34 lines
971 B
YAML
34 lines
971 B
YAML
---
|
|
date: 2026-03-08
|
|
day_of_week: Sunday
|
|
task_id: FRE-28
|
|
title: Optimize Generation Speed for Long Books
|
|
status: todo
|
|
company_id: FrenoCorp
|
|
objective: Reduce generation time for books with many segments
|
|
context: |
|
|
- Current generation is sequential and slow
|
|
- Can optimize model inference and post-processing
|
|
- Batch processing improvements needed
|
|
issue_type: enhancement
|
|
priority: medium
|
|
assignee: Atlas
|
|
parent_task: FRE-32
|
|
goal_id: MVP_Pipeline_Working
|
|
blocking_tasks: []
|
|
expected_outcome: |
|
|
- Generation time under 2x real-time for 1.7B model
|
|
- Efficient memory usage during long runs
|
|
- Configurable quality/speed tradeoffs
|
|
acceptance_criteria:
|
|
- Benchmark shows <2x real-time generation
|
|
- Memory stays stable during long books
|
|
- Speed/quality options available
|
|
|
|
notes:
|
|
- Profile generation pipeline to find bottlenecks
|
|
- Consider model quantization for speed
|
|
|
|
links:
|
|
tts_model: /home/mike/code/AudiobookPipeline/src/generation/tts_model.py
|
|
--- |