metadata

license: apache-2.0
library_name: peft
tags:
  - trl
  - sft
  - generated_from_trainer
datasets:
  - generator
base_model: mistralai/Mixtral-8x7B-v0.1
model-index:
  - name: mixtral_full
    results: []

mixtral_full

This model is a fine-tuned version of mistralai/Mixtral-8x7B-v0.1 on the generator dataset. It achieves the following results on the evaluation set:

Loss: 0.9139

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: 2
eval_batch_size: 1
seed: 42
gradient_accumulation_steps: 32
total_train_batch_size: 64
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: cosine
lr_scheduler_warmup_ratio: 0.03
num_epochs: 2.2

Training results

Training Loss	Epoch	Step	Validation Loss
1.1966	0.07	20	1.1074
1.0281	0.14	40	1.0284
0.9837	0.21	60	0.9990
0.992	0.28	80	0.9808
0.9642	0.35	100	0.9685
0.9706	0.42	120	0.9591
0.9355	0.5	140	0.9523
0.9253	0.57	160	0.9468
0.9125	0.64	180	0.9412
0.9232	0.71	200	0.9363
0.9183	0.78	220	0.9320
0.9175	0.85	240	0.9284
0.9219	0.92	260	0.9253
0.9028	0.99	280	0.9228
0.8405	1.06	300	0.9251
0.8429	1.13	320	0.9238
0.8453	1.2	340	0.9231
0.8722	1.27	360	0.9214
0.8511	1.35	380	0.9200
0.8471	1.42	400	0.9186
0.8376	1.49	420	0.9171
0.8372	1.56	440	0.9160
0.8517	1.63	460	0.9155
0.8244	1.7	480	0.9147
0.8432	1.77	500	0.9140
0.8519	1.84	520	0.9135
0.8128	1.91	540	0.9135
0.8296	1.98	560	0.9134
0.7848	2.05	580	0.9138
0.8166	2.12	600	0.9139
0.7963	2.2	620	0.9139

Framework versions

PEFT 0.7.2.dev0
Transformers 4.38.1
Pytorch 2.1.2+cu121
Datasets 2.16.1
Tokenizers 0.15.0