Select any trained model, run all 5 analysis tasks or individual tasks, then test inference with user-controlled parameters.
Select a model and load.
⬜ 1. Start ⬜ 2. Load Model (checkpoint/config/device/eval) ⬜ 3. Load Tokenizers ⬜ 4. Input (IAST) ⬜ 5. Source Tokenization ⬜ 6. Encoder (run once) ⬜ 7. KV Cache prepared ⬜ 8. Initialize x_T (MASK) ⬜ 9. Diffusion loop (T→0, with Task2/Task3 hooks) ⬜ 10. Final x0 ⬜ 11. Decode to Devanagari ⬜ 12. Evaluation/Tasks (Task4/Task5)
Runs lightweight live task analysis instead of heavy runner.
Task 5 Controls