From 81720ae0102bbfd32dc4b101cc98623e367b6bd2 Mon Sep 17 00:00:00 2001 From: Will Feng Date: Sat, 14 Mar 2026 13:53:25 -0700 Subject: [PATCH] docs: move fp8_quant to problem 1 in hackathon problems table Reorder so the unscored warm-up problem (fp8_quant) is listed first, matching the scoring note that says "Problem 1 is not scored". --- docs/helion-hackathon.md | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/docs/helion-hackathon.md b/docs/helion-hackathon.md index ed0b4a4..1aaaedc 100644 --- a/docs/helion-hackathon.md +++ b/docs/helion-hackathon.md @@ -12,8 +12,8 @@ Submit [Helion](https://github.com/pytorch/helion) kernels to the GPU MODE leade | # | Leaderboard Name | Description | |---|-----------------|-------------| -| 1 | `causal_conv1d` | Causal depthwise 1D convolution (Mamba/Mamba-2) | -| 2 | `fp8_quant` | Per-token-group FP8 E4M3 quantization (DeepSeek-V3, Llama 3, Qwen3) | +| 1 | `fp8_quant` | Per-token-group FP8 E4M3 quantization (DeepSeek-V3, Llama 3, Qwen3) | +| 2 | `causal_conv1d` | Causal depthwise 1D convolution (Mamba/Mamba-2) | | 3 | `gated_deltanet_chunk_fwd_h` | Inter-chunk state recurrence for Gated DeltaNet | | 4 | `gated_deltanet_chunk_fwd_o` | Output computation for Gated DeltaNet | | 5 | `gated_deltanet_recompute_w_u` | WY-transform forward kernel for Gated DeltaNet |