Undi95/Mistral-7B-smoll_pippa-lora

smoll_pippa

This model was trained from scratch on the smoll_pippa dataset. It achieves the following results on the evaluation set:

• Loss: 1.6736

Model description

Dataset trained on mistralai/Mistral-7B-v0.1

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 0.0002
• train_batch_size: 4
• eval_batch_size: 4
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 16
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• lr_scheduler_warmup_steps: 10
• num_epochs: 3

Training results

Training Loss	Epoch	Step	Validation Loss
1.6421	0.16	50	1.6217
1.6288	0.31	100	1.6144
1.5725	0.47	150	1.6102
1.5582	0.62	200	1.6065
1.6055	0.78	250	1.6051
1.5733	0.93	300	1.6023
1.4885	1.09	350	1.6130
1.484	1.24	400	1.6169
1.4354	1.4	450	1.6194
1.4427	1.56	500	1.6187
1.4687	1.71	550	1.6178
1.461	1.87	600	1.6174
1.327	2.02	650	1.6341
1.3015	2.18	700	1.6665
1.3328	2.33	750	1.6714
1.3453	2.49	800	1.6718
1.3458	2.64	850	1.6725
1.3016	2.8	900	1.6737
1.3018	2.95	950	1.6736

Framework versions

• Transformers 4.34.0.dev0
• Pytorch 2.0.1+cu118
• Datasets 2.14.5
• Tokenizers 0.14.0

If you want to support me, you can here.