Spaces:
Running
Running
Improve onboarding on the TensorFold Hub profile
Browse files- README.md +15 -30
- index.html +33 -58
README.md
CHANGED
|
@@ -4,45 +4,30 @@ sdk: static
|
|
| 4 |
pinned: false
|
| 5 |
---
|
| 6 |
|
| 7 |
-
<
|
| 8 |
-
<img src="./tensorfold-hero.png" alt="TensorFold" width="100%">
|
| 9 |
|
| 10 |
-
# TensorFold
|
| 11 |
|
| 12 |
-
|
| 13 |
|
| 14 |
-
[
|
| 15 |
-
</div>
|
| 16 |
|
| 17 |
-
|
| 18 |
|
| 19 |
-
|
| 20 |
|
| 21 |
-
-
|
| 22 |
-
-
|
| 23 |
-
-
|
| 24 |
-
- Measured speed, memory use, runtime versions, and known limits
|
| 25 |
-
- Clear credit, licences, and links to the original model authors
|
| 26 |
|
| 27 |
-
|
| 28 |
|
| 29 |
-
|
| 30 |
-
curl -fsSL https://tensorfold.dev/install.sh | sh
|
| 31 |
-
```
|
| 32 |
|
| 33 |
-
|
| 34 |
|
| 35 |
-
##
|
| 36 |
|
| 37 |
-
|
| 38 |
-
| --- | --- | --- |
|
| 39 |
-
| 64 GB | [Browse models](https://huggingface.co/collections/TensorFold/mlx-models-for-64gb-macs) | Published peak below 48 GB |
|
| 40 |
-
| 128 GB | [Browse models](https://huggingface.co/collections/TensorFold/mlx-models-for-128gb-macs) | Published peak below 96 GB |
|
| 41 |
-
| 256 GB | [Browse models](https://huggingface.co/collections/TensorFold/mlx-models-for-256gb-macs) | Published peak below 192 GB |
|
| 42 |
-
|
| 43 |
-
These collections are starting points with nominal context headroom, not guarantees at maximum context. Each model card records the tested runtime, prompt, output length, memory evidence, and any compatibility caveats.
|
| 44 |
-
|
| 45 |
-
TensorFold does not train the base models. Model design, training, evaluations, and upstream documentation remain the work of the original authors and contributors.
|
| 46 |
-
|
| 47 |
-
[Follow TensorFold for new model releases, runtime updates, and fixes.](https://huggingface.co/TensorFold)
|
| 48 |
|
|
|
|
|
|
| 4 |
pinned: false
|
| 5 |
---
|
| 6 |
|
| 7 |
+
<img src="https://huggingface.co/spaces/TensorFold/README/resolve/main/tensorfold-logo.png" alt="TensorFold logo" width="96">
|
|
|
|
| 8 |
|
| 9 |
+
# Welcome to TensorFold
|
| 10 |
|
| 11 |
+
TensorFold is an open-source inference engine that runs models on your own hardware and serves them through an OpenAI-compatible API. On Hugging Face, we publish model conversions for Apple silicon, along with previews for new model architectures.
|
| 12 |
|
| 13 |
+
[Setup guide](https://tensorfold.dev/guide) 路 [Browse all models](https://huggingface.co/TensorFold/models) 路 [Engine on GitHub](https://github.com/ashhart/TensorFold)
|
|
|
|
| 14 |
|
| 15 |
+
## Find a model for your Mac
|
| 16 |
|
| 17 |
+
Start with the collection for your Mac鈥檚 unified memory.
|
| 18 |
|
| 19 |
+
- [64 GB Macs](https://huggingface.co/collections/TensorFold/mlx-models-for-64gb-macs-6a9fefda17932216ec9ab457)
|
| 20 |
+
- [128 GB Macs](https://huggingface.co/collections/TensorFold/mlx-models-for-128gb-macs-6a9ff0abd31bc9abbe7922d7)
|
| 21 |
+
- [256 GB Macs](https://huggingface.co/collections/TensorFold/mlx-models-for-256gb-macs-6a9ff0ef9fed7c5bdca15e9b)
|
|
|
|
|
|
|
| 22 |
|
| 23 |
+
New to MLX? Try [Nemotron 3.5 Lightning 4-bit](https://huggingface.co/TensorFold/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-MLX-4bit). Its card includes a quick start and the conditions for a measured Apple silicon run. Check the runtime and memory notes before downloading. Longer contexts and concurrent requests can use more memory.
