Portfolio models
Every machine-learning model behind the in-browser demos of Jagrat Patel's portfolio, in one place: one folder per
model, each prepared for ONNX Runtime Web on WebGPU (with a WASM fallback), so the site that hosts the portfolio
never serves model weights. Pages load files straight from https://huggingface.co/thejagstudio/portfolio-models/resolve/main/<folder>/….
| folder | model | demo | license | runtime | size |
|---|---|---|---|---|---|
depth-anything-v2-small/ |
Depth Anything V2 Small | Autonomous Driving | Apache-2.0 | transformers.js | 77 MB |
distilbert-go-emotions/ |
DistilBERT GoEmotions (28 emotions) | Sentiment Analysis | MIT | transformers.js | 142 MB |
distilbert-ner/ |
DistilBERT NER | Sentiment Analysis | Apache-2.0 | transformers.js | 136 MB |
distilbert-sst2/ |
DistilBERT SST-2 (sentiment) | Sentiment Analysis | Apache-2.0 | transformers.js | 142 MB |
emotion-ferplus/ |
FER+ emotion CNN | Facial Emotion | MIT | ONNX Runtime Web | 18 MB |
magenta-style/ |
Arbitrary image stylization (Magenta) | Neural Style Transfer | Apache-2.0 | ONNX Runtime Web | 20 MB |
mediapipe-face/ |
MediaPipe Face Landmarker (ONNX) | Facial Emotion | Apache-2.0 | ONNX Runtime Web | 4 MB |
mediapipe-pose/ |
MediaPipe BlazePose (ONNX) | Pose Estimation | Apache-2.0 | ONNX Runtime Web | 15 MB |
nanodiffuse-q8/ |
NanoDiffuse 122M, 8-bit packed | NanoDiffuse (diffusion LLM) | Apache-2.0 | ONNX Runtime Web | 150 MB |
pp-ocrv5/ |
PP-OCRv5 mobile (English) | Document AI | Apache-2.0 | ONNX Runtime Web | 13 MB |
rtdetr-v2-r18/ |
RT-DETRv2 (ResNet-18) | Image Segmentation, Autonomous Driving | Apache-2.0 | transformers.js | 62 MB |
sd-turbo-webgpu-q4/ |
SD-Turbo, 4-bit packed (+ CLIP tokenizer) | Image Generation | Stability AI Community License | ONNX Runtime Web | 931 MB |
slimsam-77/ |
SlimSAM (77% pruned) | Image Segmentation | Apache-2.0 | transformers.js | 35 MB |
whisper-base-timestamped/ |
Whisper base (timestamped) | Speech Recognition | MIT | transformers.js | 191 MB |
whisper-tiny-timestamped/ |
Whisper tiny (timestamped) | Speech Recognition | MIT | transformers.js | 108 MB |
Each folder has its own README (source, what was changed, inputs and outputs) and LICENSE. Third-party models are redistributed under their original licenses with credit to their authors; most are copied unchanged, and conversions or re-quantisations are described in the folder's README together with the script that made them.
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support