Add scripts to fine-tune BERT (probably not needed) by Quora dataset …

…and generate embeddings by BERT.
wenlongx · May 23, 2019 · 52e2f34 · 52e2f34
1 parent d770aa6
commit 52e2f34
Show file tree

Hide file tree

Showing 3 changed files with 1,057 additions and 0 deletions.
diff --git a/bert/fine-tune-Quora.sh b/bert/fine-tune-Quora.sh
@@ -0,0 +1,16 @@
+export BERT_BASE_DIR="/home/max/Documents/NLP-playground/uncased_L-12_H-768_A-12"
+export QUORA_DIR="/home/max/Documents/NLP-playground/quora-insincere"
+
+python bert/run_classifier_quora.py \
+  --task_name=quora \
+  --do_train=true \
+  --do_eval=true \
+  --data_dir=$QUORA_DIR \
+  --vocab_file=$BERT_BASE_DIR/vocab.txt \
+  --bert_config_file=$BERT_BASE_DIR/bert_config.json \
+  --init_checkpoint=$BERT_BASE_DIR/bert_model.ckpt \
+  --max_seq_length=32 \
+  --train_batch_size=8 \
+  --learning_rate=2e-5 \
+  --num_train_epochs=3.0 \
+  --output_dir=$QUORA_DIR/bert_output