This model is a fine-tuned version of saves/qwen25-15b-sft-zh on the qwen25_1_5b_train_dpo_zh dataset. It achieves the following results on the evaluation set:

Loss: 0.3257
Rewards/chosen: -3.1545
Rewards/rejected: -11.5967
Rewards/accuracies: 0.8938
Rewards/margins: 8.4422
Logps/chosen: -219.6885
Logps/rejected: -167.2860
Logits/chosen: -2.7822
Logits/rejected: -2.7818

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: 1
eval_batch_size: 1
seed: 42
distributed_type: multi-GPU
num_devices: 2
gradient_accumulation_steps: 4
total_train_batch_size: 8
total_eval_batch_size: 2
optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED 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: 1.0

Training results

Training Loss	Epoch	Step	Validation Loss	Rewards/chosen	Rewards/rejected	Rewards/accuracies	Rewards/margins	Logps/chosen	Logps/rejected	Logits/chosen	Logits/rejected
0.1297	0.3936	3000	0.2994	-1.5824	-7.7299	0.8784	6.1475	-203.9668	-128.6177	-2.8263	-2.8328
0.1655	0.7873	6000	0.3255	-3.4943	-11.8822	0.8932	8.3879	-223.0866	-170.1410	-2.7808	-2.7804

Framework versions

Transformers 4.57.1
Pytorch 2.11.0+cu126
Datasets 4.0.0
Tokenizers 0.22.2

saiteja33/qwen25-15b-dpo-zh

qwen25-15b-dpo-zh

Model description

Intended uses & limitations

Training and evaluation data

Training procedure

Training hyperparameters

Training results

Framework versions

saiteja33/
qwen25-15b-dpo-zh