FlagEmbedding/evaluation/air_bench/arguments.py 导入错误from air_benchmark import EvalArgs as AIRBenchEvalArgs

[Marverlises]

Traceback (most recent call last):
File "/root/autodl-tmp/FlagEmbedding/quickstart.py", line 6, in
from FlagEmbedding import FlagAutoModel
File "/root/autodl-tmp/FlagEmbedding/FlagEmbedding/init.py", line 2, in
from .evaluation import *
File "/root/autodl-tmp/FlagEmbedding/FlagEmbedding/evaluation/init.py", line 1, in
from .air_bench import AIRBenchEvalModelArgs, AIRBenchEvalArgs, AIRBenchEvalRunner
File "/root/autodl-tmp/FlagEmbedding/FlagEmbedding/evaluation/air_bench/init.py", line 1, in
from .arguments import AIRBenchEvalModelArgs
File "/root/autodl-tmp/FlagEmbedding/FlagEmbedding/evaluation/air_bench/arguments.py", line 3, in
from air_benchmark import EvalArgs as AIRBenchEvalArgs
ModuleNotFoundError: No module named 'air_benchmark'

[liulizuel]

I encountered the same problem.

[zqma]

same here. please help!

[hanhainebula]

Hello, thanks for pointing out this issue! Now this issue has been fixed in #1340. The air-benchmark is necessary only when you need to perform evaluation on AIR-Bench.