qwen2.5-1.5b-diffusion-scientific-L-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: 3.8626

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: 3e-06
• 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: 10.0

Training results

Training Loss	Epoch	Step	Validation Loss
72.8125	0.3556	100	4.2313
70.7423	0.7111	200	4.2203
65.7707	1.064	300	4.2686
68.8484	1.4196	400	4.1462
65.7248	1.7751	500	4.2940
66.4324	2.128	600	4.0263
66.3626	2.4836	700	3.8795
62.3988	2.8391	800	4.1321
65.6526	3.192	900	4.1260
67.4312	3.5476	1000	3.8996
63.8859	3.9031	1100	4.2748
65.7103	4.256	1200	3.8309
63.2910	4.6116	1300	3.9817
62.6077	4.9671	1400	3.9192
62.8082	5.32	1500	3.9552
63.3571	5.6756	1600	4.1320
57.8210	6.0284	1700	4.0993
63.1211	6.384	1800	4.0743
62.5020	6.7396	1900	3.9847
64.6647	7.0924	2000	4.0498
65.7863	7.448	2100	4.0664
65.2337	7.8036	2200	4.0635
62.7605	8.1564	2300	4.1932
64.0694	8.512	2400	3.8387
60.7923	8.8676	2500	3.9723
65.5705	9.2204	2600	4.1179
63.6627	9.576	2700	3.9689
63.9980	9.9316	2800	4.2244
59.2126	10.0	2820	3.8626

Framework versions

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