Views
No views yet
1#!/usr/bin/bash
2
3python ../run_glue.py \\n --model_name_or_path google/fnet-base \\n --task_name stsb \\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 fnet-base-finetuned-stsb \\n --push_to_hub \\n --hub_strategy all_checkpoints \\n --logging_strategy epoch \\n --save_strategy epoch \\n --evaluation_strategy epoch \\n```
4
5### Training hyperparameters
6
7The following hyperparameters were used during training:
8- learning_rate: 2e-05
9- train_batch_size: 16
10- eval_batch_size: 8
11- seed: 42
12- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
13- lr_scheduler_type: linear
14- num_epochs: 3.0
15
16### Training results
17
18| Training Loss | Epoch | Step | Combined Score | Validation Loss | Pearson | Spearmanr |
19|:-------------:|:-----:|:----:|:--------------:|:---------------:|:-------:|:---------:|
20| 1.5473 | 1.0 | 360 | 0.8120 | 0.7751 | 0.8115 | 0.8125 |
21| 0.6954 | 2.0 | 720 | 0.8145 | 0.8717 | 0.8160 | 0.8130 |
22| 0.4828 | 3.0 | 1080 | 0.8238 | 0.7894 | 0.8256 | 0.8219 |
23
24
25### Framework versions
26
27- Transformers 4.11.0.dev0
28- Pytorch 1.9.0
29- Datasets 1.12.1
30- Tokenizers 0.10.3