runntime 0.0.1-beta.1 → 0.2.0-rc.1
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/README.md +12 -4
- package/dist/{classifier-ByH6OI6h.mjs → classifier-C6ZR3cyJ.mjs} +38576 -22654
- package/dist/{config-UyxIidZc.d.mts → config-DZyZ2TdH.d.mts} +21 -8
- package/dist/zoo/index.d.mts +206 -63
- package/dist/zoo/index.mjs +2217 -473
- package/dist/zoo/transformers/index.d.mts +10 -10
- package/dist/zoo/transformers/index.mjs +14 -16
- package/package.json +13 -5
package/README.md
CHANGED
|
@@ -1,10 +1,17 @@
|
|
|
1
1
|
<div align="center">
|
|
2
|
-
<
|
|
2
|
+
<picture>
|
|
3
|
+
<source media="(prefers-color-scheme: dark)" srcset="https://raw.githubusercontent.com/software-mansion/runntime/main/apps/docs/src/assets/logo-dark.svg">
|
|
4
|
+
<img alt="ruNNtime" src="https://raw.githubusercontent.com/software-mansion/runntime/main/apps/docs/src/assets/logo-light.svg" width="360">
|
|
5
|
+
</picture>
|
|
3
6
|
<p><strong>AI models on the web, built with TypeGPU.</strong></p>
|
|
4
7
|
</div>
|
|
5
8
|
|
|
6
9
|
<br />
|
|
7
10
|
|
|
11
|
+
---
|
|
12
|
+
|
|
13
|
+
<br />
|
|
14
|
+
|
|
8
15
|
**ruNNtime** is a framework for running AI models on the user's own GPU,
|
|
9
16
|
implemented using TypeGPU. It composes GPU kernels into the layers a model is
|
|
10
17
|
made of, attention, convolutions, norms, matmuls and builds common
|
|
@@ -25,8 +32,9 @@ It ships as one package, `runntime`, with two entry points:
|
|
|
25
32
|
- [1. Installation](#1-installation)
|
|
26
33
|
- [2. Set up the engine](#2-set-up-the-engine)
|
|
27
34
|
- [3. Run a model](#3-run-a-model)
|
|
35
|
+
- [Models](#models)
|
|
36
|
+
- [Coming from transformers.js](#coming-from-transformersjs)
|
|
28
37
|
- [Documentation](#documentation)
|
|
29
|
-
- [Migrating from transformers.js](#coming-from-transformers.js)
|
|
30
38
|
- [Created by Software Mansion](#created-by-software-mansion)
|
|
31
39
|
|
|
32
40
|
## Key Features
|
|
@@ -73,7 +81,7 @@ no bundler plugin is required. Follow the
|
|
|
73
81
|
>
|
|
74
82
|
> Object detection, instance segmentation, pose, depth estimation and image
|
|
75
83
|
> classification run in half precision and need the `shader-f16` feature.
|
|
76
|
-
>
|
|
84
|
+
> Embedding, speech to text and the privacy filter can run on
|
|
77
85
|
> devices without it.
|
|
78
86
|
|
|
79
87
|
### 2. Set up the engine
|
|
@@ -125,7 +133,7 @@ from `models`, or spread one and change a field.
|
|
|
125
133
|
|
|
126
134
|
| Task | Factory | Models |
|
|
127
135
|
| --------------------- | ------------------------- | ------------------------------------- |
|
|
128
|
-
|
|
|
136
|
+
| Embedding | `createEmbedder` | EmbeddingGemma 2, all-MiniLM-L6-v2 |
|
|
129
137
|
| Privacy filter | `createPrivacyFilter` | OpenAI privacy-filter (int8) |
|
|
130
138
|
| Speech to text | `createSpeechToText` | Moonshine tiny / base, streaming tiny |
|
|
131
139
|
| Object detection | `createObjectDetector` | YOLO26 n / m |
|