KnowLang/graphcodebert-codesearch-python
05
1---2license: mit3---4 5## GraphCodeBERT CodeSearch Python6 7- base model comes from ['microsoft/graphcodebert-base'](https://huggingface.co/microsoft/graphcodebert-base)8- fine tuined for the python code-search task based on [GraphCodeBERT CodeSearch](https://github.com/microsoft/CodeBERT/tree/master/GraphCodeBERT/codesearch)9- best mrr: 0.66210 11### fine tuning script12```sh13lang=python14mkdir -p ./saved_models/$lang15python run.py \16 --output_dir=./saved_models/$lang \17 --config_name=microsoft/graphcodebert-base \18 --model_name_or_path=microsoft/graphcodebert-base \19 --tokenizer_name=microsoft/graphcodebert-base \20 --lang=$lang \21 --do_train \22 --train_data_file=dataset/$lang/train.jsonl \23 --eval_data_file=dataset/$lang/valid.jsonl \24 --test_data_file=dataset/$lang/test.jsonl \25 --codebase_file=dataset/$lang/codebase.jsonl \26 --num_train_epochs 10 \27 --code_length 256 \28 --data_flow_length 64 \29 --nl_length 128 \30 --train_batch_size 16 \31 --eval_batch_size 32 \32 --learning_rate 2e-5 \33 --seed 123456 2>&1| tee saved_models/$lang/train.log34```