235B H20 configs #1132

shihaobai · 2025-12-03T07:26:31Z

No description provided.

gemini-code-assist · 2025-12-03T07:26:49Z

Summary of Changes

Hello @shihaobai, I'm Gemini Code Assist¹! 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 integrates new autotuned kernel configurations for the NVIDIA H20 GPU, enhancing the performance of several critical operations within the LightLLM framework. By providing optimized parameters for grouped matrix multiplication, Mixture of Experts (MoE) alignment and reduction, and SiLU activation, these additions aim to improve computational efficiency and throughput on the specified hardware.

Highlights

New Triton Kernel Configurations: This pull request introduces a comprehensive set of autotuned Triton kernel configurations specifically tailored for NVIDIA H20 GPUs.
Performance Optimization: The added configurations are designed to optimize the performance of critical operations, including grouped matrix multiplication, Mixture of Experts (MoE) alignment, MoE sum reduction, and SiLU activation with multiplication.
Expanded Hardware Support: These new configurations enhance support for the NVIDIA H20 GPU, providing pre-tuned parameters for various computational scenarios to improve efficiency.

Using Gemini Code Assist

The full guide for Gemini Code Assist can be found on our documentation page, here are some quick tips.

Invoking Gemini

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.

Customization

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.

Limitations & Feedback

Gemini Code Assist may make mistakes. Please leave feedback on any instances where its feedback is incorrect or counter productive. You can react with 👍 and 👎 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.

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.

Review the Privacy Notices, Generative AI Prohibited Use Policy, Terms of Service, and learn how to configure Gemini Code Assist in GitHub here. Gemini can make mistakes, so double check it and use code with caution. ↩

gemini-code-assist

Code Review

This pull request adds several new JSON configuration files for Triton kernels, specifically for the NVIDIA H20 architecture. The configurations seem appropriate for autotuning various operations. My main feedback is on the ordering of keys within these JSON files. Currently, they are sorted alphabetically, which makes them difficult to read since the keys represent numerical values. I have provided suggestions to reorder the keys numerically in each file to improve readability and maintainability. This change does not affect functionality but significantly enhances the developer experience when working with these configuration files.

gemini-code-assist · 2025-12-03T07:29:28Z

...l_routed_weight=true,out_dtype=torch.bfloat16,topk_num=1,use_fp8_w8a8=false}_NVIDIA_H20.json

+{
+  "1024": {
+    "BLOCK_SIZE_K": 32,
+    "BLOCK_SIZE_M": 16,
+    "BLOCK_SIZE_N": 128,
+    "GROUP_SIZE_M": 1,
+    "NEED_TRANS": false,
+    "num_stages": 4,
+    "num_warps": 4
+  },
+  "128": {
+    "BLOCK_SIZE_K": 32,
+    "BLOCK_SIZE_M": 16,
+    "BLOCK_SIZE_N": 128,
+    "GROUP_SIZE_M": 1,
+    "NEED_TRANS": false,
+    "num_stages": 4,
+    "num_warps": 4
+  },
+  "16384": {
+    "BLOCK_SIZE_K": 32,
+    "BLOCK_SIZE_M": 64,
+    "BLOCK_SIZE_N": 128,
+    "GROUP_SIZE_M": 16,
+    "NEED_TRANS": false,
+    "num_stages": 3,
+    "num_warps": 4
+  },
+  "2048": {
+    "BLOCK_SIZE_K": 32,
+    "BLOCK_SIZE_M": 16,
+    "BLOCK_SIZE_N": 128,
+    "GROUP_SIZE_M": 1,
+    "NEED_TRANS": false,
+    "num_stages": 4,
+    "num_warps": 4
+  },
+  "256": {
+    "BLOCK_SIZE_K": 32,
+    "BLOCK_SIZE_M": 16,
+    "BLOCK_SIZE_N": 128,
+    "GROUP_SIZE_M": 1,
+    "NEED_TRANS": false,
+    "num_stages": 4,
+    "num_warps": 4
+  },
+  "32768": {
+    "BLOCK_SIZE_K": 32,
+    "BLOCK_SIZE_M": 64,
+    "BLOCK_SIZE_N": 128,
+    "GROUP_SIZE_M": 16,
+    "NEED_TRANS": false,
+    "num_stages": 3,
+    "num_warps": 4
+  },
+  "512": {
+    "BLOCK_SIZE_K": 32,
+    "BLOCK_SIZE_M": 16,
+    "BLOCK_SIZE_N": 128,
+    "GROUP_SIZE_M": 1,
+    "NEED_TRANS": false,
+    "num_stages": 4,
+    "num_warps": 4
+  },
+  "64": {
+    "BLOCK_SIZE_K": 32,
+    "BLOCK_SIZE_M": 16,
+    "BLOCK_SIZE_N": 128,
+    "GROUP_SIZE_M": 1,
+    "NEED_TRANS": false,
+    "num_stages": 3,
+    "num_warps": 4
+  },
+  "65536": {
+    "BLOCK_SIZE_K": 32,
+    "BLOCK_SIZE_M": 64,
+    "BLOCK_SIZE_N": 128,
+    "GROUP_SIZE_M": 32,
+    "NEED_TRANS": false,
+    "num_stages": 3,
+    "num_warps": 4
+  },
+  "8": {
+    "BLOCK_SIZE_K": 32,
+    "BLOCK_SIZE_M": 16,
+    "BLOCK_SIZE_N": 128,
+    "GROUP_SIZE_M": 1,
+    "NEED_TRANS": false,
+    "num_stages": 3,
+    "num_warps": 4
+  },
+  "800": {
+    "BLOCK_SIZE_K": 32,
+    "BLOCK_SIZE_M": 16,
+    "BLOCK_SIZE_N": 128,
+    "GROUP_SIZE_M": 1,
+    "NEED_TRANS": false,
+    "num_stages": 4,
+    "num_warps": 4
+  },
+  "8192": {
+    "BLOCK_SIZE_K": 32,
+    "BLOCK_SIZE_M": 64,
+    "BLOCK_SIZE_N": 128,
+    "GROUP_SIZE_M": 16,
+    "NEED_TRANS": false,
+    "num_stages": 3,
+    "num_warps": 4
+  }
+}


