Upload start_qgen.sh
Browse files- start_qgen.sh +20 -0
start_qgen.sh
ADDED
|
@@ -0,0 +1,20 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
#! /usr/bin/env bash
|
| 2 |
+
|
| 3 |
+
python run_summarization_flax.py \
|
| 4 |
+
--output_dir ./mt5-base-qgen \
|
| 5 |
+
--model_name_or_path google/mt5-base \
|
| 6 |
+
--tokenizer_name google/mt5-base \
|
| 7 |
+
--train_file asdf.csv \
|
| 8 |
+
--text_column context \
|
| 9 |
+
--summary_column question \
|
| 10 |
+
--do_train \
|
| 11 |
+
--do_eval \
|
| 12 |
+
--num_train_epochs 4 \
|
| 13 |
+
--learning_rate 5e-5 --warmup_steps 1000 \
|
| 14 |
+
--dtype bfloat16 \
|
| 15 |
+
--per_device_train_batch_size 4 \
|
| 16 |
+
--per_device_eval_batch_size 4 \
|
| 17 |
+
--overwrite_output_dir \
|
| 18 |
+
--max_source_length 1024 \
|
| 19 |
+
--max_target_length 64 \
|
| 20 |
+
--push_to_hub
|