metadata
library_name: transformers
license: apache-2.0
base_model: alignment-handbook/zephyr-7b-sft-full
tags:
- alignment-handbook
- trl
- dpo
- generated_from_trainer
- trl
- dpo
- generated_from_trainer
datasets:
- HuggingFaceH4/ultrafeedback_binarized
model-index:
- name: zephyr-7b-align-scan-2e-07-0.39-polynomial-1.0
results: []
zephyr-7b-align-scan-2e-07-0.39-polynomial-1.0
This model is a fine-tuned version of alignment-handbook/zephyr-7b-sft-full on the HuggingFaceH4/ultrafeedback_binarized dataset. It achieves the following results on the evaluation set:
- Loss: 0.6326
- Rewards/chosen: 0.3567
- Rewards/rejected: -0.0406
- Rewards/accuracies: 0.3492
- Rewards/margins: 0.3973
- Logps/rejected: -81.2324
- Logps/chosen: -73.5766
- Logits/rejected: -2.5151
- Logits/chosen: -2.5314
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: 2e-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: polynomial
- 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 |
|---|---|---|---|---|---|---|---|---|---|---|---|
| 0.6538 | 0.3484 | 100 | 0.6320 | 0.5810 | 0.2964 | 0.3472 | 0.2846 | -80.3683 | -73.0014 | -2.5308 | -2.5469 |
| 0.6462 | 0.6969 | 200 | 0.6324 | 0.2521 | -0.1128 | 0.3492 | 0.3648 | -81.4175 | -73.8449 | -2.5055 | -2.5222 |
Framework versions
- Transformers 4.44.2
- Pytorch 2.4.0
- Datasets 2.21.0
- Tokenizers 0.19.1