Downloads · 30 days
0
raws-labs/tigris-zoo
tigris-zoo is a machine learning model from raws-labs. Use it for the machine learning task on the model card, and read the license before you ship it in a product. The card lists the license as other.
Precompiled .tgrs models with runtime requirements, input/output conventions, evaluation results, and example tensors. Each model's artifacts are under models/<model/artifacts/<artifact-id/.
Downloads · 30 days
0
Access
Public
Updated Oct 19, 2026
Repo size
753 KB
Likes
1
Public
Click a slice to open those files.
.tgrs767 KB · 60%
From the Hugging Face model README
Precompiled .tgrs models with runtime requirements, input/output conventions,
evaluation results, and example tensors. Each model's artifacts are under
models/<model>/artifacts/<artifact-id>/.
Each artifact's readme.md documents the model's task, input preprocessing,
output units, evaluation results and limitations. example-output.bin is the
reference output the recipe compared the runtime against for example-input.bin;
evaluation.json records how closely the runtime reproduced the reference.
python -m venv .venv
source .venv/bin/activate
pip install tigris-ml
tigris zoo list
tigris zoo fetch electricity-hourly -o downloaded-model
tigris codegen downloaded-model/model.tgrs --format core -o model.c
Downloads need no account. Filters select compatible builds before choosing the
newest publication. Supply the runtime separately within the current range recorded
in download.json. A null maximum means no known upper cutoff; tested releases
are listed separately. The original build manifest remains unchanged when catalog
compatibility records are updated. Fast/slow arena requirements exclude external buffers, stack,
and runtime metadata. Model quality and evaluation scope are documented per build.
Model licenses differ; inspect the downloaded license.txt and source provenance.