upgrade ppo dpo rm script

This commit is contained in:
YeAnbang
2024-05-28 03:04:39 +00:00
parent 7a7e86987d
commit 929e1e3da4
15 changed files with 169 additions and 139 deletions

View File

@@ -1,4 +1,4 @@
SAVE_DIR="/home/yeanbang/data/experiment/dataset/alpaca/test/Llama-2-7b-chat-hf"
SAVE_DIR="/home/yeanbang/data/experiment/dataset/alpaca/test/Yi-1.5-6B"
rm -rf $SAVE_DIR/cache
rm -rf $SAVE_DIR/jsonl
@@ -14,9 +14,9 @@ rm -rf $SAVE_DIR/arrow
python prepare_dataset.py --type sft \
--data_input_dirs /mnt/jfs-hdd/home/yeanbang/data/dataset/sft_data/alpaca/data_preprocessed/train \
--conversation_template_config /home/yeanbang/data/ColossalAI/applications/ColossalChat/config/conversation_template/llama2.json \
--tokenizer_dir "/mnt/jfs-hdd/share/models/Llama-2-7b-chat-hf" \
--data_input_dirs /home/yeanbang/data/experiment/dataset/sft_data/test/sft-data \
--conversation_template_config /home/yeanbang/data/ColossalAI/applications/ColossalChat/config/conversation_template/01-ai_Yi-1.5-9B-Chat.json \
--tokenizer_dir "/mnt/jfs-hdd/share/models/Yi-1.5-6B" \
--data_cache_dir $SAVE_DIR/cache \
--data_jsonl_output_dir $SAVE_DIR/jsonl \
--data_arrow_output_dir $SAVE_DIR/arrow \