The keys in this JSON file are sorted alphabetically, which makes it hard to read as they represent numerical values. For better readability and maintainability, please sort the keys numerically.

{ "8": { "BLOCK_SIZE_K": 32, "BLOCK_SIZE_M": 16, "BLOCK_SIZE_N": 128, "GROUP_SIZE_M": 1, "NEED_TRANS": false, "num_stages": 3, "num_warps": 4 }, "64": { "BLOCK_SIZE_K": 32, "BLOCK_SIZE_M": 16, "BLOCK_SIZE_N": 128, "GROUP_SIZE_M": 1, "NEED_TRANS": false, "num_stages": 3, "num_warps": 4 }, "128": { "BLOCK_SIZE_K": 32, "BLOCK_SIZE_M": 16, "BLOCK_SIZE_N": 128, "GROUP_SIZE_M": 1, "NEED_TRANS": false, "num_stages": 4, "num_warps": 4 }, "256": { "BLOCK_SIZE_K": 32, "BLOCK_SIZE_M": 16, "BLOCK_SIZE_N": 128, "GROUP_SIZE_M": 1, "NEED_TRANS": false, "num_stages": 4, "num_warps": 4 }, "512": { "BLOCK_SIZE_K": 32, "BLOCK_SIZE_M": 16, "BLOCK_SIZE_N": 128, "GROUP_SIZE_M": 1, "NEED_TRANS": false, "num_stages": 4, "num_warps": 4 }, "800": { "BLOCK_SIZE_K": 32, "BLOCK_SIZE_M": 16, "BLOCK_SIZE_N": 128, "GROUP_SIZE_M": 1, "NEED_TRANS": false, "num_stages": 4, "num_warps": 4 }, "1024": { "BLOCK_SIZE_K": 32, "BLOCK_SIZE_M": 16, "BLOCK_SIZE_N": 128, "GROUP_SIZE_M": 1, "NEED_TRANS": false, "num_stages": 4, "num_warps": 4 }, "2048": { "BLOCK_SIZE_K": 32, "BLOCK_SIZE_M": 16, "BLOCK_SIZE_N": 128, "GROUP_SIZE_M": 1, "NEED_TRANS": false, "num_stages": 4, "num_warps": 4 }, "8192": { "BLOCK_SIZE_K": 32, "BLOCK_SIZE_M": 64, "BLOCK_SIZE_N": 128, "GROUP_SIZE_M": 16, "NEED_TRANS": false, "num_stages": 3, "num_warps": 4 }, "16384": { "BLOCK_SIZE_K": 32, "BLOCK_SIZE_M": 64, "BLOCK_SIZE_N": 128, "GROUP_SIZE_M": 16, "NEED_TRANS": false, "num_stages": 3, "num_warps": 4 }, "32768": { "BLOCK_SIZE_K": 32, "BLOCK_SIZE_M": 64, "BLOCK_SIZE_N": 128, "GROUP_SIZE_M": 16, "NEED_TRANS": false, "num_stages": 3, "num_warps": 4 }, "65536": { "BLOCK_SIZE_K": 32, "BLOCK_SIZE_M": 64, "BLOCK_SIZE_N": 128, "GROUP_SIZE_M": 32, "NEED_TRANS": false, "num_stages": 3, "num_warps": 4 } }

gemini-code-assist · 2025-12-03T07:29:28Z

..._routed_weight=false,out_dtype=torch.bfloat16,topk_num=8,use_fp8_w8a8=false}_NVIDIA_H20.json

