Mistral-7B-Instruct-v0.3-lora-commonsense

This model is a fine-tuned version of mistralai/Mistral-7B-Instruct-v0.3 on an unknown dataset. It achieves the following results on the evaluation set:

Loss: 0.6863

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: 0.0001
train_batch_size: 16
eval_batch_size: 8
seed: 42
gradient_accumulation_steps: 8
total_train_batch_size: 128
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: cosine
lr_scheduler_warmup_steps: 100
num_epochs: 3

Training results

Training Loss	Epoch	Step	Validation Loss
0.8738	0.1503	200	0.8158
0.8589	0.3006	400	0.7939
0.8589	0.4510	600	0.7800
0.8589	0.6013	800	0.7725
0.8305	0.7516	1000	0.7650
0.8331	0.9019	1200	0.7506
0.7808	1.0522	1400	0.7438
0.7781	1.2026	1600	0.7350
0.7647	1.3529	1800	0.7252
0.7651	1.5032	2000	0.7228
0.7522	1.6535	2200	0.7099
0.7587	1.8038	2400	0.6997
0.7383	1.9542	2600	0.6932
0.7071	2.1045	2800	0.6949
0.6919	2.2548	3000	0.6899
0.7136	2.4051	3200	0.6884
0.6912	2.5554	3400	0.6878
0.6889	2.7057	3600	0.6867
0.6862	2.8561	3800	0.6863