Upload 3 files
Browse files- README.md +10 -0
- train_log.txt +29 -0
- training_args.json +1 -0
README.md
ADDED
@@ -0,0 +1,10 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
## TextAttack Model Card
|
2 |
+
|
3 |
+
This `bert` model was fine-tuned using TextAttack. The model was fine-tuned
|
4 |
+
for 3 epochs with a batch size of 8,
|
5 |
+
a maximum sequence length of 512, and an initial learning rate of 3e-05.
|
6 |
+
Since this was a classification task, the model was trained with a cross-entropy loss function.
|
7 |
+
The best score the model achieved on this task was 0.9558333333333333, as measured by the
|
8 |
+
eval set accuracy, found after 3 epochs.
|
9 |
+
|
10 |
+
For more information, check out [TextAttack on Github](https://github.com/QData/TextAttack).
|
train_log.txt
ADDED
@@ -0,0 +1,29 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
Writing logs to ./outputs/2024-02-27-20-27-33-802873/train_log.txt.
|
2 |
+
Wrote original training args to ./outputs/2024-02-27-20-27-33-802873/training_args.json.
|
3 |
+
***** Running training *****
|
4 |
+
Num examples = 9600
|
5 |
+
Num epochs = 3
|
6 |
+
Num clean epochs = 3
|
7 |
+
Instantaneous batch size per device = 8
|
8 |
+
Total train batch size (w. parallel, distributed & accumulation) = 8
|
9 |
+
Gradient accumulation steps = 1
|
10 |
+
Total optimization steps = 3600
|
11 |
+
==========================================================
|
12 |
+
Epoch 1
|
13 |
+
Running clean epoch 1/3
|
14 |
+
Train accuracy: 88.54%
|
15 |
+
Eval accuracy: 92.75%
|
16 |
+
Best score found. Saved model to ./outputs/2024-02-27-20-27-33-802873/best_model/
|
17 |
+
==========================================================
|
18 |
+
Epoch 2
|
19 |
+
Running clean epoch 2/3
|
20 |
+
Train accuracy: 96.02%
|
21 |
+
Eval accuracy: 95.08%
|
22 |
+
Best score found. Saved model to ./outputs/2024-02-27-20-27-33-802873/best_model/
|
23 |
+
==========================================================
|
24 |
+
Epoch 3
|
25 |
+
Running clean epoch 3/3
|
26 |
+
Train accuracy: 98.58%
|
27 |
+
Eval accuracy: 95.58%
|
28 |
+
Best score found. Saved model to ./outputs/2024-02-27-20-27-33-802873/best_model/
|
29 |
+
Wrote README to ./outputs/2024-02-27-20-27-33-802873/README.md.
|
training_args.json
ADDED
@@ -0,0 +1 @@
|
|
|
|
|
1 |
+
{"num_epochs": 3, "num_clean_epochs": 1, "attack_epoch_interval": 1, "early_stopping_epochs": null, "learning_rate": 3e-05, "num_warmup_steps": 500, "weight_decay": 0.01, "per_device_train_batch_size": 8, "per_device_eval_batch_size": 32, "gradient_accumulation_steps": 1, "random_seed": 718, "parallel": false, "load_best_model_at_end": false, "alpha": 1.0, "num_train_adv_examples": -1, "query_budget_train": null, "attack_num_workers_per_device": 1, "output_dir": "./outputs/2024-02-27-20-27-33-802873", "checkpoint_interval_steps": null, "checkpoint_interval_epochs": null, "save_last": true, "log_to_tb": false, "tb_log_dir": null, "log_to_wandb": false, "wandb_project": "textattack", "logging_interval_step": 1}
|