phi-1_5-finetuned-qlora-cluster-gsm8k-v3-smallsubset

This model is a fine-tuned version of microsoft/phi-1_5 on the None dataset. It achieves the following results on the evaluation set:

• Loss: 1.4985

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: 0.0002
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 32
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• num_epochs: 25
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss
No log	0.99	31	1.2173
No log	1.98	62	1.1653
No log	2.98	93	1.1535
No log	4.0	125	1.1498
No log	4.99	156	1.1562
No log	5.98	187	1.1682
1.0347	6.98	218	1.1832
1.0347	8.0	250	1.1934
1.0347	8.99	281	1.2183
1.0347	9.98	312	1.2468
1.0347	10.98	343	1.2760
1.0347	12.0	375	1.3096
0.7791	12.99	406	1.3348
0.7791	13.98	437	1.3695
0.7791	14.98	468	1.3935
0.7791	16.0	500	1.4104
0.7791	16.99	531	1.4235
0.7791	17.98	562	1.4546
0.7791	18.98	593	1.4709
0.5995	20.0	625	1.4790
0.5995	20.99	656	1.4889
0.5995	21.98	687	1.4942
0.5995	22.98	718	1.4954
0.5995	24.0	750	1.4982
0.5995	24.8	775	1.4985

Framework versions

• PEFT 0.11.1
• Transformers 4.37.2
• Pytorch 2.3.0
• Datasets 2.19.1
• Tokenizers 0.15.1