forked from NVIDIA/Megatron-LM
-
Notifications
You must be signed in to change notification settings - Fork 0
Expand file tree
/
Copy pathmmlu.sh
More file actions
19 lines (15 loc) · 634 Bytes
/
mmlu.sh
File metadata and controls
19 lines (15 loc) · 634 Bytes
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
#!/bin/bash
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 --sequence-parallel"
${LAUNCH_SCRIPT} ${SCRIPT_DIR}/mmlu.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}