Skip to content
This repository has been archived by the owner on Oct 25, 2024. It is now read-only.

Set limitation for model mapping #461

Closed
wants to merge 16 commits into from
Closed

Conversation

changwangss
Copy link
Contributor

@changwangss changwangss commented Oct 13, 2023

Type of Change

the mapping is only used for IPEX 2.0 smoothquant.
the root cause is

  1. the tensors on both sides of the plus sign cannot be recognized when tracing.
  2. example_inputs only support tuple, the order in tuple should be the same with forward parameters order.

If you use torch.jit.trace in other situations, you may also need this mapping, please let me know.

Description

change description:
Advance the order of processing sq_config and load the fp32 model with special processing for sq.

Expected Behavior & Potential Risk

the expected behavior that triggered by this PR

How has this PR been tested?

how to reproduce the test (including hardware information)

Dependency Change?

any library dependency introduced or removed

Signed-off-by: changwangss <chang1.wang@intel.com>
@changwangss
Copy link
Contributor Author

@VincyZhang the CI PR ut is canceled, it seems due to no space, please check.

@VincyZhang
Copy link
Contributor

@VincyZhang the CI PR ut is canceled, it seems due to no space, please check.

This is not due to space, just hangs, it's strange that can not import tensorflow after importing torch. See this issue, tensorflow/tensorflow#60109

@VincyZhang
Copy link
Contributor

@changwangss you can ask @PenghuiCheng for help, he solved this problem before.

@hshen14
Copy link
Contributor

hshen14 commented Oct 14, 2023

@changwangss you can ask @PenghuiCheng for help, he solved this problem before.

Lazy import?

changwangss and others added 11 commits October 16, 2023 13:19
Signed-off-by: changwangss <chang1.wang@intel.com>
Signed-off-by: Wenxin Zhang <wenxin.zhang@intel.com>
* Support StreamingLLM on CPU

Signed-off-by: zhenwei-intel <zhenwei.liu@intel.com>
Signed-off-by: changwangss <chang1.wang@intel.com>
Signed-off-by: Wenxin Zhang <wenxin.zhang@intel.com>
Signed-off-by: Wang, Chang <chang1.wang@intel.com>
Co-authored-by: Wenxin Zhang <wenxin.zhang@intel.com>
@changwangss
Copy link
Contributor Author

close, #568 merged.

Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.
Labels
bug Something isn't working
Projects
None yet
Development

Successfully merging this pull request may close these issues.

6 participants