How woduq1414/ko-microgpt works
ko-microgpt combines a Korean/English name-generation model with a scroll-based React lesson that explains the model’s data, tokenization, embedding, attention, loss, training, and inference stages. The frontend is the main user-facing workflow; the Python model is a separate training and inference runtime. The graph preserves the documented lesson capabilities, but does not infer connections to snapshot files or unsampled chapter components beyond the relationships shown in the source index and excerpts.
Main entry points: App.jsx Lesson app · chapterUtils.js Token and math utilities · ChapterFiveTrainingDemo.jsx Loss and gradients · ChapterSixTrainingDemo.jsx Training updates
Components 22
Lesson application
- Lesson app
app/src/App.jsx - Language and intro
app/src/components/sections/HeroSection.jsx - Chapter sections
app/src/components/sections/ChapterSectionShell.jsx - Lesson descriptions
app/src/constants/lessonSections.js - Localization
app/src/constants/localization.js
Model lessons
- Name data
app/src/components/chapters/ChapterOneDataCloud.jsx - Jamo cloud
app/src/components/chapters/ChapterTwoJamoCloud.jsx - Tokenization
app/src/components/chapters/ChapterTwoTokenizationDemo.jsx - Embeddings
app/src/components/chapters/ChapterThreeEmbeddingDemo.jsx - Attention
app/src/components/chapters/ChapterFourAttentionDemo.jsx - Loss and gradients
app/src/components/chapters/ChapterFiveTrainingDemo.jsx - Training updates
app/src/components/chapters/ChapterSixTrainingDemo.jsx - Inference demo
app/src/components/chapters/ChapterSevenInferenceDemo.jsx - Model comparison
app/src/components/sections/ChapterEightSection.jsx
Model runtime
- Training pipeline
model/ko_main.py - Inference entrypoint
model/ko_inference.py - Name datasets
model/data - Trained checkpoints
model/checkpoints
Lesson support
- Token and math utilities
app/src/components/chapters/shared/chapterUtils.js - Lesson constants
app/src/components/chapters/shared/chapterConstants.js - Unavailable state
app/src/components/common/SectionStateCard.jsx
Other components
- Learner
Connections 27
- Learner → Lesson app: uses
README.md - Lesson app → Language and intro: renders
app/src/App.jsx - Lesson app → Chapter sections: renders
app/src/App.jsx - Lesson app → Name data: renders
app/src/App.jsx - Lesson app → Jamo cloud: renders
app/src/App.jsx - Lesson app → Tokenization: renders
app/src/App.jsx - Lesson app → Embeddings: renders
app/src/App.jsx - Lesson app → Attention: renders
app/src/App.jsx - Lesson app → Loss and gradients: renders
app/src/App.jsx - Lesson app → Training updates: renders
app/src/App.jsx - Lesson app → Inference demo: renders
app/src/App.jsx - Lesson app → Model comparison: renders
app/src/App.jsx - Lesson app → Localization: reads language settings
app/src/App.jsx - Tokenization → Token and math utilities: uses token helpers
app/src/components/chapters/ChapterTwoTokenizationDemo.jsx - Embeddings → Token and math utilities: uses math and token helpers
app/src/components/chapters/ChapterThreeEmbeddingDemo.jsx - Attention → Token and math utilities: uses attention helpers
app/src/components/chapters/ChapterFourAttentionDemo.jsx - Loss and gradients → Token and math utilities: uses shared helpers
app/src/components/chapters/ChapterFiveTrainingDemo.jsx - Training updates → Token and math utilities: uses shared helpers
app/src/components/chapters/ChapterSixTrainingDemo.jsx - Inference demo → Token and math utilities: uses inference helpers
app/src/components/chapters/ChapterSevenInferenceDemo.jsx - Token and math utilities → Lesson constants: imports constants
app/src/components/chapters/shared/chapterUtils.js - Loss and gradients → Unavailable state: uses fallback
app/src/components/chapters/ChapterFiveTrainingDemo.jsx - Training updates → Unavailable state: uses fallback
app/src/components/chapters/ChapterSixTrainingDemo.jsx - Inference demo → Unavailable state: uses fallback
app/src/components/chapters/ChapterSevenInferenceDemo.jsx - Training pipeline → Name datasets: loads training names
model/ko_main.py - Training pipeline → Trained checkpoints: saves model
README.md - Inference entrypoint → Training pipeline: calls inference
model/ko_inference.py - Inference entrypoint → Trained checkpoints: uses checkpoint
README.md