How v4nnuth/deploy_mstsne works
Ms t-SNE Explorer is a browser-based benchmark runner for comparing L-BFGS-B, SGD, Momentum, and Adam across datasets. A researcher configures an experiment, the Vue client creates a session and streams a run request to FastAPI, which loads and embeds the data, evaluates optimizers, and produces results and figures for the UI. The README documents the broader workflow, code/formula inspection, Adam learning-rate search, and GPU/CPU support. Several frontend and backend connections are explicit in excerpts or the README; unsampled details are not expanded into speculative calls.
Main entry points: main.py FastAPI run service · App.vue Application shell · useExperiment.js Session client · mstsne.py Optimizer execution
Components 17
Experiment interface
- Application shell
frontend/src/App.vue - Run configuration
frontend/src/pages/ConfigurePage.vue - Workflow canvas
frontend/src/components/WorkflowCanvas.vue - Stage inspector
frontend/src/components/NodePanel.vue - Code and formulas
frontend/src/components/CodeEditorSidebar.vue - Session client
frontend/src/composables/useExperiment.js
Run orchestration
- FastAPI run service
backend/api/main.py - Dataset loaders
backend/dataset_loaders.py - Benchmark datasets
dataset
Embedding analysis
- Distance and similarity
backend/engine/mstsne.py - Optimizer execution
backend/engine/mstsne.py - R_NX and AUC
backend/engine/mstsne.py
Review outputs
- Figure generation
backend/api/main.py - Results view
frontend/src/pages/ResultsPage.vue - Figures view
frontend/src/pages/FiguresPage.vue
Other components
- Researcher
- CuPy or NumPy
Connections 18
- Researcher → Run configuration: sets run options
README.md - Run configuration → Application shell: submits configuration
README.md - Application shell → Session client: starts session and run
frontend/src/App.vue - Application shell → Workflow canvas: shows run progress
frontend/src/App.vue - Application shell → Stage inspector: opens stage details
frontend/src/App.vue - Application shell → Code and formulas: opens code editor
frontend/src/App.vue - Session client → FastAPI run service: sends WebSocket run
frontend/src/composables/useExperiment.js - FastAPI run service → Dataset loaders: loads dataset
backend/api/main.py - Dataset loaders → Benchmark datasets: reads benchmark data
backend/dataset_loaders.py - FastAPI run service → Distance and similarity: computes distances
backend/api/main.py - FastAPI run service → Optimizer execution: runs optimizers
backend/api/main.py - Optimizer execution → R_NX and AUC: evaluates embeddings
README.md - FastAPI run service → Figure generation: generates plots
backend/api/main.py - FastAPI run service → Session client: streams run updates
README.md - Session client → Results view: provides scores
frontend/src/App.vue - Session client → Figures view: provides figures
frontend/src/App.vue - Distance and similarity → CuPy or NumPy: uses GPU or CPU
README.md - Optimizer execution → CuPy or NumPy: uses GPU or CPU
README.md