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

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: 13.9174

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
561742.3125	0.1117	10	3654.1487
0.3373	0.2235	20	16.0614
4.7407	0.3352	30	20.0628
13.0883	0.4469	40	16.9850
0.2847	0.5587	50	14.4631
0.8503	0.6704	60	16.1353
0.3988	0.7821	70	16.5311
0.1785	0.8939	80	16.1144
0.1524	1.0	90	17.1748
0.1447	1.1117	100	15.2639
0.4107	1.2235	110	14.7690
0.4008	1.3352	120	12.9608
0.0564	1.4469	130	11.9789
1.7456	1.5587	140	14.9167
0.1389	1.6704	150	15.7580
0.0553	1.7821	160	17.4344
0.1096	1.8939	170	14.6135
0.0714	2.0	180	17.3837
0.0931	2.1117	190	16.1148
0.0919	2.2235	200	15.9285
0.5142	2.3352	210	14.0383
0.2968	2.4469	220	16.4832
0.0483	2.5587	230	16.1801
0.0327	2.6704	240	13.8616
0.0405	2.7821	250	17.9884
0.0927	2.8939	260	13.9174

Framework versions

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