+{
+  "1": {
+    "BLOCK_SIZE_K": 128,
+    "BLOCK_SIZE_M": 16,
+    "BLOCK_SIZE_N": 64,
+    "GROUP_SIZE_M": 1,
+    "NEED_TRANS": false,
+    "num_stages": 4,
+    "num_warps": 4
+  },
+  "100": {
+    "BLOCK_SIZE_K": 128,
+    "BLOCK_SIZE_M": 16,
+    "BLOCK_SIZE_N": 64,
+    "GROUP_SIZE_M": 1,
+    "NEED_TRANS": false,
+    "num_stages": 3,
+    "num_warps": 4
+  },
+  "1024": {
+    "BLOCK_SIZE_K": 64,
+    "BLOCK_SIZE_M": 64,
+    "BLOCK_SIZE_N": 64,
+    "GROUP_SIZE_M": 64,
+    "NEED_TRANS": false,
+    "num_stages": 4,
+    "num_warps": 4
+  },
+  "128": {
+    "BLOCK_SIZE_K": 128,
+    "BLOCK_SIZE_M": 16,
+    "BLOCK_SIZE_N": 64,
+    "GROUP_SIZE_M": 1,
+    "NEED_TRANS": false,
+    "num_stages": 3,
+    "num_warps": 4
+  },
+  "16": {
+    "BLOCK_SIZE_K": 128,
+    "BLOCK_SIZE_M": 16,
+    "BLOCK_SIZE_N": 32,
+    "GROUP_SIZE_M": 1,
+    "NEED_TRANS": false,
+    "num_stages": 3,
+    "num_warps": 4
+  },
+  "2048": {
+    "BLOCK_SIZE_K": 64,
+    "BLOCK_SIZE_M": 64,
+    "BLOCK_SIZE_N": 64,
+    "GROUP_SIZE_M": 64,
+    "NEED_TRANS": false,
+    "num_stages": 4,
+    "num_warps": 4
+  },
+  "256": {
+    "BLOCK_SIZE_K": 128,
+    "BLOCK_SIZE_M": 16,
+    "BLOCK_SIZE_N": 64,
+    "GROUP_SIZE_M": 64,
+    "NEED_TRANS": false,
+    "num_stages": 3,
+    "num_warps": 4
+  },
+  "32": {
+    "BLOCK_SIZE_K": 128,
+    "BLOCK_SIZE_M": 16,
+    "BLOCK_SIZE_N": 64,
+    "GROUP_SIZE_M": 1,
+    "NEED_TRANS": false,
+    "num_stages": 4,
+    "num_warps": 4
+  },
+  "4096": {
+    "BLOCK_SIZE_K": 64,
+    "BLOCK_SIZE_M": 64,
+    "BLOCK_SIZE_N": 64,
+    "GROUP_SIZE_M": 64,
+    "NEED_TRANS": false,
+    "num_stages": 4,
+    "num_warps": 4
+  },
+  "64": {
+    "BLOCK_SIZE_K": 128,
+    "BLOCK_SIZE_M": 16,
+    "BLOCK_SIZE_N": 64,
+    "GROUP_SIZE_M": 1,
+    "NEED_TRANS": false,
+    "num_stages": 3,
+    "num_warps": 4
+  },
+  "8": {
+    "BLOCK_SIZE_K": 128,
+    "BLOCK_SIZE_M": 16,
+    "BLOCK_SIZE_N": 64,
+    "GROUP_SIZE_M": 32,
+    "NEED_TRANS": false,
+    "num_stages": 3,
+    "num_warps": 4
+  },
+  "8192": {
+    "BLOCK_SIZE_K": 64,
+    "BLOCK_SIZE_M": 64,
+    "BLOCK_SIZE_N": 64,
+    "GROUP_SIZE_M": 64,
+    "NEED_TRANS": false,
+    "num_stages": 4,
+    "num_warps": 4
+  }
+}


The keys in this JSON file are sorted alphabetically, which makes it hard to read as they represent numerical values. For better readability and maintainability, please sort the keys numerically.

{ "1": { "BLOCK_SIZE_K": 128, "BLOCK_SIZE_M": 16, "BLOCK_SIZE_N": 64, "GROUP_SIZE_M": 1, "NEED_TRANS": false, "num_stages": 4, "num_warps": 4 }, "8": { "BLOCK_SIZE_K": 128, "BLOCK_SIZE_M": 16, "BLOCK_SIZE_N": 64, "GROUP_SIZE_M": 32, "NEED_TRANS": false, "num_stages": 3, "num_warps": 4 }, "16": { "BLOCK_SIZE_K": 128, "BLOCK_SIZE_M": 16, "BLOCK_SIZE_N": 32, "GROUP_SIZE_M": 1, "NEED_TRANS": false, "num_stages": 3, "num_warps": 4 }, "32": { "BLOCK_SIZE_K": 128, "BLOCK_SIZE_M": 16, "BLOCK_SIZE_N": 64, "GROUP_SIZE_M": 1, "NEED_TRANS": false, "num_stages": 4, "num_warps": 4 }, "64": { "BLOCK_SIZE_K": 128, "BLOCK_SIZE_M": 16, "BLOCK_SIZE_N": 64, "GROUP_SIZE_M": 1, "NEED_TRANS": false, "num_stages": 3, "num_warps": 4 }, "100": { "BLOCK_SIZE_K": 128, "BLOCK_SIZE_M": 16, "BLOCK_SIZE_N": 64, "GROUP_SIZE_M": 1, "NEED_TRANS": false, "num_stages": 3, "num_warps": 4 }, "128": { "BLOCK_SIZE_K": 128, "BLOCK_SIZE_M": 16, "BLOCK_SIZE_N": 64, "GROUP_SIZE_M": 1, "NEED_TRANS": false, "num_stages": 3, "num_warps": 4 }, "256": { "BLOCK_SIZE_K": 128, "BLOCK_SIZE_M": 16, "BLOCK_SIZE_N": 64, "GROUP_SIZE_M": 64, "NEED_TRANS": false, "num_stages": 3, "num_warps": 4 }, "1024": { "BLOCK_SIZE_K": 64, "BLOCK_SIZE_M": 64, "BLOCK_SIZE_N": 64, "GROUP_SIZE_M": 64, "NEED_TRANS": false, "num_stages": 4, "num_warps": 4 }, "2048": { "BLOCK_SIZE_K": 64, "BLOCK_SIZE_M": 64, "BLOCK_SIZE_N": 64, "GROUP_SIZE_M": 64, "NEED_TRANS": false, "num_stages": 4, "num_warps": 4 }, "4096": { "BLOCK_SIZE_K": 64, "BLOCK_SIZE_M": 64, "BLOCK_SIZE_N": 64, "GROUP_SIZE_M": 64, "NEED_TRANS": false, "num_stages": 4, "num_warps": 4 }, "8192": { "BLOCK_SIZE_K": 64, "BLOCK_SIZE_M": 64, "BLOCK_SIZE_N": 64, "GROUP_SIZE_M": 64, "NEED_TRANS": false, "num_stages": 4, "num_warps": 4 } }

gemini-code-assist · 2025-12-03T07:29:28Z

...otune_kernel_configs/triton_3.4.0/NVIDIA_H20/moe_align_fused:v1/{topk_num=8}_NVIDIA_H20.json

