Gemma2B-LORAfied

This model is a fine-tuned version of google/gemma-2b on the databricks/databricks-dolly-15k dataset. It achieves the following results on the evaluation set:

• Loss: 2.0206

Training Hardware

This model was trained using:

• GPU: Intel(R) Data Center GPU Max 1100
• CPU: Intel(R) Xeon(R) Platinum 8480+

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 1e-05
• train_batch_size: 2
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 8
• total_train_batch_size: 16
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.05
• training_steps: 1480

Training results

Training Loss	Epoch	Step	Validation Loss
2.927	1.64	100	2.5783
2.4568	3.28	200	2.2983
2.2609	4.92	300	2.1769
2.1671	6.56	400	2.1051
2.1065	8.2	500	2.0739
2.0844	9.84	600	2.0567
2.0643	11.48	700	2.0455
2.0511	13.11	800	2.0374
2.0435	14.75	900	2.0318
2.0304	16.39	1000	2.0276
2.0245	18.03	1100	2.0248
2.0247	19.67	1200	2.0228
2.0096	21.31	1300	2.0212
2.0183	22.95	1400	2.0206

Framework versions

• PEFT 0.10.0
• Transformers 4.39.3
• Pytorch 2.0.1a0+cxx11.abi
• Datasets 2.18.0
• Tokenizers 0.15.2