---
library_name: transformers
base_model: W-61/llama-3-8b-base-sft-hh-harmless-4xh200
tags:
- alignment-handbook
- new-dpo
- generated_from_trainer
datasets:
- Anthropic/hh-rlhf
model-index:
- name: llama-3-8b-base-new-dpo-hh-harmless-s_star1.0-4xh200-batch-64-20260421-213851
  results: []
---

<!-- This model card has been generated automatically according to the information the Trainer had access to. You
should probably proofread and complete it, then remove this comment. -->

# llama-3-8b-base-new-dpo-hh-harmless-s_star1.0-4xh200-batch-64-20260421-213851

This model is a fine-tuned version of [W-61/llama-3-8b-base-sft-hh-harmless-4xh200](https://huggingface.co/W-61/llama-3-8b-base-sft-hh-harmless-4xh200) on the Anthropic/hh-rlhf dataset.
It achieves the following results on the evaluation set:
- Loss: 0.5467
- Fcm Dpo/beta: 0.2268
- Fcm Dpo/q T: 0.3412
- Fcm Dpo/delta: -0.0017
- Fcm Dpo/margin: 4.4089
- Margin Dpo/margin Mean: 4.4089
- Margin Dpo/margin Std: 7.2504
- Logps/chosen: -82.5570
- Logps/rejected: -91.6554
- Logps/ref Chosen: -74.8595
- Logps/ref Rejected: -79.5490
- Logits/chosen: 0.2724
- Logits/rejected: 0.2290

## 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: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
- lr_scheduler_type: cosine
- lr_scheduler_warmup_ratio: 0.1
- num_epochs: 1

### Training results

| Training Loss | Epoch  | Step | Validation Loss | Fcm Dpo/beta | Fcm Dpo/q T | Fcm Dpo/delta | Fcm Dpo/margin | Margin Dpo/margin Mean | Margin Dpo/margin Std | Logps/chosen | Logps/rejected | Logps/ref Chosen | Logps/ref Rejected | Logits/chosen | Logits/rejected |
|:-------------:|:------:|:----:|:---------------:|:------------:|:-----------:|:-------------:|:--------------:|:----------------------:|:---------------------:|:------------:|:--------------:|:----------------:|:------------------:|:-------------:|:---------------:|
| 0.9878        | 0.3023 | 200  | 0.5709          | 0.4418       | 0.3475      | 0.0198        | 2.2173         | 2.2173                 | 3.8804                | -79.5152     | -86.4220       | -74.8595         | -79.5490           | 0.2313        | 0.1912          |
| 0.966         | 0.6047 | 400  | 0.5573          | 0.2747       | 0.3451      | 0.0194        | 3.5687         | 3.5687                 | 6.0461                | -81.6850     | -89.9432       | -74.8595         | -79.5490           | 0.2557        | 0.2135          |
| 1.122         | 0.9070 | 600  | 0.5467          | 0.2268       | 0.3412      | -0.0017       | 4.4089         | 4.4089                 | 7.2504                | -82.5570     | -91.6554       | -74.8595         | -79.5490           | 0.2724        | 0.2290          |


### Framework versions

- Transformers 4.51.0
- Pytorch 2.3.1+cu121
- Datasets 2.21.0
- Tokenizers 0.21.4