kubapok/ireland-news-word-gap-prediction
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

a

Browse Source
This commit is contained in:
Jakub Pokrywka 2021-11-04 12:24:59 +01:00
parent 2898bee8e9
commit aea7951a1c
5 changed files with 20 additions and 12 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

8
how_to_run
View File

@ -4,3 +4,11 @@ instal datasets: pip install datasets
install tokenizers: pip install tokenizers install tokenizers: pip install tokenizers
to run: go do a solution directory and: bash run.sh to run: go do a solution directory and: bash run.sh
adjust:
per_device_train_batch_size
per_device_eval_batch_size
gradient_accumulation_steps
if you want to run scripts on other dataset remeber to adjust max_seq_length

6
regular_roberta_from_scratch/2_run.sh
View File

@ -8,9 +8,9 @@ python run_mlm.py \
--validation_file ./dev-0_in.csv \ --validation_file ./dev-0_in.csv \
--do_train \ --do_train \
--do_eval \ --do_eval \
--per_device_train_batch_size=1 \ --per_device_train_batch_size=256 \
--per_device_eval_batch_size=1 \ --per_device_eval_batch_size=256 \
--gradient_accumulation_steps=4 \ --gradient_accumulation_steps=1 \
--fp16 False \ --fp16 False \
--save_steps 1000 \ --save_steps 1000 \
--eval_steps 1000 \ --eval_steps 1000 \

6
roberta_ft/2_run.sh
View File

@ -8,9 +8,9 @@ python run_mlm.py \
--validation_file ./dev-0_in.csv \ --validation_file ./dev-0_in.csv \
--do_train \ --do_train \
--do_eval \ --do_eval \
--per_device_train_batch_size=32 \ --per_device_train_batch_size=256 \
--per_device_eval_batch_size=32 \ --per_device_eval_batch_size=256 \
--gradient_accumulation_steps=8 \ --gradient_accumulation_steps=1 \
--fp16 False \ --fp16 False \
--save_steps 1000 \ --save_steps 1000 \
--eval_steps 1000 \ --eval_steps 1000 \

6
roberta_with_year_from_scratch/2_run.sh
View File

@ -8,9 +8,9 @@ python run_mlm.py \
--validation_file ./dev-0_in.csv \ --validation_file ./dev-0_in.csv \
--do_train \ --do_train \
--do_eval \ --do_eval \
--per_device_train_batch_size=1 \ --per_device_train_batch_size=256 \
--per_device_eval_batch_size=1 \ --per_device_eval_batch_size=256 \
--gradient_accumulation_steps=4 \ --gradient_accumulation_steps=1 \
--fp16 False \ --fp16 False \
--save_steps 1000 \ --save_steps 1000 \
--eval_steps 1000 \ --eval_steps 1000 \

6
roberta_with_year_ft/2_run.sh
View File

@ -8,9 +8,9 @@ python run_mlm.py \
--validation_file ./dev-0_in.csv \ --validation_file ./dev-0_in.csv \
--do_train \ --do_train \
--do_eval \ --do_eval \
--per_device_train_batch_size=32 \ --per_device_train_batch_size=256 \
--per_device_eval_batch_size=32 \ --per_device_eval_batch_size=256 \
--gradient_accumulation_steps=8 \ --gradient_accumulation_steps=1 \
--fp16 False \ --fp16 False \
--save_steps 1000 \ --save_steps 1000 \
--eval_steps 1000 \ --eval_steps 1000 \