formalmathatepfl/deepseek-math-7B-finetuned

deepseek-math-lean-sft

This model is a fine-tuned version of deepseek-ai/deepseek-math-7b-base on the lean_sft dataset. It achieves the following results on the evaluation set:

• Loss: 0.0597

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-05
• train_batch_size: 2
• eval_batch_size: 2
• seed: 42
• distributed_type: multi-GPU
• num_devices: 8
• gradient_accumulation_steps: 2
• total_train_batch_size: 32
• total_eval_batch_size: 16
• 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: 0.05
• num_epochs: 1.0

Training results

Training Loss	Epoch	Step	Validation Loss
0.0744	0.1692	1000	0.0767
0.068	0.3384	2000	0.0688
0.0637	0.5077	3000	0.0648
0.0601	0.6769	4000	0.0620
0.0608	0.8461	5000	0.0601

Framework versions

• Transformers 4.57.3
• Pytorch 2.9.0+cu129
• Datasets 4.0.0
• Tokenizers 0.22.1