wangbo/ComfyUI
1
0
Fork 0
mirror of https://github.com/comfyanonymous/ComfyUI.git synced 2026-07-03 21:20:49 +08:00
Code Issues Actions 22 Packages Projects Releases Wiki Activity

Clarify tensor usage and structural metadata guidelines

Browse Source
This commit is contained in:
comfyanonymous 2026-07-02 16:58:26 -07:00 committed by GitHub
parent 8a207df688
commit 25b6a15700
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 8 additions and 5 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

13
AGENTS.md
View File

@ -177,11 +177,14 @@
- Do not use tensors as general-purpose Python data structures. Keep metadata, - Do not use tensors as general-purpose Python data structures. Keep metadata,
bookkeeping, counters, flags, shape math, padding math, index planning, memory bookkeeping, counters, flags, shape math, padding math, index planning, memory
estimates, and control-flow decisions in plain Python values unless the data estimates, and control-flow decisions in plain Python values unless the data
must participate directly in tensor computation. Split points, slice must participate directly in tensor computation. Do not create tensors for
boundaries, sequence offsets, and similar structural indices should be Python structural metadata that is only used for Python-side control flow. Sequence
ints/lists after validation, not CPU tensors kept only to drive Python-side lengths, cumulative offsets, split indices, window counts, slice boundaries,
control flow. Avoid creating temporary tensors just to use tensor methods for and repeat counts should be kept as Python ints/lists from the point they are
scalar or structural calculations. computed. Do not build them as CPU/GPU tensors and then cast, move, validate,
or convert them back to Python for `split`, `tensor_split`, indexing plans,
loops, or cache keys. Avoid creating temporary tensors just to use tensor
methods for scalar or structural calculations.
- Avoid unnecessary casts and transfers. Preserve the intended compute dtype, - Avoid unnecessary casts and transfers. Preserve the intended compute dtype,
storage dtype, bias dtype, and original tensor shape metadata. storage dtype, bias dtype, and original tensor shape metadata.
- Keep model-native latent layout handling inside the model or latent-format - Keep model-native latent layout handling inside the model or latent-format