| --- |
| license: apache-2.0 |
| base_model: Qwen/Qwen3-1.7B |
| tags: |
| - quantization |
| - trellis-coded-quantization |
| - qtip |
| - webgpu |
| - 3-bit |
| - on-device |
| library_name: transformers |
| --- |
| |
| # Qwen3-1.7B — 3-bit trellis-quantized for WebGPU |
|
|
| A **3-bit trellis-coded quantization (TCQ)** of [Qwen3-1.7B](https://huggingface.co/Qwen/Qwen3-1.7B), |
| packed to run its **full forward pass in a web browser on WebGPU** — no CUDA, no server. |
|
|
| This is part of **[Trellis WebGPU](https://github.com/Azimml/trellis-webgpu)** — |
| the first trellis-coded quantization decoder (the QTIP / EXL3 quality tier) to run |
| outside CUDA. See the repo for the quantizer, the WGSL kernels, and the full |
| verification harness. |
|
|
| ## Quality (WikiText-2 perplexity, no fine-tuning) |
|
|
| | | fp16 | **TCQ 3-bit (this model)** | vs fp16 | |
| |---|---|---|---| |
| | Qwen3-1.7B | 15.53 | **18.00** | 1.16× | |
|
|
| Trellis-coded quantization is the current quality frontier for low-bit LLM weights, |
| beating scalar quantization (GPTQ / AWQ / GGUF) at equal bitrate. The K=2 quantizer in |
| this project reproduces the QTIP paper's published rate–distortion (MSE 0.0739 vs 0.0733). |
|
|
| ## This packed model |
|
|
| - **28 layers**, packed to **~3 bits/weight** (0.63 GB on disk). |
| - peaks at **~1.0 GB VRAM** — measured live in Chrome on a laptop RTX 3050 Ti; runs comfortably on a 4 GB (or smaller) GPU. |
| - Verified: the full model runs through the exact shipping WGSL shaders and generates |
| coherent, factually-correct text; kernels match NumPy to `<1e-6`; 135M logits are |
| top-5 exact vs PyTorch. |
|
|
| ## Format |
|
|
| This is **not** a standard `transformers` checkpoint. It is a packed 3-bit format |
| (`manifest.json` + sharded `.bin` weights + IP metadata) designed for the WebGPU runtime |
| in the [Trellis WebGPU](https://github.com/Azimml/trellis-webgpu) repo. To run it: |
|
|
| ```bash |
| git clone https://github.com/Azimml/trellis-webgpu |
| # place these files under web/model_1p7b/ , then: |
| cd web && python3 -m http.server 8000 # open index.html in a WebGPU browser |
| # or verify headlessly on your GPU: |
| python scripts/run_packed_headless.py web/model_1p7b "The capital of France is" 40 |
| ``` |
|
|
| ## License & credit |
|
|
| Quantized derivative of [Qwen3-1.7B](https://huggingface.co/Qwen/Qwen3-1.7B) (Apache-2.0), |
| distributed under the base model's license. Method: QTIP (Tseng et al., NeurIPS 2024) |
| and EXL3 (turboderp). Quantization + WebGPU port: independent reimplementation. |
|
|
