Views
No views yet
1cd ./cdb_cn
2wget -O cards.cdb https://cdn02.moecube.com:444/ygopro-database/zh-CN/cards.cdb
3
4cd ./cdb_jp
5wget -O cards.cdb https://cdn02.moecube.com:444/ygopro-database/ja_JP/cards.cdbgit clone https://github.com/mycard/ygopro-scripts.gitpython all_in_one.pyinstruction:
下面是卡片的信息,请根据这些信息生成lua脚本:{name},{desc},{tag},卡密为{id}output:
{code}finetune_deepseekcoder.py1deepspeed finetune.py \
2 --model_name_or_path $MODEL_PATH \
3 --data_path $DATA_PATH \
4 --output_dir $OUTPUT_PATH \
5 --num_train_epochs 3 \
6 --model_max_length 4096 \
7 --per_device_train_batch_size 8 \
8 --per_device_eval_batch_size 1 \
9 --gradient_accumulation_steps 4 \
10 --evaluation_strategy "no" \
11 --save_strategy "epoch" \
12 --save_total_limit 5 \
13 --learning_rate 2e-5 \
14 --warmup_steps 10 \
15 --logging_steps 100 \
16 --lr_scheduler_type "cosine" \
17 --gradient_checkpointing True \
18 --report_to "tensorboard" \
19 --deepspeed configs/ds_config_zero3.json \
20 --bf16 True| 模型路径 | ROUGE-1 | ROUGE-2 | ROUGE-L | BLEU | BERTScore Precision | BERTScore Recall | BERTScore F1 |
|---|---|---|---|---|---|---|---|
| base model | 0.0753 | 0.0125 | 0.0539 | 0.0010 | 0.6216 | 0.6621 | 0.6400 |
| on_cn+jp | 0.4603 | 0.4214 | 0.4302 | 0.1183 | 0.8841 | 0.8541 | 0.8673 |
| on_cn | 0.3042 | 0.2610 | 0.2750 | 0.0769 | 0.7955 | 0.7647 | 0.7767 |