DeepSeek/ESFT
1
0
Fork 0
mirror of https://github.com/deepseek-ai/ESFT synced 2025-06-26 18:15:50 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity

update eval and readme

Browse Source
This commit is contained in:
ZihanWang314
2024-08-09 18:06:57 +08:00
parent 809d0e377e
commit 36c6194349
10 changed files with 333 additions and 152 deletions
Download Patch File Download Diff File Expand all files Collapse all files

18
scripts/eval.sh
View File

@@ -1,12 +1,24 @@
# first, download adapter models and put them to the corresponding directories
# first: download adapter models and put them to the corresponding directories
python scripts/eval.py \
python eval_multigpu.py \
--eval_datasets=translation \
--base_model_path=deepseek-ai/ESFT-vanilla-lite \
--adapter_dir=all_models/adapters/token \
--output_dir=results/completions/token \
--max_new_tokens=512 \
--openai_api_key=REPLACE_WITH_YOUR_KEY \
--eval_batch_size=2
--eval_batch_size=2 \
--world_size=4 \
--gpus_per_rank=2
# this script is used for single-gpu training and has been deprecated. If you have no multiple gpus, you can set above world_size=1 and gpus_per_rank=1
# python scripts/eval.py \
# --eval_datasets=translation \
# --base_model_path=deepseek-ai/ESFT-vanilla-lite \
# --adapter_dir=all_models/adapters/token \
# --output_dir=results/completions/token \
# --max_new_tokens=512 \
# --openai_api_key=REPLACE_WITH_YOUR_KEY \
# --eval_batch_size=2