Downloads · 30 days
0
wakirin/supertonic-2-trt-engines-fp16-sm86-cuda12
supertonic-2-trt-engines-fp16-sm86-cuda12 is a machine learning model from wakirin. Use it for the machine learning task on the model card, and read the license before you ship it in a product.
Prebuilt ONNX Runtime TensorRT EP engines for slnww/gpu-worker-supertonic:cuda12, downloaded by services/gpu-worker/scripts/download-models.sh into /opt/trtcache.
Downloads · 30 days
0
Access
Public
Updated Oct 22, 2026
Repo size
295 MB
Likes
0
Public
Click a slice to open those files.
.engine145 MB · 100%
From the Hugging Face model README
Prebuilt ONNX Runtime TensorRT EP engines for slnww/gpu-worker-supertonic:cuda12,
downloaded by services/gpu-worker/scripts/download-models.sh into /opt/trt_cache.
ORT names each engine by a hash of the model graph and the ORT / CUDA / TensorRT versions, so these files only hit inside the exact environment below. After any change to those versions or to the ONNX models, rebuild inside the new image and re-upload, or every boot silently rebuilds (~5 min per worker).
Built 2026-10-22 in:
slnww/gpu-worker-supertonic:cuda12 (digest sha256:90b519759c51…)Used as the default for every sm86 card without its own repo: engines built on
another GPU model load fine but TensorRT warns, and speed varies — 3060-built
engines measured within 5% of native on an A5000; 3090-built ones were 24-29%
slower on a 3060/A5000, hence a separate ...-rtx3090-cuda12 repo.
| model | engine |
|---|---|
| duration_predictor | ..._10894559101746328646_... |
| text_encoder | ..._8296872225004375479_... |
| vector_estimator | ..._9244645762570505081_... |
| vocoder | ..._742995754182254538_... |