This is an ONNX export of XPENG-AI/X-AuT, a Qwen3-ASR-0.6B model whose audio encoder was pruned from 18 to 14 layers. The files are laid out for onnxruntime-web on WebGPU. Demo: mrfakename/x-aut-webgpu. License: CC-BY-NC-4.0 (non-commercial), inherited from the original model. All credit for the model goes to XPENG-AI. This is an unofficial conversion. The decoder was built with the onnxruntime-genai builder (excludeembeds). 1. Log-mel features (Whisper-style, 128 bins, matching preprocessorconfig.json): - 16 kHz audio - nfft 400, hop 160, periodic Hann window, reflect pad - log10, clamped to max − 8, then (x + 4) / 4 - frames = len // 160 2. Encoder: run the encoder to get N audio…
Open weights
cc-by-nc-4.0
onnxruntime
ONNX export of SurjoLabs/Surjo-Image-Preview for the browser (onnxruntime-web, WebGPU), together with its text encoder LiquidAI/LFM2.5-Encoder-230M and VAE decoder fp16w means weights are stored in fp16 and computed in fp32 (image PSNR about 60 dB against PyTorch fp32). An int4 text encoder was tried and dropped: it changed the pictures too much. Sampling: start from standard normal noise, t = i/steps, dt = 1/steps, 50 steps, cfg 3.0. onnx/parity.json has the image PSNR against the PyTorch fp32 pipeline.
Open weights
apache-2.0
onnxruntime
ONNX export of udgamlabs/Chitra-Tiny, a 212M-parameter text-to-image model (256×256), for onnxruntime-web on WebGPU. Used by the mrfakename/chitra-tiny-webgpu Space. Weights are stored in fp16 and cast to fp32 inside each graph: half the download, fp32 math. The DiT can't run in fp16 because its residual stream goes far past the fp16 range. Sampling (same as chitratiny.py): tokenize with padding to 77, encode the prompt and the negative prompt (empty by default), start from Gaussian noise, then Euler steps over t = linspace(1, 0, steps+1) with x += (t[i+1] - t[i]) (vuncond + g (vcond - vuncond)), then decode. Defaults: 50 steps, guidance 7. tools/exportonnx.py reproduces the files…
Open weights
apache-2.0
onnx
ONNX files for nvidia/parakeet-unified-en-0.6b (also mirrored at algorithco/parakeet-unified-en-0.6b), sized for ONNX Runtime Web with WebGPU. They are used by the parakeet-unified-webgpu Space. The int8 decoder is taken from the same repo. - The 48 pointwise (kernel size 1) Conv1d layers were rewritten as MatMul so that the weight-only quantiser covers them. Then the encoder was quantised with ONNX Runtime's MatMulNBitsQuantizer. - preprocessor.onnx is the same NeMo 128-mel front end used in mrfakename/parakeet-ultra-ONNX, with the same settings as this model. The encoder uses full-context attention, which is the model's offline mode. The chunked streaming attention masks are not part of…
Open weights
other
onnx
