You are viewing a plain text version of this content. The canonical link for it is here.

Posted to dev@tvm.apache.org by WANG HUNG-HSIANG via dev <de...@tvm.apache.org> on 2025/12/04 06:56:34 UTC

[apache/tvm] My posit dev (PR #18543)

You can view, comment on, or merge this pull request online at:

  https://github.com/apache/tvm/pull/18543

-- Commit Summary --

  * set(CMAKE_CXX_STANDARD 17 to 20)
  * Modify UNIVERSAL_PATH
  * Fix bugs in the original version
  * Use Hyper-threading
  * Bring Your Own Datatypes
  * Fix bug in constant folding for custom datatypes
  * Fix function name error
  * Fix sqrt require_float_dtype set to false
  * Add &quot;bring our own datatype&quot; test program
  * Comment pow datatype check
  * Add posit operation
  * Comment softmax datatype check
  * set unary arith op require_float_dtype to false
  * Unsupported numpy or ml_dtypes dtype(&#39;O&#39;) when importing ONNX model using Relax frontend
  * add _convert_struct_info

-- File Changes --

    A 3rdparty/universal (1)
    M CMakeLists.txt (2)
    M cmake/modules/contrib/Posit.cmake (2)
    A docs/how_to/tutorials/TinyLlama-1.1B-Chat-v1.0 (1)
    M docs/how_to/tutorials/optimize_llm.py (8)
    A docs/how_to/tutorials/register.py (138)
    A example/BYODT/register.py (275)
    A example/BYODT/sch_handed.py (116)
    A example/BYODT/test_posites2.py (148)
    A example/LLM/GPT-2/Compile_GPT2.py (259)
    A example/LLM/GPT-2/ToMixedPrecision.md (219)
    A example/LLM/GPT-2/benchmark_result.csv (4)
    A example/LLM/GPT-2/out_fp16 (883)
    A example/LLM/GPT-2/out_mixed (1289)
    A example/LLM/GPT-2/run_benchmark_tvm (22)
    A example/LLM/GPT-2/test.py (216)
    A example/LLM/GPT-2/test_comparison.py (395)
    A example/LLM/Qwen/Compile_Qwen.py (257)
    A example/LLM/Qwen/NUMERICAL_SENSITIVITY_REPORT.md (137)
    A example/LLM/Qwen/analyze_numerical_sensitivity.py (232)
    A example/LLM/Qwen/benchmark_result.csv (8)
    A example/LLM/Qwen/check_normalization_precision.py (134)
    A example/LLM/Qwen/root_cause_summary.py (187)
    A example/LLM/Qwen/run_benchmark_tvm (28)
    A example/LLM/Qwen/test.py (219)
    A example/LLM/Qwen/test_comparison.py (418)
    A example/LLM/Whisper/Compile_Decoder.py (213)
    A example/LLM/Whisper/Compile_Decoder_with_past.py (221)
    A example/LLM/Whisper/Compile_Encoder.py (211)
    A example/LLM/Whisper/benchmark_result.csv (3)
    A example/LLM/Whisper/data.npy (0)
    A example/LLM/Whisper/run_benchmark_tvm (20)
    A example/LLM/Whisper/test.py (305)
    A example/LLM/llama/Compile_Llama.py (257)
    A example/LLM/llama/benchmark_result.csv (17)
    A example/LLM/llama/run_benchmark_tvm (28)
    A example/LLM/llama/sch_handed.py (221)
    A example/LLM/llama/test.py (218)
    A example/LLM/llama/test_comparison.py (418)
    A python/tvm/relax/frontend/change_datatype.py (108)
    M python/tvm/relax/frontend/onnx/onnx_frontend.py (34)
    M python/tvm/relax/transform/legalize_ops/datatype.py (3)
    M python/tvm/target/datatype.py (7)
    M src/relax/op/nn/nn.cc (12)
    M src/relax/op/tensor/unary.cc (12)
    M src/runtime/threading_backend.cc (4)
    M src/target/datatype/posit/posit-wrapper.cc (447)
    M src/target/llvm/codegen_nvptx.cc (4)
    M src/tir/op/op.cc (2)

-- Patch Links --

https://github.com/apache/tvm/pull/18543.patch
https://github.com/apache/tvm/pull/18543.diff

-- 
Reply to this email directly or view it on GitHub:
https://github.com/apache/tvm/pull/18543
You are receiving this because you are subscribed to this thread.

Message ID: &lt;apache/tvm/pull/18543@github.com&gt;

Re: [apache/tvm] My posit dev (PR #18543)

Posted by "gemini-code-assist[bot] via dev" <de...@tvm.apache.org>.

gemini-code-assist[bot] left a comment (apache/tvm#18543)

> [!WARNING]
> Gemini encountered an error creating the review. You can try again by commenting `/gemini review`.

