qwen2_5_omni_wo_video_1016

This model was trained from scratch on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.6904
• Token Acc: 0.7885

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: 0.0001
• train_batch_size: 2
• eval_batch_size: 1
• seed: 42
• distributed_type: multi-GPU
• num_devices: 8
• gradient_accumulation_steps: 4
• total_train_batch_size: 64
• total_eval_batch_size: 8
• optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.95) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: cosine
• lr_scheduler_warmup_ratio: 0.05
• num_epochs: 3.0

Training results

Training Loss	Epoch	Step	Validation Loss	Token Acc
0.9505	0.1768	50	0.9682	0.7253
0.9444	0.3537	100	0.9573	0.7272
0.9103	0.5305	150	0.9360	0.7337
0.8897	0.7073	200	0.9147	0.7370
0.8645	0.8842	250	0.8898	0.7423
0.7806	1.0601	300	0.8696	0.7475
0.7602	1.2370	350	0.8452	0.7527
0.7295	1.4138	400	0.8240	0.7588
0.6995	1.5906	450	0.7984	0.7627
0.6855	1.7675	500	0.7722	0.7688
0.6592	1.9443	550	0.7455	0.7741
0.5814	2.1202	600	0.7314	0.7785
0.5640	2.2971	650	0.7165	0.7818
0.5678	2.4739	700	0.7033	0.7852
0.5414	2.6508	750	0.6962	0.7866
0.5421	2.8276	800	0.6915	0.7881
0.5421	3.0	849	0.6904	0.7885

Framework versions

• Transformers 4.57.1
• Pytorch 2.9.0+cu128
• Datasets 3.6.0
• Tokenizers 0.22.1