Skip to content
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

在 SPU 中实现 Tensorflow 前端支持 #386

Open
Candicepan opened this issue Nov 2, 2023 · 2 comments
Open

在 SPU 中实现 Tensorflow 前端支持 #386

Candicepan opened this issue Nov 2, 2023 · 2 comments
Labels
challenge OSCP SecretFlow Open Source Contribution Plan

Comments

@Candicepan
Copy link
Contributor

Candicepan commented Nov 2, 2023

此 ISSUE 为 隐语开源共建计划(SecretFlow Open Source Contribution Plan,简称 SF OSCP)任务 ISSUE,欢迎社区开发者参与共建~
若有感兴趣想要认领的任务,但还未报名,辛苦先完成报名进行哈~

任务介绍

  • 任务名称:Tensorflow 前端支持
  • 技术方向:SPU/Frontend
  • 任务难度:挑战🌟🌟🌟
  • 任务期望完成时间:8 周

详细要求

SPU 对于 TensorFlow 的支持尚不完整,请验证更多的 TensorFlow 模型可以正确的在 SPU 上执行。
并尝试改善对于 tf.Tensor, tf.DataSet 等功能的支持程度

能力要求

  • 了解 TensorFlow
  • 熟悉 tf.function
  • 对 SPU 编译器前端有一定了解

操作说明

@Candicepan Candicepan added challenge OSCP SecretFlow Open Source Contribution Plan labels Nov 2, 2023
@Candicepan Candicepan moved this to Needs Triage in OSCP Phase 3 Nov 3, 2023
@Candicepan Candicepan removed this from OSCP Phase 3 Mar 8, 2024
@Candicepan Candicepan added this to OSCP Mar 8, 2024
@Candicepan Candicepan moved this to Needs Triage in OSCP Mar 8, 2024
@aoshujiaocheng
Copy link

/secretflow/device/device/spu.py还是只支持JAX吗?
运行TensorFlow前端时,spu.py文件line 1410 def _spu_compile(fn, copts, *meta_args, **meta_kwargs): 函数报错。
该函数下line 1427-1440
executable, output_tree = spu_fe.compile( **spu_fe.Kind.JAX**, fn, meta_args, meta_kwargs, input_name, input_vis, lambda output_flat: [ _generate_output_uuid() for _ in range(len(output_flat)) ], static_argnums=(), static_argnames=None, copts=copts, )
是否默认只能使用JAX作为frontend?

@anakinxc
Copy link
Contributor

/secretflow/device/device/spu.py还是只支持JAX吗? 运行TensorFlow前端时,spu.py文件line 1410 def _spu_compile(fn, copts, *meta_args, **meta_kwargs): 函数报错。 该函数下line 1427-1440 executable, output_tree = spu_fe.compile( **spu_fe.Kind.JAX**, fn, meta_args, meta_kwargs, input_name, input_vis, lambda output_flat: [ _generate_output_uuid() for _ in range(len(output_flat)) ], static_argnums=(), static_argnames=None, copts=copts, ) 是否默认只能使用JAX作为frontend?

spu 本身在能力支持上 jax 是第一梯队,torch 是第二梯队,tf 目前只是一个 pure demo 的能力。如果有兴趣和诉求,欢迎认领这个任务来完善 tf 的支持。

secretflow hardcode 前端类型的问题,请在 secretflow repo 内创建一个新的 issue 来 track 这个问题

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
challenge OSCP SecretFlow Open Source Contribution Plan
Projects
Status: Needs Triage
Development

No branches or pull requests

3 participants