+{
+  "1": {
+    "BLOCK_SIZE": 256,
+    "num_warps": 8
+  },
+  "100": {
+    "BLOCK_SIZE": 128,
+    "num_warps": 8
+  },
+  "1024": {
+    "BLOCK_SIZE": 128,
+    "num_warps": 8
+  },
+  "128": {
+    "BLOCK_SIZE": 128,
+    "num_warps": 8
+  },
+  "16": {
+    "BLOCK_SIZE": 256,
+    "num_warps": 8
+  },
+  "2048": {
+    "BLOCK_SIZE": 128,
+    "num_warps": 4
+  },
+  "256": {
+    "BLOCK_SIZE": 128,
+    "num_warps": 8
+  },
+  "32": {
+    "BLOCK_SIZE": 512,
+    "num_warps": 8
+  },
+  "4096": {
+    "BLOCK_SIZE": 128,
+    "num_warps": 8
+  },
+  "64": {
+    "BLOCK_SIZE": 128,
+    "num_warps": 8
+  },
+  "8": {
+    "BLOCK_SIZE": 256,
+    "num_warps": 8
+  },
+  "8192": {
+    "BLOCK_SIZE": 128,
+    "num_warps": 8
+  }
+}


The keys in this JSON file are sorted alphabetically, which makes it hard to read as they represent numerical values. For better readability and maintainability, please sort the keys numerically.

{ "1": { "BLOCK_SIZE": 256, "num_warps": 8 }, "8": { "BLOCK_SIZE": 256, "num_warps": 8 }, "16": { "BLOCK_SIZE": 256, "num_warps": 8 }, "32": { "BLOCK_SIZE": 512, "num_warps": 8 }, "64": { "BLOCK_SIZE": 128, "num_warps": 8 }, "100": { "BLOCK_SIZE": 128, "num_warps": 8 }, "128": { "BLOCK_SIZE": 128, "num_warps": 8 }, "256": { "BLOCK_SIZE": 128, "num_warps": 8 }, "1024": { "BLOCK_SIZE": 128, "num_warps": 8 }, "2048": { "BLOCK_SIZE": 128, "num_warps": 4 }, "4096": { "BLOCK_SIZE": 128, "num_warps": 8 }, "8192": { "BLOCK_SIZE": 128, "num_warps": 8 } }

gemini-code-assist · 2025-12-03T07:29:28Z

..._H20/moe_sum_reduce:v1/{hidden_dim=4096,out_dtype=torch.bfloat16,topk_num=8}_NVIDIA_H20.json

+{
+  "1": {
+    "BLOCK_DIM": 64,
+    "BLOCK_M": 4,
+    "NUM_STAGE": 2,
+    "num_warps": 8
+  },
+  "100": {
+    "BLOCK_DIM": 1024,
+    "BLOCK_M": 1,
+    "NUM_STAGE": 1,
+    "num_warps": 8
+  },
+  "1024": {
+    "BLOCK_DIM": 256,
+    "BLOCK_M": 1,
+    "NUM_STAGE": 1,
+    "num_warps": 2
+  },
+  "128": {
+    "BLOCK_DIM": 1024,
+    "BLOCK_M": 1,
+    "NUM_STAGE": 1,
+    "num_warps": 8
+  },
+  "16": {
+    "BLOCK_DIM": 512,
+    "BLOCK_M": 1,
+    "NUM_STAGE": 1,
+    "num_warps": 8
+  },
+  "2048": {
+    "BLOCK_DIM": 1024,
+    "BLOCK_M": 1,
+    "NUM_STAGE": 4,
+    "num_warps": 2
+  },
+  "256": {
+    "BLOCK_DIM": 512,
+    "BLOCK_M": 1,
+    "NUM_STAGE": 1,
+    "num_warps": 2
+  },
+  "32": {
+    "BLOCK_DIM": 1024,
+    "BLOCK_M": 1,
+    "NUM_STAGE": 1,
+    "num_warps": 8
+  },
+  "4096": {
+    "BLOCK_DIM": 1024,
+    "BLOCK_M": 1,
+    "NUM_STAGE": 4,
+    "num_warps": 2
+  },
+  "64": {
+    "BLOCK_DIM": 1024,
+    "BLOCK_M": 1,
+    "NUM_STAGE": 1,
+    "num_warps": 8
+  },
+  "8": {
+    "BLOCK_DIM": 512,
+    "BLOCK_M": 1,
+    "NUM_STAGE": 4,
+    "num_warps": 16
+  },
+  "8192": {
+    "BLOCK_DIM": 1024,
+    "BLOCK_M": 1,
+    "NUM_STAGE": 4,
+    "num_warps": 1
+  }
+}


The keys in this JSON file are sorted alphabetically, which makes it hard to read as they represent numerical values. For better readability and maintainability, please sort the keys numerically.

