fine-tuning-Phi2-with-webglm-qa-with-lora

This model is a fine-tuned version of microsoft/phi-2 on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.1032

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: 2
• seed: 42
• gradient_accumulation_steps: 5
• total_train_batch_size: 10
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 50
• training_steps: 500
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss
No log	0.2	10	8.0361
No log	0.4	20	6.2064
No log	0.6	30	2.7739
No log	0.8	40	0.6071
4.4774	1.0	50	0.5329
4.4774	1.2	60	0.4635
4.4774	1.39	70	0.4081
4.4774	1.59	80	0.3576
4.4774	1.79	90	0.3173
0.3338	1.99	100	0.2889
0.3338	2.19	110	0.2645
0.3338	2.39	120	0.2471
0.3338	2.59	130	0.2301
0.3338	2.79	140	0.2121
0.1964	2.99	150	0.1992
0.1964	3.19	160	0.1913
0.1964	3.39	170	0.1793
0.1964	3.59	180	0.1713
0.1964	3.78	190	0.1642
0.1501	3.98	200	0.1579
0.1501	4.18	210	0.1531
0.1501	4.38	220	0.1511
0.1501	4.58	230	0.1455
0.1501	4.78	240	0.1379
0.1248	4.98	250	0.1333
0.1248	5.18	260	0.1313
0.1248	5.38	270	0.1308
0.1248	5.58	280	0.1271
0.1248	5.78	290	0.1244
0.1097	5.98	300	0.1208
0.1097	6.18	310	0.1178
0.1097	6.37	320	0.1164
0.1097	6.57	330	0.1155
0.1097	6.77	340	0.1125
0.0976	6.97	350	0.1108
0.0976	7.17	360	0.1109
0.0976	7.37	370	0.1093
0.0976	7.57	380	0.1085
0.0976	7.77	390	0.1079
0.0917	7.97	400	0.1072
0.0917	8.17	410	0.1064
0.0917	8.37	420	0.1058
0.0917	8.57	430	0.1054
0.0917	8.76	440	0.1047
0.0855	8.96	450	0.1040
0.0855	9.16	460	0.1034
0.0855	9.36	470	0.1032
0.0855	9.56	480	0.1032
0.0855	9.76	490	0.1032
0.0833	9.96	500	0.1032

Framework versions

• PEFT 0.7.1
• Transformers 4.36.2
• Pytorch 2.0.0
• Datasets 2.15.0
• Tokenizers 0.15.0