math_ft

This model is a fine-tuned version of PrunaAI/deepseek-ai-deepseek-math-7b-rl-bnb-4bit-smashed on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.5111

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: 0.0002
• train_batch_size: 32
• eval_batch_size: 32
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 128
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 2
• num_epochs: 10
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss
1.0364	0.9655	7	0.7955
0.7271	1.9310	14	0.6199
0.604	2.8966	21	0.5675
0.4951	4.0	29	0.5419
0.5483	4.9655	36	0.5294
0.531	5.9310	43	0.5218
0.5217	6.8966	50	0.5168
0.4511	8.0	58	0.5131
0.5083	8.9655	65	0.5115
0.4632	9.6552	70	0.5111

Framework versions

• PEFT 0.11.1
• Transformers 4.40.2
• Pytorch 2.1.0+cu121
• Datasets 2.19.1
• Tokenizers 0.19.1