update readme
Browse files
README.md
CHANGED
@@ -10,4 +10,24 @@ size_categories: 10K<n<100K
|
|
10 |
**QR-AN Dataset: a classification dataset of french Parliament questions-answers.**
|
11 |
|
12 |
This is a dataset for theme/topic classification, made of questions and answers from https://www2.assemblee-nationale.fr/recherche/resultats_questions . \
|
13 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
10 |
**QR-AN Dataset: a classification dataset of french Parliament questions-answers.**
|
11 |
|
12 |
This is a dataset for theme/topic classification, made of questions and answers from https://www2.assemblee-nationale.fr/recherche/resultats_questions . \
|
13 |
+
|
14 |
+
It contains 188 unbalanced classes, 80k questions-answers divided into 3 splits: train (60k), val (10k) and test (10k). \
|
15 |
+
|
16 |
+
Compatible with [run_glue.py](https://github.com/huggingface/transformers/tree/master/examples/pytorch/text-classification) script:
|
17 |
+
```
|
18 |
+
export MODEL_NAME=camembert-base
|
19 |
+
export MAX_SEQ_LENGTH=512
|
20 |
+
|
21 |
+
python run_glue.py \
|
22 |
+
--model_name_or_path $MODEL_NAME \
|
23 |
+
--dataset_name cassandra-themis/QR-AN \
|
24 |
+
--do_train \
|
25 |
+
--do_eval \
|
26 |
+
--max_seq_length $MAX_SEQ_LENGTH \
|
27 |
+
--per_device_train_batch_size 8 \
|
28 |
+
--gradient_accumulation_steps 4 \
|
29 |
+
--learning_rate 2e-5 \
|
30 |
+
--num_train_epochs 1 \
|
31 |
+
--max_eval_samples 500 \
|
32 |
+
--output_dir tmp/QR-AN
|
33 |
+
```
|