metadata
license: apache-2.0
base_model: mistralai/Mistral-7B-Instruct-v0.1
tags:
- trl
- sft
- generated_from_trainer
datasets:
- super_glue
metrics:
- accuracy
model-index:
- name: original_glue_boolq
results: []
original_glue_boolq
This model is a fine-tuned version of mistralai/Mistral-7B-Instruct-v0.1 on the super_glue dataset. It achieves the following results on the evaluation set:
- Loss: 0.3115
- Accuracy: 0.8735
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: 1
- 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 |
---|---|---|---|---|
0.369 | 0.05 | 50 | 0.4328 | 0.8120 |
0.2708 | 0.1 | 100 | 0.4051 | 0.8283 |
0.6276 | 0.15 | 150 | 0.4020 | 0.8452 |
0.4395 | 0.2 | 200 | 0.3671 | 0.8452 |
0.3282 | 0.25 | 250 | 0.3746 | 0.8445 |
0.2967 | 0.3 | 300 | 0.3557 | 0.8523 |
0.2483 | 0.35 | 350 | 0.3862 | 0.8622 |
0.384 | 0.4 | 400 | 0.3765 | 0.8565 |
0.334 | 0.45 | 450 | 0.3628 | 0.8601 |
0.2671 | 0.5 | 500 | 0.3290 | 0.8664 |
0.2478 | 0.55 | 550 | 0.3421 | 0.8650 |
0.1814 | 0.6 | 600 | 0.3233 | 0.8693 |
0.3332 | 0.65 | 650 | 0.3451 | 0.8728 |
0.2063 | 0.7 | 700 | 0.3709 | 0.8678 |
0.2614 | 0.75 | 750 | 0.3530 | 0.8763 |
0.4273 | 0.8 | 800 | 0.3383 | 0.8721 |
0.1319 | 0.85 | 850 | 0.3360 | 0.8735 |
0.196 | 0.9 | 900 | 0.3096 | 0.8735 |
0.3564 | 0.95 | 950 | 0.3354 | 0.8770 |
0.3145 | 1.0 | 1000 | 0.3421 | 0.8784 |
0.1344 | 1.05 | 1050 | 0.4273 | 0.8735 |
0.4227 | 1.1 | 1100 | 0.3555 | 0.8707 |
0.1696 | 1.15 | 1150 | 0.3399 | 0.8742 |
0.5423 | 1.2 | 1200 | 0.3405 | 0.8813 |
Framework versions
- Transformers 4.35.2
- Pytorch 2.1.1+cu121
- Datasets 2.15.0
- Tokenizers 0.15.0