metadata

tags:
  - trl
  - sft
  - generated_from_trainer
datasets:
  - super_glue
metrics:
  - accuracy
model-index:
  - name: mistral_sparse_80_percent_boolq_1000
    results: []

mistral_sparse_80_percent_boolq_1000

This model is a fine-tuned version of on the super_glue dataset. It achieves the following results on the evaluation set:

Loss: 0.3381
Accuracy: 0.8664

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: 1e-05
train_batch_size: 2
eval_batch_size: 4
seed: 2
distributed_type: multi-GPU
num_devices: 2
gradient_accumulation_steps: 2
total_train_batch_size: 8
total_eval_batch_size: 8
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: linear
num_epochs: 10

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
0.4991	0.05	50	0.5522	0.7216
0.3812	0.1	100	0.4342	0.8141
0.369	0.15	150	0.4112	0.8170
0.4132	0.2	200	0.4139	0.8382
0.4219	0.25	250	0.3940	0.8339
0.4144	0.3	300	0.3803	0.8481
0.1534	0.35	350	0.3786	0.8516
0.4855	0.4	400	0.3821	0.8502
0.2109	0.45	450	0.3583	0.8516
0.3026	0.5	500	0.3675	0.8558
0.2903	0.55	550	0.3744	0.8537
0.2988	0.6	600	0.3573	0.8587
0.3432	0.65	650	0.3396	0.8657
0.3156	0.7	700	0.3299	0.8671
0.4978	0.75	750	0.3623	0.8657
0.4523	0.8	800	0.3240	0.8700
0.2367	0.85	850	0.3393	0.8678
0.3334	0.9	900	0.3252	0.8834
0.3286	0.95	950	0.3605	0.8742
0.1659	1.0	1000	0.3269	0.8742
0.2373	1.05	1050	0.3256	0.8792
0.5102	1.1	1100	0.3633	0.8749

Framework versions

Transformers 4.35.2
Pytorch 2.1.1+cu121
Datasets 2.15.0
Tokenizers 0.15.0