Instructions to use KexuanShi/Megatron-LM with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- NeMo
How to use KexuanShi/Megatron-LM with NeMo:
# tag did not correspond to a valid NeMo domain.
- Notebooks
- Google Colab
- Kaggle
| SCRIPT_DIR="$(dirname "$(readlink -f "$0")")" | |
| # Common arguments and base model specific arguments | |
| source "${SCRIPT_DIR}/conf/arguments.sh" | |
| # Extra arguments of this script | |
| MLM_DEFAULT_ARGS="--finetune --auto-detect-ckpt-format --export-te-mcore-model" | |
| if [ -z ${MLM_MODEL_CKPT} ]; then | |
| printf "${MLM_ERROR} Variable ${PURPLE}MLM_MODEL_CKPT${WHITE} must be set!\n" | |
| exit 1 | |
| fi | |
| TOOLS_DIR="$(realpath ${SCRIPT_DIR}/../../../tools)" | |
| ${LAUNCH_SCRIPT} ${TOOLS_DIR}/run_text_generation_server.py \ | |
| ${MODEL_ARGS} \ | |
| --tensor-model-parallel-size ${TP} \ | |
| --expert-tensor-parallel-size ${ETP} \ | |
| --expert-model-parallel-size ${EP} \ | |
| --pipeline-model-parallel-size ${PP} \ | |
| --tokenizer-model ${TOKENIZER_MODEL} \ | |
| --load ${MLM_MODEL_CKPT} \ | |
| ${MLM_DEFAULT_ARGS} ${MLM_EXTRA_ARGS} | |