{ "1": { "BLOCK_DIM": 64, "BLOCK_M": 4, "NUM_STAGE": 2, "num_warps": 8 }, "8": { "BLOCK_DIM": 512, "BLOCK_M": 1, "NUM_STAGE": 4, "num_warps": 16 }, "16": { "BLOCK_DIM": 512, "BLOCK_M": 1, "NUM_STAGE": 1, "num_warps": 8 }, "32": { "BLOCK_DIM": 1024, "BLOCK_M": 1, "NUM_STAGE": 1, "num_warps": 8 }, "64": { "BLOCK_DIM": 1024, "BLOCK_M": 1, "NUM_STAGE": 1, "num_warps": 8 }, "100": { "BLOCK_DIM": 1024, "BLOCK_M": 1, "NUM_STAGE": 1, "num_warps": 8 }, "128": { "BLOCK_DIM": 1024, "BLOCK_M": 1, "NUM_STAGE": 1, "num_warps": 8 }, "256": { "BLOCK_DIM": 512, "BLOCK_M": 1, "NUM_STAGE": 1, "num_warps": 2 }, "1024": { "BLOCK_DIM": 256, "BLOCK_M": 1, "NUM_STAGE": 1, "num_warps": 2 }, "2048": { "BLOCK_DIM": 1024, "BLOCK_M": 1, "NUM_STAGE": 4, "num_warps": 2 }, "4096": { "BLOCK_DIM": 1024, "BLOCK_M": 1, "NUM_STAGE": 4, "num_warps": 2 }, "8192": { "BLOCK_DIM": 1024, "BLOCK_M": 1, "NUM_STAGE": 4, "num_warps": 1 } }

gemini-code-assist · 2025-12-03T07:29:29Z

...riton_3.4.0/NVIDIA_H20/silu_and_mul_fwd:v1/{N=1536,out_dtype=torch.bfloat16}_NVIDIA_H20.json