|
| 24 |
|
| 25 |
+
## What鈥檚 new
|
|
|
|
|
|
|
| 26 |
|
| 27 |
+
Tencent Hy4-flash with DFly2 is available as a checkpoint preview in [MLX 4-bit](https://huggingface.co/TensorFold/Hy4-flash-MLX-4bit-MTP) and [MLX 8-bit](https://huggingface.co/TensorFold/Hy4-flash-MLX-8bit-MTP). Conversion checks have passed, and TensorFold runtime support is pending.
|
| 28 |
|
| 29 |
+
## Need a hand?
|
| 30 |
|
| 31 |
+
Use the **Community** tab on a model鈥檚 page for questions about that checkpoint. For engine issues, use [GitHub Issues](https://github.com/ashhart/TensorFold/issues). Include your hardware and memory, runtime version, and model ID so we can help.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 32 |
|
| 33 |
+
The original model authors receive credit in the checkpoint cards, alongside links to the source models and their licences.
|
index.html
CHANGED
|
@@ -1,61 +1,36 @@
|
|
| 1 |
<!doctype html>
|
| 2 |
<html lang="en">
|
| 3 |
-
|
| 4 |
-
|
| 5 |
-
|
| 6 |
-
|
| 7 |
-
|
| 8 |
-
|
| 9 |
-
|
| 10 |
-
|
| 11 |
-
|
| 12 |
-
|
| 13 |
-
|
| 14 |
-
|
| 15 |
-
|
| 16 |
-
|
| 17 |
-
|
| 18 |
-
|
| 19 |
-
|
| 20 |
-
|
| 21 |
-
|
| 22 |
-
|
| 23 |
-
|
| 24 |
-
|
| 25 |
-
|
| 26 |
-
|
| 27 |
-
|
| 28 |
-
|
| 29 |
-
|
| 30 |
-
|
| 31 |
-
|
| 32 |
-
|
| 33 |
-
|
| 34 |
-
|
| 35 |
-
|
| 36 |
-
</p>
|
| 37 |
-
|
| 38 |
-
<br />
|
| 39 |
-
|
| 40 |
-
<h2>What you will find here</h2>
|
| 41 |
-
<ul>
|
| 42 |
-
<li><strong>Apple Silicon:</strong> MLX quantized checkpoints with measured speed and memory use.</li>
|
| 43 |
-
<li><strong>Drafted decoding:</strong> MTP and DFlash assets when the upstream model provides a compatible drafter.</li>
|
| 44 |
-
<li><strong>NVIDIA CUDA:</strong> tested recipes for supported GPUs and DGX Spark systems.</li>
|
| 45 |
-
<li><strong>Clear release notes:</strong> runtime versions, setup steps, known limits, licences, and upstream credit.</li>
|
| 46 |
-
</ul>
|
| 47 |
-
|
| 48 |
-
<br />
|
| 49 |
-
|
| 50 |
-
<h2>Run TensorFold</h2>
|
| 51 |
-
<p><code>curl -fsSL https://tensorfold.dev/install.sh | sh</code></p>
|
| 52 |
-
|
| 53 |
-
<br />
|
| 54 |
-
|
| 55 |
-
<p>
|
| 56 |
-
TensorFold does not train the base models. Model design, training, evaluations, and
|
| 57 |
-
upstream documentation remain the work of the original authors and contributors.
