34 lines
964 B
YAML
34 lines
964 B
YAML
---
|
|
date: 2026-03-08
|
|
day_of_week: Sunday
|
|
task_id: FRE-29
|
|
title: Parallelize Segment Generation
|
|
status: todo
|
|
company_id: FrenoCorp
|
|
objective: Generate multiple segments in parallel when possible
|
|
context: |
|
|
- Current generation processes segments sequentially
|
|
- GPU can handle multiple inference requests
|
|
- Need to manage concurrency and memory carefully
|
|
issue_type: enhancement
|
|
priority: medium
|
|
assignee: Atlas
|
|
parent_task: FRE-32
|
|
goal_id: MVP_Pipeline_Working
|
|
blocking_tasks: []
|
|
expected_outcome: |
|
|
- Multiple segments generated concurrently
|
|
- Memory usage controlled via batch size
|
|
- Speedup proportional to GPU capability
|
|
acceptance_criteria:
|
|
- Parallel generation mode available
|
|
- Configurable max concurrent segments
|
|
- No OOM errors with reasonable batch sizes
|
|
|
|
notes:
|
|
- Use torch.inference_mode() for efficiency
|
|
- Monitor GPU memory usage
|
|
|
|
links:
|
|
batch_processor: /home/mike/code/AudiobookPipeline/src/generation/batch_processor.py
|
|
--- |