Views
No views yet
1#!/usr/bin/bash
2
3
4python ../run_glue.py \\n --model_name_or_path bert-base-cased \\n --task_name sst2 \\n --do_train \\n --do_eval \\n --max_seq_length 512 \\n --per_device_train_batch_size 16 \\n --learning_rate 2e-5 \\n --num_train_epochs 3 \\n --output_dir bert-base-cased-finetuned-sst2 \\n --push_to_hub \\n --hub_strategy all_checkpoints \\n --logging_strategy epoch \\n --save_strategy epoch \\n --evaluation_strategy epoch \\n```
5
6### Training hyperparameters
7
8The following hyperparameters were used during training:
9- learning_rate: 2e-05
10- train_batch_size: 16
11- eval_batch_size: 8
12- seed: 42
13- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
14- lr_scheduler_type: linear
15- num_epochs: 3.0
16
17### Training results
18
19| Training Loss | Epoch | Step | Accuracy | Validation Loss |
20|:-------------:|:-----:|:-----:|:--------:|:---------------:|
21| 0.233 | 1.0 | 4210 | 0.9174 | 0.2841 |
22| 0.1261 | 2.0 | 8420 | 0.9278 | 0.3310 |
23| 0.0768 | 3.0 | 12630 | 0.9232 | 0.3649 |
24
25
26### Framework versions
27
28- Transformers 4.11.0.dev0
29- Pytorch 1.9.0
30- Datasets 1.12.1
31- Tokenizers 0.10.3