Search results

INT8 Krea2 models broken on Tesla T4 (SM75): CUDA invalid argument or corrupted outputs

Status: Open.

#55 In Comfy-Org/comfy-kitchen;

· PhoenixCreation opened on Jun 26, 2026

Python Python 3.13.11

Status: Open.

#45 In Comfy-Org/comfy-kitchen;

· saykorz opened on May 25, 2026

100 cpu when import

Status: Open.

#44 In Comfy-Org/comfy-kitchen;

· qwegjx opened on May 24, 2026

FP8 E4M3 quantized models inference is slower than same model quantized in MXFP8

Status: Open.

#38 In Comfy-Org/comfy-kitchen;

· matalama80td3l opened on May 18, 2026

The --no-cuda argument can't be passed through pyproject.toml

Status: Open.

#31 In Comfy-Org/comfy-kitchen;

· yurivict opened on Mar 19, 2026

Clarification regarding the readme example

Status: Open.

#27 In Comfy-Org/comfy-kitchen;

· sayakpaul opened on Jan 29, 2026

Hope to support older CUDA version backends

Status: Open.

#21 In Comfy-Org/comfy-kitchen;

· love530love opened on Jan 15, 2026

RTX30xx FP8 feather matmul triton kernel?

Status: Open.

#16 In Comfy-Org/comfy-kitchen;

· phazei opened on Jan 7, 2026

memory bug

Status: Open.

#10 In Comfy-Org/comfy-kitchen;

· rong7932 opened on Jan 6, 2026

INT8 support

Status: Open.

#9 In Comfy-Org/comfy-kitchen;

· kanttouchthis opened on Jan 6, 2026

cannot import name '_C' from partially initialized module 'comfy_kitchen.backends.cuda

Status: Open.

#5 In Comfy-Org/comfy-kitchen;

· lyh111901886 opened on Jan 6, 2026

Bug: Scales do not persist in fp32 when LoRA is applied.

Status: Open.

#2 In Comfy-Org/comfy-kitchen;

· silveroxides opened on Jan 5, 2026