Discussion-Phi-4-multimodal-instruct-audio-dimp-alpha

This model is a fine-tuned version of microsoft/Phi-4-multimodal-instruct on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 13.3911

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: 4e-05
• train_batch_size: 1
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 16
• total_train_batch_size: 16
• optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.95) and epsilon=1e-07 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 50
• num_epochs: 3

Training results

Training Loss	Epoch	Step	Validation Loss
614700.625	0.1117	10	3857.8706
0.0989	0.2235	20	16.7370
2.1703	0.3352	30	15.6149
22.0924	0.4469	40	14.3853
0.1729	0.5587	50	15.3640
0.4787	0.6704	60	12.0508
0.4442	0.7821	70	11.0306
0.0653	0.8939	80	11.0157
0.0255	1.0	90	13.1966
0.0258	1.1117	100	13.2358
0.3599	1.2235	110	15.2634
0.0537	1.3352	120	16.0396
0.0199	1.4469	130	14.4705
0.1635	1.5587	140	12.5128
0.0398	1.6704	150	13.3947
0.0157	1.7821	160	12.4733
0.0494	1.8939	170	12.6727
0.0103	2.0	180	12.1961
0.0284	2.1117	190	13.5831
0.0193	2.2235	200	12.4818
0.0481	2.3352	210	13.1176
0.0625	2.4469	220	15.4632
0.0243	2.5587	230	12.9704
0.0132	2.6704	240	15.8782
0.0128	2.7821	250	14.8728
0.0239	2.8939	260	13.3911

Framework versions

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