Downloads · 30 days
1
5% of all-time downloads
Kanha-AI/b0f84795-9aab-4c3a-9b83-83d1e6fe23e6-medium
b0f84795-9aab-4c3a-9b83-83d1e6fe23e6-medium is a text generation model from Kanha-AI. Use it when you need the model to write or continue text. It is set up for mlc-llm. The card lists the license as other.
This model was trained by Kanha and runs entirely on-device via WebGPU.
Downloads · 30 days
1
5% of all-time downloads
All-time downloads
19
Public
Repo size
659 MB
Likes
0
Public
Click a slice to open those files.
.bin659 MB · 99%
From the Hugging Face model README
This model was trained by Kanha and runs entirely on-device via WebGPU.
This is a custom-trained chatbot checkpoint, fine-tuned on a specific website's content using Kanha. It is optimized for on-device inference in the browser via WebGPU — no server-side API calls, no per-query costs, no data leaves the device.
This model is not general-purpose. It is trained to answer questions about a specific website's content.
Install the SDK and embed the bot:
<script type="module" src="https://cdn.jsdelivr.net/npm/kanha-ai/dist/widget.js"></script>
<kanha-bot model-url="https://huggingface.co/Kanha-AI/b0f84795-9aab-4c3a-9b83-83d1e6fe23e6-medium/resolve/main/" model-size="medium" model-lib="https://huggingface.co/Kanha-AI/kanha-liquid-lfm2.5-1.2b-instruct-sft-v1-MLC/resolve/b5b3e59834bf187e37cfe51acc7edc3c05a08bb8/kanha-liquid-lfm2.5-1.2b-tuned-q4f16_1.wasm"></kanha-bot>
Or with React:
import { KanhaBot } from 'kanha-ai';
<KanhaBot modelUrl="https://huggingface.co/Kanha-AI/b0f84795-9aab-4c3a-9b83-83d1e6fe23e6-medium/resolve/main/" modelSize="medium" modelLib="https://huggingface.co/Kanha-AI/kanha-liquid-lfm2.5-1.2b-instruct-sft-v1-MLC/resolve/b5b3e59834bf187e37cfe51acc7edc3c05a08bb8/kanha-liquid-lfm2.5-1.2b-tuned-q4f16_1.wasm" />
MLC model library: https://huggingface.co/Kanha-AI/kanha-liquid-lfm2.5-1.2b-instruct-sft-v1-MLC/resolve/b5b3e59834bf187e37cfe51acc7edc3c05a08bb8/kanha-liquid-lfm2.5-1.2b-tuned-q4f16_1.wasm
Kanha crawls your website, generates training data, fine-tunes a compact model, and serves it directly in your visitors' browsers via WebGPU.