+{
+  "1": {
+    "BLOCK_M": 256,
+    "BLOCK_N": 32,
+    "NUM_STAGES": 1,
+    "num_warps": 1
+  },
+  "1024": {
+    "BLOCK_M": 8,
+    "BLOCK_N": 256,
+    "NUM_STAGES": 4,
+    "num_warps": 1
+  },
+  "1152": {
+    "BLOCK_M": 8,
+    "BLOCK_N": 256,
+    "NUM_STAGES": 1,
+    "num_warps": 1
+  },
+  "128": {
+    "BLOCK_M": 1,
+    "BLOCK_N": 256,
+    "NUM_STAGES": 2,
+    "num_warps": 1
+  },
+  "1280": {
+    "BLOCK_M": 8,
+    "BLOCK_N": 128,
+    "NUM_STAGES": 1,
+    "num_warps": 1
+  },
+  "1408": {
+    "BLOCK_M": 8,
+    "BLOCK_N": 256,
+    "NUM_STAGES": 2,
+    "num_warps": 1
+  },
+  "1536": {
+    "BLOCK_M": 8,
+    "BLOCK_N": 128,
+    "NUM_STAGES": 1,
+    "num_warps": 1
+  },
+  "1664": {
+    "BLOCK_M": 8,
+    "BLOCK_N": 256,
+    "NUM_STAGES": 1,
+    "num_warps": 1
+  },
+  "16768": {
+    "BLOCK_M": 1,
+    "BLOCK_N": 256,
+    "NUM_STAGES": 4,
+    "num_warps": 1
+  },
+  "17024": {
+    "BLOCK_M": 1,
+    "BLOCK_N": 256,
+    "NUM_STAGES": 4,
+    "num_warps": 1
+  },
+  "17152": {
+    "BLOCK_M": 1,
+    "BLOCK_N": 256,
+    "NUM_STAGES": 4,
+    "num_warps": 1
+  },
+  "17280": {
+    "BLOCK_M": 1,
+    "BLOCK_N": 256,
+    "NUM_STAGES": 4,
+    "num_warps": 1
+  },
+  "17536": {
+    "BLOCK_M": 1,
+    "BLOCK_N": 256,
+    "NUM_STAGES": 4,
+    "num_warps": 1
+  },
+  "17792": {
+    "BLOCK_M": 1,
+    "BLOCK_N": 256,
+    "NUM_STAGES": 4,
+    "num_warps": 1
+  },
+  "1792": {
+    "BLOCK_M": 8,
+    "BLOCK_N": 256,
+    "NUM_STAGES": 1,
+    "num_warps": 1
+  },
+  "1920": {
+    "BLOCK_M": 8,
+    "BLOCK_N": 256,
+    "NUM_STAGES": 1,
+    "num_warps": 1
+  },
+  "2048": {
+    "BLOCK_M": 8,
+    "BLOCK_N": 256,
+    "NUM_STAGES": 1,
+    "num_warps": 1
+  },
+  "256": {
+    "BLOCK_M": 1,
+    "BLOCK_N": 256,
+    "NUM_STAGES": 1,
+    "num_warps": 1
+  },
+  "2688": {
+    "BLOCK_M": 8,
+    "BLOCK_N": 256,
+    "NUM_STAGES": 4,
+    "num_warps": 1
+  },
+  "2816": {
+    "BLOCK_M": 8,
+    "BLOCK_N": 256,
+    "NUM_STAGES": 4,
+    "num_warps": 1
+  },
+  "2944": {
+    "BLOCK_M": 8,
+    "BLOCK_N": 256,
+    "NUM_STAGES": 1,
+    "num_warps": 1
+  },
+  "3072": {
+    "BLOCK_M": 8,
+    "BLOCK_N": 256,
+    "NUM_STAGES": 4,
+    "num_warps": 1
+  },
+  "3200": {
+    "BLOCK_M": 8,
+    "BLOCK_N": 256,
+    "NUM_STAGES": 4,
+    "num_warps": 1
+  },
+  "32768": {
+    "BLOCK_M": 1,
+    "BLOCK_N": 256,
+    "NUM_STAGES": 4,
+    "num_warps": 1
+  },
+  "33280": {
+    "BLOCK_M": 8,
+    "BLOCK_N": 256,
+    "NUM_STAGES": 4,
+    "num_warps": 1
+  },
+  "33536": {
+    "BLOCK_M": 8,
+    "BLOCK_N": 256,
+    "NUM_STAGES": 4,
+    "num_warps": 1
+  },
+  "33792": {
+    "BLOCK_M": 8,
+    "BLOCK_N": 256,
+    "NUM_STAGES": 4,
+    "num_warps": 1
+  },
+  "33920": {
+    "BLOCK_M": 8,
+    "BLOCK_N": 256,
+    "NUM_STAGES": 4,
+    "num_warps": 1
+  },
+  "34048": {
+    "BLOCK_M": 8,
+    "BLOCK_N": 256,
+    "NUM_STAGES": 4,
+    "num_warps": 1
+  },
+  "34176": {
+    "BLOCK_M": 1,
+    "BLOCK_N": 256,
+    "NUM_STAGES": 4,
+    "num_warps": 1
+  },
+  "512": {
+    "BLOCK_M": 8,
+    "BLOCK_N": 256,
+    "NUM_STAGES": 4,
+    "num_warps": 1
+  },
+  "64768": {
+    "BLOCK_M": 1,
+    "BLOCK_N": 256,
+    "NUM_STAGES": 4,
+    "num_warps": 1
+  },
+  "65792": {
+    "BLOCK_M": 8,
+    "BLOCK_N": 256,
+    "NUM_STAGES": 4,
+    "num_warps": 1
+  },
+  "66176": {
+    "BLOCK_M": 8,
+    "BLOCK_N": 256,
+    "NUM_STAGES": 4,
+    "num_warps": 1
+  },
+  "66560": {
+    "BLOCK_M": 8,
+    "BLOCK_N": 256,
+    "NUM_STAGES": 4,
+    "num_warps": 1
+  },
+  "66816": {
+    "BLOCK_M": 8,
+    "BLOCK_N": 256,
+    "NUM_STAGES": 4,
+    "num_warps": 1
+  },
+  "67328": {
+    "BLOCK_M": 8,
+    "BLOCK_N": 256,
+    "NUM_STAGES": 4,
+    "num_warps": 1
+  },
+  "68352": {
+    "BLOCK_M": 8,
+    "BLOCK_N": 256,
+    "NUM_STAGES": 4,
+    "num_warps": 1
+  },
+  "768": {
+    "BLOCK_M": 8,
+    "BLOCK_N": 256,
+    "NUM_STAGES": 4,
+    "num_warps": 1
+  },
+  "8576": {
+    "BLOCK_M": 8,
+    "BLOCK_N": 256,
+    "NUM_STAGES": 4,
+    "num_warps": 1
+  },
+  "8832": {
+    "BLOCK_M": 32,
+    "BLOCK_N": 256,
+    "NUM_STAGES": 4,
+    "num_warps": 1
+  },
+  "896": {
+    "BLOCK_M": 8,
+    "BLOCK_N": 256,
+    "NUM_STAGES": 1,
+    "num_warps": 4
+  },
+  "9216": {
+    "BLOCK_M": 32,
+    "BLOCK_N": 256,
+    "NUM_STAGES": 4,
+    "num_warps": 1
+  },
+  "9344": {
+    "BLOCK_M": 32,
+    "BLOCK_N": 256,
+    "NUM_STAGES": 4,
+    "num_warps": 1
+  },
+  "9472": {
+    "BLOCK_M": 8,
+    "BLOCK_N": 256,
+    "NUM_STAGES": 4,
+    "num_warps": 1
+  },
+  "9600": {
+    "BLOCK_M": 8,
+    "BLOCK_N": 256,
+    "NUM_STAGES": 4,
+    "num_warps": 1
+  },
+  "9856": {
+    "BLOCK_M": 32,
+    "BLOCK_N": 256,
+    "NUM_STAGES": 4,
+    "num_warps": 1
+  }
+}


The keys in this JSON file are sorted alphabetically, which makes it hard to read as they represent numerical values. For better readability and maintainability, please sort the keys numerically.

