Jean Rojas jeanc18rlos
jeanc18rlos
/ cog.yaml
Created
May 13, 2026 19:35
SAM2 hybrid — Replicate encoder microservice. cog package (cog.yaml + predict.py + push.sh) and browser-side client. Companion to https://jeanrojas.com/blog/splitting-sam2-encoder-decoder
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| build: | |
| gpu: true | |
| cuda: "12.1" | |
| python_version: "3.11" | |
| python_packages: | |
| - "torch==2.4.0" | |
| - "torchvision==0.19.0" | |
| - "onnxruntime-gpu==1.20.0" | |
| - "Pillow==10.4.0" | |
| - "numpy==1.26.4" |
jeanc18rlos
/ sam2-decoder.ts
Created
May 13, 2026 19:32
SAM2 hybrid — browser-side decoder (lib/sam2-decoder.ts). Loads decoder.onnx on WebGPU, fetches the encoder embedding bundle, runs per-click segmentation. Companion to https://jeanrojas.com/blog/splitting-sam2-encoder-decoder
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| // lib/sam2-decoder.ts | |
| // Browser-side decoder for SAM2 hybrid deployment. | |
| // Loads sam2.1_hiera_large.decoder.onnx (~16 MB) via onnxruntime-web on WebGPU. | |
| // Pairs with the encoder bundle produced by the companion notebook. | |
| // Full article: https://jeanrojas.com/blog/splitting-sam2-encoder-decoder | |
| import * as ort from "onnxruntime-web/webgpu"; | |
| export interface Embedding { | |
| imageEmbed: ort.Tensor; |
jeanc18rlos
/ 1_install.sh
Created
May 13, 2026 19:30
SAM2 hybrid encoder notebook — 5 cells: install, download checkpoint, export ONNX, encode image, package for browser. Companion to https://jeanrojas.com/blog/splitting-sam2-encoder-decoder
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| # Cell 1 — pin a working torch + the ONNX toolchain. | |
| pip install --quiet \ | |
| torch==2.4.0 \ | |
| torchvision==0.19.0 \ | |
| onnx onnxscript onnxsim onnxruntime samexporter | |
| pip install --quiet git+https://github.com/facebookresearch/segment-anything-2.git |