phi-1_5-finetuned-qlora-cluster-gsm8k-v2

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: 1.9003

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
1.0468	1.0	233	1.1314
0.9635	2.0	467	1.1069
0.9293	3.0	701	1.1129
0.8905	4.0	935	1.1269
0.8478	5.0	1168	1.1509
0.7686	6.0	1402	1.1727
0.7125	7.0	1636	1.2254
0.6637	8.0	1870	1.2571
0.6155	9.0	2103	1.3230
0.574	10.0	2337	1.3985
0.5273	11.0	2571	1.4532
0.451	12.0	2805	1.5160
0.4102	13.0	3038	1.5888
0.3802	14.0	3272	1.6469
0.3586	15.0	3506	1.6916
0.3391	16.0	3740	1.7576
0.3194	17.0	3973	1.7898
0.293	18.0	4207	1.8284
0.2815	19.0	4441	1.8460
0.2739	20.0	4675	1.8681
0.2693	21.0	4908	1.8821
0.2646	22.0	5142	1.8908
0.2614	23.0	5376	1.8954
0.2577	24.0	5610	1.8993
0.2566	24.92	5825	1.9003

Framework versions

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