Repository navigation

All pull requests

Labels (9)Milestones (0)New pull request

Open

(23)

Closed

(98)

Rotate timed trials over distinct input sets to stop input-identity memoization (#171)
#173utsav1033 opened Oct 14, 2026
Skip untimed correct samples in geometric mean speedup
#170shaurya416 opened Sep 25, 2026
Fix kernel_static_checker.py false positives on try/except and pass statements
#161cubeerea opened Aug 19, 20263 tasks done
fix: eval_kernel_against_ref catches only RuntimeError on model load, not other exceptions
#160MattFisher opened Jul 20, 2026
Fix: Resolve ignored groups parameter in Model initialization
#158diyarora opened Jun 28, 2026
SOL Model
#153fmgreco opened May 15, 2026
fix(eval): preserve integer tensor dtype during precision cast
#152moria97 opened Apr 16, 20264 tasks done
Fix dimensions of 97_ScaledDotProductAttention
#149arielge opened Apr 6, 2026
Moved cuda synch outside of warmup loop
#145Tyler-Hilbert opened Mar 22, 2026
Fix pytorch_wrap regex to catch both nn.X and torch.nn.X
#140nataliakokoromyti opened Feb 24, 2026Collaborator
add kernel evaluation notebook
#136shira-g opened Jan 25, 2026
fix: Fixes for bugs encountered during generation & evaluation
#134peri044 opened Jan 22, 2026
2
Add support to non-fp32 input dtype in level3/33_VanillaRNN.py and level3/35_LSTM.py
#131mzweilin opened Jan 20, 2026
Triton TLX Backend Support
#128Willy-Chan opened Jan 20, 2026Collaborator
cuTile Backend Support
#126Willy-Chan opened Jan 17, 2026Collaborator
[wip] fix some hackable level 1 problems
#118bkal01 opened Jan 7, 2026Collaborator
4
☑️ Fix problems solvable by a constant fill or eliminating a redundant operation
#108EssamWisam opened Dec 20, 2025
18
Benchmarking guide
#106PaliC opened Dec 17, 2025Collaborator
14
Fix Issue #97: Use Heavy-Tailed Distribution for MSELoss to Prevent Partial Computation from Passing Verification
#98doux-jy opened Nov 28, 2025
Tt prompt generation
#90chilinghan opened Nov 11, 2025
fix metric calculation caused by mismatched pair comparisons
#60kleinercubs opened Aug 23, 2025
1
adjusted level 3 sizes
#58AffectionateCurry opened Jul 24, 2025Collaborator
2
Investigate Sakana Kernel
#25simonguozirui opened Feb 23, 2025Collaborator

Navigation Menu

Repository navigation

All pull requests

Rotate timed trials over distinct input sets to stop input-identity memoization (#171)

Skip untimed correct samples in geometric mean speedup

Fix kernel_static_checker.py false positives on try/except and pass statements

fix: eval_kernel_against_ref catches only RuntimeError on model load, not other exceptions

Fix: Resolve ignored groups parameter in Model initialization

SOL Model

fix(eval): preserve integer tensor dtype during precision cast

Fix dimensions of 97_ScaledDotProductAttention

Moved cuda synch outside of warmup loop

Fix pytorch_wrap regex to catch both nn.X and torch.nn.X

add kernel evaluation notebook

fix: Fixes for bugs encountered during generation & evaluation

Add support to non-fp32 input dtype in `level3/33_VanillaRNN.py` and `level3/35_LSTM.py`

Triton TLX Backend Support

cuTile Backend Support

[wip] fix some hackable level 1 problems

☑️ Fix problems solvable by a constant fill or eliminating a redundant operation

Benchmarking guide

Fix Issue #97: Use Heavy-Tailed Distribution for MSELoss to Prevent Partial Computation from Passing Verification

Tt prompt generation

fix metric calculation caused by mismatched pair comparisons

adjusted level 3 sizes

Investigate Sakana Kernel

Pull requests