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

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: 16.1311

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
199892.9688	0.1117	10	775.1600
0.3377	0.2235	20	18.9439
18.127	0.3352	30	15.0859
3.0535	0.4469	40	17.2034
0.0927	0.5587	50	17.3699
0.1183	0.6704	60	16.5855
0.0634	0.7821	70	17.7980
0.0472	0.8939	80	17.8210
0.0222	1.0	90	17.0549
0.0432	1.1117	100	19.1342
0.2647	1.2235	110	18.6049
0.0742	1.3352	120	18.7196
0.0208	1.4469	130	16.5360
0.0454	1.5587	140	16.6369
0.0419	1.6704	150	16.4503
0.0328	1.7821	160	16.6961
0.053	1.8939	170	16.4221
0.0173	2.0	180	17.6572
0.0145	2.1117	190	16.6156
0.0126	2.2235	200	18.8278
0.0208	2.3352	210	16.4277
0.0186	2.4469	220	15.9193
0.0108	2.5587	230	18.5243
0.0068	2.6704	240	15.9373
0.0108	2.7821	250	19.0050
0.025	2.8939	260	16.1311

Framework versions

• Transformers 4.48.2
• Pytorch 2.4.1+cu124
• Datasets 3.6.0
• Tokenizers 0.21.1