Datasets:

Modalities:
Text
ArXiv:
File size: 628 Bytes
f9dff7a
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
OUT_ROOT="./out/pretrain/qwen2_7b_question_focus_lr_plus"
#"./out/pretrain/qwen2_7b_question_focus"

LOG_ROOT="./out/pretrain/qwen2_7b_question_focus_lr_plus/teelog.txt"
#"./out/pretrain/qwen2_7b_question_focus/teelog.txt"

DATA_ROOT="litgpt/data/arxiv_test_qwen2_tokenized/"

CKPT_0="./checkpoints/Qwen/Qwen2-7B/"

CONFIG_FILE="config_hub/pretrain/qwen2_7b_test.yaml"


export CUDA_VISIBLE_DEVICES="4,5,6,7"

touch "$LOG_ROOT"


litgpt pretrain \
    --config $CONFIG_FILE \
    --train.max_tokens 52428800 \
    --out_dir $OUT_ROOT \
    --initial_checkpoint_dir $CKPT_0 \
    --data_dir $DATA_ROOT \
    2>&1 | tee $LOG_ROOT