{ "1": { "BLOCK_M": 256, "BLOCK_N": 32, "NUM_STAGES": 1, "num_warps": 1 }, "128": { "BLOCK_M": 1, "BLOCK_N": 256, "NUM_STAGES": 2, "num_warps": 1 }, "256": { "BLOCK_M": 1, "BLOCK_N": 256, "NUM_STAGES": 1, "num_warps": 1 }, "512": { "BLOCK_M": 8, "BLOCK_N": 256, "NUM_STAGES": 4, "num_warps": 1 }, "768": { "BLOCK_M": 8, "BLOCK_N": 256, "NUM_STAGES": 4, "num_warps": 1 }, "896": { "BLOCK_M": 8, "BLOCK_N": 256, "NUM_STAGES": 1, "num_warps": 4 }, "1024": { "BLOCK_M": 8, "BLOCK_N": 256, "NUM_STAGES": 4, "num_warps": 1 }, "1152": { "BLOCK_M": 8, "BLOCK_N": 256, "NUM_STAGES": 1, "num_warps": 1 }, "1280": { "BLOCK_M": 8, "BLOCK_N": 128, "NUM_STAGES": 1, "num_warps": 1 }, "1408": { "BLOCK_M": 8, "BLOCK_N": 256, "NUM_STAGES": 2, "num_warps": 1 }, "1536": { "BLOCK_M": 8, "BLOCK_N": 128, "NUM_STAGES": 1, "num_warps": 1 }, "1664": { "BLOCK_M": 8, "BLOCK_N": 256, "NUM_STAGES": 1, "num_warps": 1 }, "1792": { "BLOCK_M": 8, "BLOCK_N": 256, "NUM_STAGES": 1, "num_warps": 1 }, "1920": { "BLOCK_M": 8, "BLOCK_N": 256, "NUM_STAGES": 1, "num_warps": 1 }, "2048": { "BLOCK_M": 8, "BLOCK_N": 256, "NUM_STAGES": 1, "num_warps": 1 }, "2688": { "BLOCK_M": 8, "BLOCK_N": 256, "NUM_STAGES": 4, "num_warps": 1 }, "2816": { "BLOCK_M": 8, "BLOCK_N": 256, "NUM_STAGES": 4, "num_warps": 1 }, "2944": { "BLOCK_M": 8, "BLOCK_N": 256, "NUM_STAGES": 1, "num_warps": 1 }, "3072": { "BLOCK_M": 8, "BLOCK_N": 256, "NUM_STAGES": 4, "num_warps": 1 }, "3200": { "BLOCK_M": 8, "BLOCK_N": 256, "NUM_STAGES": 4, "num_warps": 1 }, "8576": { "BLOCK_M": 8, "BLOCK_N": 256, "NUM_STAGES": 4, "num_warps": 1 }, "8832": { "BLOCK_M": 32, "BLOCK_N": 256, "NUM_STAGES": 4, "num_warps": 1 }, "9216": { "BLOCK_M": 32, "BLOCK_N": 256, "NUM_STAGES": 4, "num_warps": 1 }, "9344": { "BLOCK_M": 32, "BLOCK_N": 256, "NUM_STAGES": 4, "num_warps": 1 }, "9472": { "BLOCK_M": 8, "BLOCK_N": 256, "NUM_STAGES": 4, "num_warps": 1 }, "9600": { "BLOCK_M": 8, "BLOCK_N": 256, "NUM_STAGES": 4, "num_warps": 1 }, "9856": { "BLOCK_M": 32, "BLOCK_N": 256, "NUM_STAGES": 4, "num_warps": 1 }, "16768": { "BLOCK_M": 1, "BLOCK_N": 256, "NUM_STAGES": 4, "num_warps": 1 }, "17024": { "BLOCK_M": 1, "BLOCK_N": 256, "NUM_STAGES": 4, "num_warps": 1 }, "17152": { "BLOCK_M": 1, "BLOCK_N": 256, "NUM_STAGES": 4, "num_warps": 1 }, "17280": { "BLOCK_M": 1, "BLOCK_N": 256, "NUM_STAGES": 4, "num_warps": 1 }, "17536": { "BLOCK_M": 1, "BLOCK_N": 256, "NUM_STAGES": 4, "num_warps": 1 }, "17792": { "BLOCK_M": 1, "BLOCK_N": 256, "NUM_STAGES": 4, "num_warps": 1 }, "32768": { "BLOCK_M": 1, "BLOCK_N": 256, "NUM_STAGES": 4, "num_warps": 1 }, "33280": { "BLOCK_M": 8, "BLOCK_N": 256, "NUM_STAGES": 4, "num_warps": 1 }, "33536": { "BLOCK_M": 8, "BLOCK_N": 256, "NUM_STAGES": 4, "num_warps": 1 }, "33792": { "BLOCK_M": 8, "BLOCK_N": 256, "NUM_STAGES": 4, "num_warps": 1 }, "33920": { "BLOCK_M": 8, "BLOCK_N": 256, "NUM_STAGES": 4, "num_warps": 1 }, "34048": { "BLOCK_M": 8, "BLOCK_N": 256, "NUM_STAGES": 4, "num_warps": 1 }, "34176": { "BLOCK_M": 1, "BLOCK_N": 256, "NUM_STAGES": 4, "num_warps": 1 }, "64768": { "BLOCK_M": 1, "BLOCK_N": 256, "NUM_STAGES": 4, "num_warps": 1 }, "65792": { "BLOCK_M": 8, "BLOCK_N": 256, "NUM_STAGES": 4, "num_warps": 1 }, "66176": { "BLOCK_M": 8, "BLOCK_N": 256, "NUM_STAGES": 4, "num_warps": 1 }, "66560": { "BLOCK_M": 8, "BLOCK_N": 256, "NUM_STAGES": 4, "num_warps": 1 }, "66816": { "BLOCK_M": 8, "BLOCK_N": 256, "NUM_STAGES": 4, "num_warps": 1 }, "67328": { "BLOCK_M": 8, "BLOCK_N": 256, "NUM_STAGES": 4, "num_warps": 1 }, "68352": { "BLOCK_M": 8, "BLOCK_N": 256, "NUM_STAGES": 4, "num_warps": 1 } }

gemini-code-assist · 2025-12-03T07:29:29Z

...triton_3.4.0/NVIDIA_H20/silu_and_mul_fwd:v1/{N=192,out_dtype=torch.bfloat16}_NVIDIA_H20.json

