qwen2.5-1.5b-diffusion-scientific-L

This model is a fine-tuned version of lamm-mit/qwen2.5-1.5b-diffusion-ultrachat on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 3.6680

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: 1e-05
• train_batch_size: 2
• eval_batch_size: 2
• seed: 42
• gradient_accumulation_steps: 16
• total_train_batch_size: 32
• 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: linear
• lr_scheduler_warmup_steps: 0.03
• num_epochs: 3.0

Training results

Training Loss	Epoch	Step	Validation Loss
64.1562	0.3556	100	3.7350
64.3663	0.7111	200	3.8210
59.7814	1.064	300	3.9045
63.2812	1.4196	400	3.8106
60.3536	1.7751	500	3.9681
61.2504	2.128	600	3.7179
61.4287	2.4836	700	3.5746
57.6944	2.8391	800	3.8365
59.3825	3.0	846	3.6680

Framework versions

• Transformers 5.14.1
• Pytorch 2.13.0+cu130
• Datasets 4.8.5
• Tokenizers 0.22.2