-- 
Reply to this email directly or view it on GitHub:
https://github.com/apache/tvm/pull/18543#issuecomment-3610595275
You are receiving this because you are subscribed to this thread.

Message ID: <ap...@github.com>

Re: [apache/tvm] My posit dev (PR #18543)

Posted by WANG HUNG-HSIANG via dev <de...@tvm.apache.org>.

Closed #18543.

-- 
Reply to this email directly or view it on GitHub:
https://github.com/apache/tvm/pull/18543#event-21351096967
You are receiving this because you are subscribed to this thread.

Message ID: <ap...@github.com>

Re: [apache/tvm] My posit dev (PR #18543)

Posted by "gemini-code-assist[bot] via dev" <de...@tvm.apache.org>.

gemini-code-assist[bot] left a comment (apache/tvm#18543)

## Summary of Changes

Hello @aa32801052, I'm Gemini Code Assist[^1]! I'm currently reviewing this pull request and will post my feedback shortly. In the meantime, here's a summary to help you and other reviewers quickly get up to speed!

This pull request introduces several key developments, including upgrading the C++ standard, integrating posit datatypes, updating tutorials, adding parallelization directives, providing BYODT examples, and enhancing the GPT-2 compilation process. These changes collectively aim to improve performance, enable custom data types, and provide better documentation and examples.

### Highlights

* **CMake Update**: The CMakeLists.txt file has been updated to use C++ standard 20.
* **Posit Integration**: The Posit.cmake module has been modified to correctly include the Universal path, and new files have been added to register posit datatypes and operations.
* **Tutorial Updates**: The optimize_llm.py tutorial has been updated to use the CPU device and a specific path for TinyLlama weights, and the TinyLlama tutorial was added.
* **Parallelization**: A new script sch_handed.py was added to introduce parallel directives to all functions.
* **BYODT Example**: New example files were added to demonstrate the usage of BYODT (Bring Your Own Datatype) with posit numbers, including registration and a test script.
* **GPT-2 Compilation**: The GPT-2 compilation script has been updated to include FP16 and mixed precision models, along with automatic optimization and posit compilation.
* **Mixed Precision Strategy**: A document describing the mixed precision strategy has been added.
* **Benchmark Results**: A CSV file containing benchmark results has been added.







<details>
<summary><b>Using Gemini Code Assist</b></summary>
<br>

The full guide for Gemini Code Assist can be found on our [documentation page](https://developers.google.com/gemini-code-assist/docs/review-github-code), here are some quick tips.

<b>Invoking Gemini</b>

You can request assistance from Gemini at any point by creating a comment using either `/gemini <command>` or `@gemini-code-assist <command>`. Below is a summary of the supported commands on the current page.

Feature | Command | Description
--- | --- | ---
Code Review | `/gemini review` | Performs a code review for the current pull request in its current state.
Pull Request Summary | `/gemini summary` | Provides a summary of the current pull request in its current state.
Comment | @gemini-code-assist | Responds in comments when explicitly tagged, both in pull request comments and review comments.
Help | `/gemini help` | Displays a list of available commands.

<b>Customization</b>

To customize Gemini Code Assist for GitHub experience, repository maintainers can create a configuration file and/or provide a custom code review style guide (such as PEP-8 for Python) by creating and adding files to a `.gemini/` folder in the base of the repository. Detailed instructions can be found [here](https://developers.google.com/gemini-code-assist/docs/customize-gemini-behavior-github).

<b>Limitations & Feedback</b>

Gemini Code Assist may make mistakes. Please leave feedback on any instances where its feedback is incorrect or counter productive. You can react with :thumbsup: and :thumbsdown: on @gemini-code-assist comments. If you're interested in giving your feedback about your experience with Gemini Code Assist for Github and other Google products, sign up [here](https://google.qualtrics.com/jfe/form/SV_2cyuGuTWsEw84yG).

<b>You can also get AI-powered code generation, chat, as well as code reviews directly in the IDE at no cost with the [Gemini Code Assist IDE Extension](https://cloud.google.com/products/gemini/code-assist).</b>
</details>




[^1]: Review the [Privacy Notices](https://policies.google.com/privacy), [Generative AI Prohibited Use Policy](https://policies.google.com/terms/generative-ai/use-policy), [Terms of Service](https://policies.google.com/terms), and learn how to configure Gemini Code Assist in GitHub [here](https://developers.google.com/gemini-code-assist/docs/customize-gemini-behavior-github). Gemini can make mistakes, so double check it and [use code with caution](https://support.google.com/legal/answer/13505487).


-- 
Reply to this email directly or view it on GitHub:
https://github.com/apache/tvm/pull/18543#issuecomment-3610584527
You are receiving this because you are subscribed to this thread.

Message ID: <ap...@github.com>