+{
+  "1024": {
+    "BLOCK_M": 1,
+    "BLOCK_N": 256,
+    "NUM_STAGES": 1,
+    "num_warps": 1
+  },
+  "128": {
+    "BLOCK_M": 1,
+    "BLOCK_N": 256,
+    "NUM_STAGES": 2,
+    "num_warps": 8
+  },
+  "16384": {
+    "BLOCK_M": 1,
+    "BLOCK_N": 256,
+    "NUM_STAGES": 1,
+    "num_warps": 1
+  },
+  "2048": {
+    "BLOCK_M": 8,
+    "BLOCK_N": 64,
+    "NUM_STAGES": 4,
+    "num_warps": 1
+  },
+  "256": {
+    "BLOCK_M": 1,
+    "BLOCK_N": 256,
+    "NUM_STAGES": 2,
+    "num_warps": 8
+  },
+  "32768": {
+    "BLOCK_M": 8,
+    "BLOCK_N": 256,
+    "NUM_STAGES": 4,
+    "num_warps": 1
+  },
+  "512": {
+    "BLOCK_M": 1,
+    "BLOCK_N": 256,
+    "NUM_STAGES": 2,
+    "num_warps": 1
+  },
+  "64": {
+    "BLOCK_M": 1,
+    "BLOCK_N": 128,
+    "NUM_STAGES": 1,
+    "num_warps": 4
+  },
+  "65536": {
+    "BLOCK_M": 32,
+    "BLOCK_N": 256,
+    "NUM_STAGES": 4,
+    "num_warps": 1
+  },
+  "8": {
+    "BLOCK_M": 1,
+    "BLOCK_N": 32,
+    "NUM_STAGES": 1,
+    "num_warps": 1
+  },
+  "800": {
+    "BLOCK_M": 1,
+    "BLOCK_N": 256,
+    "NUM_STAGES": 2,
+    "num_warps": 1
+  },
+  "8192": {
+    "BLOCK_M": 8,
+    "BLOCK_N": 128,
+    "NUM_STAGES": 1,
+    "num_warps": 1
+  }
+}


The keys in this JSON file are sorted alphabetically, which makes it hard to read as they represent numerical values. For better readability and maintainability, please sort the keys numerically.

{ "8": { "BLOCK_M": 1, "BLOCK_N": 32, "NUM_STAGES": 1, "num_warps": 1 }, "64": { "BLOCK_M": 1, "BLOCK_N": 128, "NUM_STAGES": 1, "num_warps": 4 }, "128": { "BLOCK_M": 1, "BLOCK_N": 256, "NUM_STAGES": 2, "num_warps": 8 }, "256": { "BLOCK_M": 1, "BLOCK_N": 256, "NUM_STAGES": 2, "num_warps": 8 }, "512": { "BLOCK_M": 1, "BLOCK_N": 256, "NUM_STAGES": 2, "num_warps": 1 }, "800": { "BLOCK_M": 1, "BLOCK_N": 256, "NUM_STAGES": 2, "num_warps": 1 }, "1024": { "BLOCK_M": 1, "BLOCK_N": 256, "NUM_STAGES": 1, "num_warps": 1 }, "2048": { "BLOCK_M": 8, "BLOCK_N": 64, "NUM_STAGES": 4, "num_warps": 1 }, "8192": { "BLOCK_M": 8, "BLOCK_N": 128, "NUM_STAGES": 1, "num_warps": 1 }, "16384": { "BLOCK_M": 1, "BLOCK_N": 256, "NUM_STAGES": 1, "num_warps": 1 }, "32768": { "BLOCK_M": 8, "BLOCK_N": 256, "NUM_STAGES": 4, "num_warps": 1 }, "65536": { "BLOCK_M": 32, "BLOCK_N": 256, "NUM_STAGES": 4, "num_warps": 1 } }

235B H20 configs

0b8229e

gemini-code-assist bot reviewed Dec 3, 2025

View reviewed changes

235b fp8 h20 config

adada2b

hiworldwzj merged commit 160c1b0 into main Dec 3, 2025
1 check passed

hiworldwzj deleted the 235b_h20 branch December 3, 2025 09:29

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

235B H20 configs #1132

235B H20 configs #1132

Uh oh!

shihaobai commented Dec 3, 2025

Uh oh!

gemini-code-assist bot commented Dec 3, 2025

Uh oh!

gemini-code-assist bot left a comment

Uh oh!

gemini-code-assist bot Dec 3, 2025

Uh oh!

gemini-code-assist bot Dec 3, 2025

Uh oh!

gemini-code-assist bot Dec 3, 2025

Uh oh!

gemini-code-assist bot Dec 3, 2025

Uh oh!

gemini-code-assist bot Dec 3, 2025

Uh oh!

gemini-code-assist bot Dec 3, 2025

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

235B H20 configs #1132

235B H20 configs #1132

Uh oh!

Conversation

shihaobai commented Dec 3, 2025

Uh oh!

gemini-code-assist bot commented Dec 3, 2025

Summary of Changes

Highlights

Footnotes

Uh oh!

gemini-code-assist bot left a comment

Choose a reason for hiding this comment

Code Review

Uh oh!

gemini-code-assist bot Dec 3, 2025

Choose a reason for hiding this comment

Uh oh!

gemini-code-assist bot Dec 3, 2025

Choose a reason for hiding this comment

Uh oh!

gemini-code-assist bot Dec 3, 2025

Choose a reason for hiding this comment

Uh oh!

gemini-code-assist bot Dec 3, 2025

Choose a reason for hiding this comment

Uh oh!

gemini-code-assist bot Dec 3, 2025

Choose a reason for hiding this comment

Uh oh!

gemini-code-assist bot Dec 3, 2025

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants