-
Notifications
You must be signed in to change notification settings - Fork 100
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
尝试复现examples/python/ml/flax_gpt2报错 #245
Comments
Hi @Mingbo-Lee 请使用 gcc 11.2 |
使用gcc 11.2 后,再自行排除一些错误,问题已经解决,非常感谢 @anakinxc |
出现新的报错
安装最新版本的jax jaxlib 和 spu 都没用
jax&jaxlib == 0.4.12 /0.4.13 都报错 |
更新一下 flax 看看 |
更新最新版本flax,还是报错
|
有点奇怪…… cpu 的跑起来没问题?顺便问一下 grpc 的版本是多少? |
1.49.1
|
更新到了最新版本1.56.0,还是报错
|
这种方式很容易复现:https://www.secretflow.org.cn/docs/secretflow/latest/zh-Hans/tutorial/gpt2_with_spu |
CPU跑起来没问题:
|
SecretFlow 的这个 tutorial 是基于 sf&ray 的。。spu 的是基于 spu 自己实现的一个简单的 distributed framework |
好的 非常感谢 |
正在尝试复现。。。稍等哈 |
好的 |
Hi @Mingbo-Lee 我刚刚从头试了一下,没复现,我来描述一下我的 step 拉一个新的 secretflow/spu-ci:latest
pip install -r requirements.txt
pip install 'transformers[flax]'
bazel build //examples/python/... -c opt
找两个 terminals
第一个跑 bazel-bin/examples/python/utils/nodectl --config `pwd`/examples/python/ml/flax_gpt2/3pc.json up
第二个跑 bazel-bin/examples/python/ml/flax_gpt2/flax_gpt2 --config `pwd`/examples/python/ml/flax_gpt2/3pc.json output
pip list 结果如下
要不试一下新建一个新的 python env? |
我新建一个新的Python env, 成功复现,非常感谢
|
Issue Type
Bug
Modules Involved
Documentation/Tutorial/Example
Have you reproduced the bug with SPU HEAD?
Yes
Installation Kind
binary
SPU Version
spu 0.4.1b1
OS Platform and Distribution
Linux Ubuntu 20.04.6 LTS
Python Version
3.8.17
Compiler Version
GCC 9.4.0
Current Behavior?
我尝试复现examples/python/ml/flax_gpt2
按照 https://github.com/secretflow/spu/tree/main/examples/python/ml/flax_gpt2 内的指令一步步复现
执行到
bazel run -c opt //examples/python/utils:nodectl -- --config
pwd
/examples/python/ml/flax_gpt2/3pc.json up出现报错
Standalone code to reproduce the issue
Relevant log output
The text was updated successfully, but these errors were encountered: