
- Rewards/accuracies: 0.7847
- Rewards/margins: 0.1201
- Logps/rejected: -1097.3336
- Logps/chosen: -485.8272
- Logits/rejected: -1.1127
- Logits/chosen: -0.0114
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-07
- train_batch_size: 8
- eval_batch_size: 8
- seed: 42
- distributed_type: multi-GPU
- num_devices: 4
- gradient_accumulation_steps: 2
- total_train_batch_size: 64
- total_eval_batch_size: 32
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: cosine
- lr_scheduler_warmup_ratio: 0.1
- num_epochs: 1
Training results
| Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen |
|---|---|---|---|---|---|---|---|---|---|---|---|
| 0.5991 | 0.7 | 100 | 0.6540 | 0.0513 | -0.0539 | 0.7778 | 0.1052 | -1095.5646 | -485.5474 | -1.1199 | -0.0017 |
Framework versions
- Transformers 4.39.0.dev0
- Pytorch 2.1.2
- Datasets 2.14.6
- Tokenizers 0.15.2