search

apache / tvm

Open deep learning compiler stack for cpu, gpu and specialized accelerators
https://tvm.apache.org/
Apache License 2.0
11.41k stars 3.4k forks source link

[Bug] #17090

Closed wangsheng115 closed 3 weeks ago

wangsheng115 commented 3 weeks ago

I run command: mlc_chat convert_weight ..................

An error happened, this is error message: TVMError: Op(relax.nn.rms_norm) requires the input gamma, beta, etc., to have size same as the lengths of the data on the given axes. However, there exists [T.int64(2304)] and [T.int64(4096)] that are unequal

platform: macOs Sonoma 14.5

How to avoid this problem?

Hzfengsy commented 3 weeks ago

If you'd like to ask about mlc-llm, please open the issue at https://github.com/mlc-ai/mlc-llm but not TVM repo.

BTW, mlc_chat is the old interface that has been deprecated by the latest mlc-llm, please check your version too.