jerseyjerry/task-5-microsoft-Phi-3-mini-4k-instruct-20250301

lora

This model is a fine-tuned version of microsoft/Phi-3-mini-4k-instruct on the flock_task5_tranning dataset. It achieves the following results on the evaluation set:

• Loss: 0.0052

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: 5e-05
• train_batch_size: 2
• eval_batch_size: 1
• seed: 42
• distributed_type: multi-GPU
• num_devices: 2
• gradient_accumulation_steps: 2
• total_train_batch_size: 8
• total_eval_batch_size: 2
• optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: cosine
• lr_scheduler_warmup_ratio: 0.05
• num_epochs: 60

Training results

Training Loss	Epoch	Step	Validation Loss
1.5231	2.5	10	1.6286
1.4069	5.0	20	1.4773
1.2518	7.5	30	1.3657
1.3069	10.0	40	1.2441
1.0816	12.5	50	1.0924
1.0063	15.0	60	0.9201
0.666	17.5	70	0.7236
0.5723	20.0	80	0.5105
0.3671	22.5	90	0.3136
0.2108	25.0	100	0.1737
0.1203	27.5	110	0.0830
0.069	30.0	120	0.0397
0.0233	32.5	130	0.0212
0.0158	35.0	140	0.0129
0.0104	37.5	150	0.0093
0.0081	40.0	160	0.0076
0.0073	42.5	170	0.0066
0.0072	45.0	180	0.0060
0.0062	47.5	190	0.0056
0.0063	50.0	200	0.0054
0.0068	52.5	210	0.0053
0.0064	55.0	220	0.0052
0.0061	57.5	230	0.0052
0.0056	60.0	240	0.0052

Framework versions

• PEFT 0.12.0
• Transformers 4.48.3
• Pytorch 2.6.0+cu124
• Datasets 3.2.0
• Tokenizers 0.21.0