NoManDeRY/DPO-Shift-Qwen-2-7B-UltraChat200K-SFT

qwen-2-7b-paged_adamw_32bit

This is a model released from the preprint: DPO-Shift: Shifting the Distribution of Direct Preference Optimization. Please refer to our repository for more details.

This model is a fine-tuned version of Qwen/Qwen2-7B on the ultrachat_200k_train dataset. It achieves the following results on the evaluation set:

• Loss: 0.8906

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-05
• train_batch_size: 4
• eval_batch_size: 8
• seed: 42
• distributed_type: multi-GPU
• num_devices: 8
• gradient_accumulation_steps: 4
• total_train_batch_size: 128
• total_eval_batch_size: 64
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 1.0

Training results

Training Loss	Epoch	Step	Validation Loss
0.9206	0.1232	200	0.9238
0.9521	0.2463	400	0.9254
0.9654	0.3695	600	0.9204
0.9188	0.4926	800	0.9126
0.967	0.6158	1000	0.9037
0.8783	0.7389	1200	0.8964
0.8915	0.8621	1400	0.8918
0.9246	0.9852	1600	0.8906

Framework versions

• Transformers 4.44.2
• Pytorch 2.4.0+cu121
• Datasets 2.21.0
• Tokenizers 0.19.1