Pre-built Python wheels for ninja, flash-linear-attention, and causal-conv1d compiled for modern NVIDIA GPUs, authored by Feargal and last updated on April 11, 2026. This collection is organized to help developers bypass resource-intensive C++ compilation times in CI/CD pipelines and Docker builds. The dataset is open-sourced and hosted on the Hugging Face platform.
Use Cases
- Accelerate CI/CD pipeline builds based on pre-compiled CUDA wheels mentioned in the description
- Speed up Docker image creation for GPU-based development environments based on the described compilation bypass
- Deploy GPU-accelerated causal convolutional and linear attention layers without local compilation based on the provided pre-built packages
Strengths
- Specifically compiled for modern NVIDIA GPUs, as stated in the description
- Aims to bypass highly resource-intensive C++ compilation times, a concrete benefit mentioned
Limitations
- Description metadata is limited; actual data quality requires manual inspection after download
- Column-level documentation is absent; field semantics must be inferred after download
- File formats and dataset size are unknown, which may limit suitability assessment
Provenance
- Source
- Hugging Face (Feargal)
- Collection Method
- Collection of pre-built software packages
- Time Range
- Last updated April 2026
- Freshness
- Last updated 2026-04-11 22:08:54
- Geography
- null