Triton GPU Latency: PyTorch Model Runtimes with Triton Kernel Implementations

A large dataset of PyTorch problems paired with candidate Triton-kernel implementations and their measured GPU runtimes. Each row contains a self-contained Python program defining a reference PyTorch model and a Triton-reimplemented version. The dataset was generated by MakoraGenerate and is hosted by makora-ai, with a last update timestamp of 2026-06-12.

Use Cases

• Benchmarking Triton kernel performance against standard PyTorch operations based on the described runtime measurements.
• Training or evaluating models for automatic GPU kernel generation based on the described program pairs.
• Analyzing patterns in GPU runtime for different computational problems based on the described reference and re-implemented models.

Strengths

• Each row is a self-contained Python program, which suggests a reproducible and isolated test case.
• The dataset is described as large, implying a substantial number of examples for analysis.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Freshness should be verified as the last update timestamp is in the future (2026-06-12).

Provenance

Source

makora-ai

Collection Method

Generated by MakoraGenerate, pairing problems (mostly from KernelBench) with Triton-kernel implementations.

Freshness

Last updated 2026-06-12 15:14:03