|
| 58 |
-
</p>
|
| 59 |
-
</main>
|
| 60 |
-
</body>
|
| 61 |
</html>
|
|
|
|
| 1 |
<!doctype html>
|
| 2 |
<html lang="en">
|
| 3 |
+
<head>
|
| 4 |
+
<meta charset="utf-8">
|
| 5 |
+
<meta name="viewport" content="width=device-width, initial-scale=1">
|
| 6 |
+
<meta name="description" content="Find a TensorFold checkpoint, check runtime support, and get started with local inference.">
|
| 7 |
+
<title>TensorFold on Hugging Face</title>
|
| 8 |
+
<style>
|
| 9 |
+
:root { color-scheme: light dark; }
|
| 10 |
+
.tensorfold-org-card { font: 1rem/1.6 system-ui, sans-serif; max-width: 44rem; margin: 2rem auto; padding: 0 1.25rem 2rem; }
|
| 11 |
+
.tensorfold-org-card h1, .tensorfold-org-card h2 { line-height: 1.25; }
|
| 12 |
+
.tensorfold-org-card h1 { margin-top: .5rem; }
|
| 13 |
+
</style>
|
| 14 |
+
</head>
|
| 15 |
+
<body>
|
| 16 |
+
<main class="tensorfold-org-card">
|
| 17 |
+
<img src="https://huggingface.co/spaces/TensorFold/README/resolve/main/tensorfold-logo.png" alt="TensorFold logo" width="96">
|
| 18 |
+
<h1>Welcome to TensorFold</h1>
|
| 19 |
+
<p>TensorFold is an open-source inference engine that runs models on your own hardware and serves them through an OpenAI-compatible API. On Hugging Face, we publish model conversions for Apple silicon, along with previews for new model architectures.</p>
|
| 20 |
+
<p><a href="https://tensorfold.dev/guide">Setup guide</a> 路 <a href="https://huggingface.co/TensorFold/models">Browse all models</a> 路 <a href="https://github.com/ashhart/TensorFold">Engine on GitHub</a></p>
|
| 21 |
+
<h2>Find a model for your Mac</h2>
|
| 22 |
+
<p>Start with the collection for your Mac鈥檚 unified memory.</p>
|
| 23 |
+
<ul>
|
| 24 |
+
<li><a href="https://huggingface.co/collections/TensorFold/mlx-models-for-64gb-macs-6a9fefda17932216ec9ab457">64 GB Macs</a></li>
|
| 25 |
+
<li><a href="https://huggingface.co/collections/TensorFold/mlx-models-for-128gb-macs-6a9ff0abd31bc9abbe7922d7">128 GB Macs</a></li>
|
| 26 |
+
<li><a href="https://huggingface.co/collections/TensorFold/mlx-models-for-256gb-macs-6a9ff0ef9fed7c5bdca15e9b">256 GB Macs</a></li>
|
| 27 |
+
</ul>
|
| 28 |
+
<p>New to MLX? Try <a href="https://huggingface.co/TensorFold/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-MLX-4bit">Nemotron 3.5 Lightning 4-bit</a>. Its card includes a quick start and the conditions for a measured Apple silicon run. Check the runtime and memory notes before downloading. Longer contexts and concurrent requests can use more memory.</p>
|
| 29 |
+
<h2>What鈥檚 new</h2>
|
| 30 |
+
<p>Tencent Hy4-flash with DFly2 is available as a checkpoint preview in <a href="https://huggingface.co/TensorFold/Hy4-flash-MLX-4bit-MTP">MLX 4-bit</a> and <a href="https://huggingface.co/TensorFold/Hy4-flash-MLX-8bit-MTP">MLX 8-bit</a>. Conversion checks have passed, and TensorFold runtime support is pending.</p>
|
| 31 |
+
<h2>Need a hand?</h2>
|
| 32 |
+
<p>Use the <strong>Community</strong> tab on a model鈥檚 page for questions about that checkpoint. For engine issues, use <a href="https://github.com/ashhart/TensorFold/issues">GitHub Issues</a>. Include your hardware and memory, runtime version, and model ID so we can help.</p>
|
| 33 |
+
<p>The original model authors receive credit in the checkpoint cards, alongside links to the source models and their licences.</p>
|
| 34 |
+
</main>
|
| 35 |
+
</body>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 36 |
</html>
|