kto6

This model is a fine-tuned version of deepseek-ai/deepseek-coder-1.3b-base on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.5
• Eval/rewards/chosen: -448.8974
• Eval/logps/chosen: -4526.3562
• Eval/rewards/rejected: -433.3957
• Eval/logps/rejected: -4408.4109
• Eval/rewards/margins: -15.5017
• Eval/kl: 0.0

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: 0.05
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 16
• total_train_batch_size: 128
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• lr_scheduler_warmup_steps: 200
• num_epochs: 10
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss
0.5029	1.1552	100	0.5028	0.0
0.502	2.3105	200	0.5000	0.0
0.4998	3.4657	300	0.5000	0.0
0.5002	4.6209	400	0.5000	0.0
0.501	5.7762	500	0.5000	0.0
0.4994	6.9314	600	0.5000	0.0
0.4998	8.0866	700	0.5000	0.0
0.501	9.2419	800	0.5	0.0

Framework versions

• Transformers 4.45.0
• Pytorch 2.5.1+cu124
• Datasets 2.19.2
• Tokenizers 0.20.3