ymoslem/qwen-audio-en-de-ft-decoder-32-24layers-iter-pruning-chrf-4bs-1e5lr-cosine-3epoch-acl-kd1-50

Qwen2-Audio-7B EN-DE Speech Translation

This model is a fine-tuned version of ymoslem/qwen-audio-en-de-ft-decoder-32-24layers-iterative-pruning-chrf on the ymoslem/ACL-6060 dataset. It achieves the following results on the evaluation set:

• Loss: 0.9771

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: 4
• eval_batch_size: 4
• seed: 42
• optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: cosine
• num_epochs: 3

Training results

Training Loss	Epoch	Step	Validation Loss
1.5107	0.3497	100	1.2779
0.9759	0.6993	200	1.1098
0.3299	1.0490	300	1.0876
0.3006	1.3986	400	1.0439
0.307	1.7483	500	1.0281
0.0463	2.0979	600	0.9985
0.1783	2.4476	700	1.0037
0.1508	2.7972	800	0.9771

Framework versions

• Transformers 4.51.3
• Pytorch 2.4.1+cu124
• Datasets 3.6.0
• Tokenizers 0.21.1