Reusing a subprocess after unsloth patches torch internals causes
inspect.getsource() failures when loading a different model type.
Each load now gets a clean Python interpreter.
Replaces cmd_queue-based cancel polling with a shared mp.Event.
Fixes two issues:
- Loading a new model while generating no longer hangs (cancel is instant)
- Subprocess shuts down cleanly after explicit stop generation
Inference now runs in a persistent subprocess, solving the same
transformers version-switching problem that was fixed for training.
The subprocess stays alive between requests (model in GPU memory)
and is only restarted when switching transformers versions.
New files:
- core/inference/worker.py: subprocess entry point with command loop
- core/inference/orchestrator.py: parent-side proxy with same API
Modified:
- core/inference/__init__.py: exports orchestrator as default backend
- routes/inference.py: removed in-process ensure_transformers_version()
These are standalone benchmark scripts that were force-added despite being
gitignored. They have no test functions and run network calls at module
level, which breaks pytest collection in CI.
- Add 200-sample parallel probe using ThreadPoolExecutor + safe_num_proc
to estimate download speed and failure rate before full conversion
- Abort with clear error if >=30% of probe images fail to download
- Show estimated download time in the training overlay modal
- Parallel batch conversion for URL-based datasets (vs sequential for local)
- Add warning field to /check-format response for URL-based image datasets
- Display URL warning in dataset preview dialog (amber banner)
- Thread progress_callback from trainer through format_and_template_dataset
to convert_to_vlm_format for real-time status updates
Place Train Split Start / End inputs inside the Advanced collapsible
with descriptive tooltips clarifying they slice the training split.
Revert the selectors component to its original eval-split-only layout.
Place Slice Start and Slice End inputs alongside the Eval Split
selector in a single row (grid-cols-3) so the dataset card stays
compact. Remove the duplicate controls from the Advanced section.
Add Start/End index inputs under Advanced in the dataset card,
allowing users to slice a dataset by row range before training.
Wired end-to-end: frontend store, API payload, backend Pydantic
model, and trainer dataset loading (inclusive on both ends).
Place Train Split Start / End inputs inside the Advanced collapsible
with descriptive tooltips clarifying they slice the training split.
Revert the selectors component to its original eval-split-only layout.
Place Slice Start and Slice End inputs alongside the Eval Split
selector in a single row (grid-cols-3) so the dataset card stays
compact. Remove the duplicate controls from the Advanced section.
Add Start/End index inputs under Advanced in the dataset card,
allowing users to slice a dataset by row range before training.
Wired end-to-end: frontend store, API payload, backend Pydantic
model, and trainer dataset loading (inclusive on both ends).
