wangbo/ComfyUI
1
0
Fork 0
mirror of https://github.com/comfyanonymous/ComfyUI.git synced 2026-05-25 00:17:23 +08:00
Code Issues Actions 20 Packages Projects Releases Wiki Activity

Use cond_shapes in multigpu memory-fit check (parity with single-GPU path)

Browse Source 
The multigpu cond-batching loop called model.memory_required(input_shape) without conditioning shapes, while the single-GPU path at line 279 passes cond_shapes. Large conditioning tensors (e.g. video prompts, control inputs) were therefore under-counted, risking OOM at runtime when the chosen batch size was too large. Match the single-GPU pattern by building cond_shapes from each batched cond's conditioning dict and passing it to memory_required.

Amp-Thread-ID: https://ampcode.com/threads/T-019e43b8-8258-70fd-ab3a-53e4c97f85d5
Co-authored-by: Amp <amp@ampcode.com>
This commit is contained in:
Jedrzej Kosinski 2026-05-20 19:52:03 -07:00
parent dd85851efe
commit ac0a90c323
1 changed files with 5 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

6
comfy/samplers.py
View File

@ -433,7 +433,11 @@ def _calc_cond_batch_multigpu(model: BaseModel, conds: list[list[dict]], x_in: t
for i in range(1, len(to_batch_temp) + 1): for i in range(1, len(to_batch_temp) + 1):
batch_amount = to_batch_temp[:len(to_batch_temp)//i] batch_amount = to_batch_temp[:len(to_batch_temp)//i]
input_shape = [len(batch_amount) * first_shape[0]] + list(first_shape)[1:] input_shape = [len(batch_amount) * first_shape[0]] + list(first_shape)[1:]
if model.memory_required(input_shape) * 1.5 < free_memory: cond_shapes = collections.defaultdict(list)
for tt in batch_amount:
for k, v in to_run[tt][0].conditioning.items():
cond_shapes[k].append(v.size())
if model.memory_required(input_shape, cond_shapes=cond_shapes) * 1.5 < free_memory:
to_batch = batch_amount to_batch = batch_amount
break break