Buckets:
Luôn luôn sync file bắt đầu từ source code local (Vì dùng cursor để debug) => tải lên Colab, Kaggle, ThueGPU.vn
Đang Training data is on MINDsmall
LƯU Ý:
- Running data_preprocess.py script and Tensorboard on Computer
- Glove: https://nlp.stanford.edu/data/glove.840B.300d.zip
- entity_embedding_dim = 100 instead of 300
- Training on GPU P100 in Kaggle, A4000 on ThueGPU.vn (~24h)
- Test training on Google Colab
Using the REMARK Tip: ~~~~ The REMARK environment variable lets you customize the run name in TensorBoard for better readability:
- REMARK="num-filters-300-window-size-5" python3 src/train.py
Ubuntu setup note:
sudo apt-get update
sudo apt-get install python3-pip-whl python3-setuptools-whl python3.10-venv
python3 -m venv venv
source venv/bin/activate
pip install -r requirements.txt
export MODEL_NAME=NAML_LSTUR_DKN
python3 train.py
Xet Storage Details
- Size:
- 964 Bytes
- Xet hash:
- acd68ded07b361e8e43d0ec69303b732c473717ad35842d8b9c0f6e3d6b72399
·
Xet efficiently stores files, intelligently splitting them into unique chunks and accelerating uploads and downloads. More info.