docs: Remove numpy from auto-excluded build packages#581
Conversation
|
|
||
|
|
||
| <Tip> | ||
|
|
There was a problem hiding this comment.
Citation: PR #263 renames BASE_IMAGE_PACKAGES to SIZE_PROHIBITIVE_PACKAGES and removes numpy from the list. The root cause was that the blacklist was defined by GPU base image contents, not size constraints, which silently broke CPU endpoints using python-slim that needed numpy. See changes in src/runpod_flash/cli/commands/build.py.
View source
|
|
||
| - `torch`, `torchvision`, `torchaudio` | ||
| - `numpy`, `triton` | ||
| - `torch`, `torchvision`, `torchaudio`, `triton` |
There was a problem hiding this comment.
Citation: Updated to match the new SIZE_PROHIBITIVE_PACKAGES constant in src/runpod_flash/cli/commands/build.py. Each package is annotated with its approximate size: torch (~500 MB), torchvision (~50 MB), torchaudio (~30 MB), triton (~150 MB).
View source
|
Just a reminder: If you'd like me to act on any feedback you have via Github comments, just type @Promptless in your suggestion and I'll get right on it! (I won't show up in the user dropdown, but I'll process any request that has @Promptless in the comment body.) |
|
Preview deployment for your docs. Learn more about Mintlify Previews.
|
Open this suggestion in Promptless to view citations and reasoning process
Updates the
flash builddocs to remove numpy from the auto-excluded packages list. NumPy was incorrectly being stripped from CPU endpoint builds because CPU base images (python-slim) don't ship it pre-installed. Now clarifies that only CUDA/GPU packages (torch,torchvision,torchaudio,triton) are auto-excluded.Trigger Events
runpod/flash PR #263: fix(build): remove numpy from dependency blacklist, rename to SIZE_PROHIBITIVE_PACKAGES
Promptless Research (6 files, 1 GitHub PR)
.long_term_context/product_knowledge/product_overview.md.long_term_context/doc_workflow/client_instructions.mdflash/cli/build.mdxflash/apps/build-app.mdxflash/troubleshooting.mdx.long_term_context/style/client_style_guide.mdAgent Response
Tip: Request one-off documentation tasks in the Dashboard under New Task 🚀