---
library_name: transformers
license: apache-2.0
base_model: jackf857/qwen3-8b-base-sft-hh-harmless-4xh200-batch-64-20260417-214452
tags:
- alignment-handbook
- new-dpo
- generated_from_trainer
datasets:
- Anthropic/hh-rlhf
model-index:
- name: qwen3-8b-base-new-dpo-hh-harmless-4xh200-batch-64-q_t-0.45-s_star-0.4
  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. -->

# qwen3-8b-base-new-dpo-hh-harmless-4xh200-batch-64-q_t-0.45-s_star-0.4

This model is a fine-tuned version of [jackf857/qwen3-8b-base-sft-hh-harmless-4xh200-batch-64-20260417-214452](https://huggingface.co/jackf857/qwen3-8b-base-sft-hh-harmless-4xh200-batch-64-20260417-214452) on the Anthropic/hh-rlhf dataset.
It achieves the following results on the evaluation set:
- Loss: 0.5698
- Fcm Dpo/beta: 0.0074
- Margin Dpo/margin Mean: 51.5411
- Margin Dpo/margin Std: 86.1362
- Logps/chosen: -172.7695
- Logps/rejected: -234.1051
- Logps/ref Chosen: -86.9018
- Logps/ref Rejected: -96.6964
- Logits/chosen: 1.3822
- Logits/rejected: 1.2498

## 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 | Margin Dpo/margin Mean | Margin Dpo/margin Std | Logps/chosen | Logps/rejected | Logps/ref Chosen | Logps/ref Rejected | Logits/chosen | Logits/rejected |
|:-------------:|:------:|:----:|:---------------:|:------------:|:----------------------:|:---------------------:|:------------:|:--------------:|:----------------:|:------------------:|:-------------:|:---------------:|
| 1.3228        | 0.1512 | 100  | 0.6546          | 0.1000       | 0.8918                 | 1.9761                | -86.2353     | -96.9216       | -86.9018         | -96.6964           | 1.6911        | 1.5774          |
| 1.1508        | 0.3023 | 200  | 0.5664          | 0.0827       | 5.5014                 | 9.7680                | -86.8311     | -102.1271      | -86.9018         | -96.6964           | 1.6408        | 1.5124          |
| 1.1662        | 0.4535 | 300  | 0.5601          | 0.0232       | 19.9301                | 33.9240               | -106.0360    | -135.7607      | -86.9018         | -96.6964           | 1.3810        | 1.2532          |
| 1.2381        | 0.6047 | 400  | 0.5746          | 0.0092       | 39.8591                | 68.7826               | -144.7731    | -194.4268      | -86.9018         | -96.6964           | 1.7610        | 1.6208          |
| 1.1309        | 0.7559 | 500  | 0.5676          | 0.0079       | 51.0216                | 86.4987               | -169.9242    | -230.7403      | -86.9018         | -96.6964           | 1.5455        | 1.4084          |
| 1.183         | 0.9070 | 600  | 0.5698          | 0.0074       | 51.5411                | 86.1362               | -172.7695    | -234.1051      | -86.9018         | -96.6964           | 1.3822        | 1.2498          |


### Framework versions

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