Repository navigation

Issues

New issue

Dynamic VRAM streaming crashes all generations with HostBuffer.read_file_slice failed → CUDA OOM (regression after Aug 3 2026 update) (CORE-398)
#15255 · roumani76-arch opened on Aug 3, 2026
78

LTXVImgToVideo: first-frame anchoring broken in v0.44.x — first ~30 frames render black (regression vs v0.30.0)
Comfy-Org/ComfyUI#17618 · opened on Nov 8, 2026
Version mismatch in older stable releases (including v0.14.2 Windows portable)
Comfy-Org/ComfyUI#17616 · opened on Nov 8, 2026
sam3D template node is throwing error
Comfy-Org/ComfyUI#17613 · opened on Nov 8, 2026
3
[BUG][v0.43.0 and above] Flux2 fused VAE + --fast fp16_accumulation crashes on Turing/sm_75 (CUTLASS Mma<16,8,16> / CUDA 719)
Comfy-Org/ComfyUI#17612 · opened on Nov 8, 2026
TextGenerate crashes on Qwen3-VL 8B since #17361: linear(): argument 'weight' must be Tensor, not NoneType
Comfy-Org/ComfyUI#17610 · opened on Nov 8, 2026
1
[Bug] Process crashes with CUDA error on second prompt execution with Dynamic VRAM enabled (Qwen 3.x Text Encoder)
Comfy-Org/ComfyUI#17609 · opened on Nov 8, 2026
关于relayvsr后台LOG进度条兼容问题
Comfy-Org/ComfyUI#17607 · opened on Nov 8, 2026
[Bug] Silent crash / Application Error in aimdo.dll during video generation on RTX 50-series
Comfy-Org/ComfyUI#17606 · opened on Nov 8, 2026
Merge List of Tiles to Image corrupts borders with FP16 intermediates
Comfy-Org/ComfyUI#17604 · opened on Nov 8, 2026
Minimaxh3 Tiled Artifact
Comfy-Org/ComfyUI#17600 · opened on Nov 7, 2026
Unload leaks int8_convrot (comfy-kitchen) weights — VRAM/RAM never released, next job OOMs (ROCm; plain fp8/fp16 unloads cleanly)
Comfy-Org/ComfyUI#17597 · opened on Nov 7, 2026
2
Severe UI Regressions in v0.44: Volatile History Assets, Missing Durations, and Broken Native Video Nodes
Comfy-Org/ComfyUI#17594 · opened on Nov 7, 2026
Asset scan aborts with SQLITE_CANTOPEN on a transient WAL handle race (no retry in _write_in_batches)
Comfy-Org/ComfyUI#17593 · opened on Nov 7, 2026
help with trellis2
Comfy-Org/ComfyUI#17592 · opened on Nov 7, 2026
How to enable SSD->VRAM directly?
Comfy-Org/ComfyUI#17590 · opened on Nov 7, 2026
pid1.5 4k image upscale much slower than before
Comfy-Org/ComfyUI#17588 · opened on Nov 7, 2026
AMD portable build: hardcoded path D:\a\ComfyUI\... causes offload-arch error
Comfy-Org/ComfyUI#17587 · opened on Nov 7, 2026
1
Security hardening: use folder_paths.is_within_directory (os.path.realpath) and os.path.basename consistently in server.py / user_manager.py
Comfy-Org/ComfyUI#17580 · opened on Nov 6, 2026
[Windows/NVIDIA] Severe system-wide UI stutter during generation on 0.44.1 (comfy-aimdo 0.6.0) - fine on 0.42.3 (0.5.6), independent of model size
Comfy-Org/ComfyUI#17578 · opened on Nov 6, 2026
2
For Security Add "None" / Clear option to resource loader nodes (Load LoRA, Image, Video, Audio)
Comfy-Org/ComfyUI#17577 · opened on Nov 6, 2026
Text Encode Qwen Image 2.1 failed
Comfy-Org/ComfyUI#17574 · opened on Nov 6, 2026
ROCm gfx1100: Comfy model compiler + auto-detected fast_disk adds ~30 s per sampling step; either --disable-comfy-compiler or --disable-fast-disk fixes it
Comfy-Org/ComfyUI#17573 · opened on Nov 6, 2026
ROCm gfx1100: contiguous-k/v SDPA workaround is gfx1151-only, MiniMax H3 sampling 2.2x slower on RX 7900 XTX
Comfy-Org/ComfyUI#17572 · opened on Nov 6, 2026
Execution cache silently replays stale outputs when a reference image file changes (autogrow images inputs are not folded into downstream cache signatures)
Comfy-Org/ComfyUI#17552 · opened on Nov 5, 2026
2
AttributeError: 'ImageModelDescriptor' object has no attribute 'patcher'
Comfy-Org/ComfyUI#17551 · opened on Nov 5, 2026
1

Navigation Menu

Repository navigation

Issues

Issues

LTXVImgToVideo: first-frame anchoring broken in v0.44.x — first ~30 frames render black (regression vs v0.30.0)

Version mismatch in older stable releases (including v0.14.2 Windows portable)

sam3D template node is throwing error

[BUG][v0.43.0 and above] Flux2 fused VAE + `--fast fp16_accumulation` crashes on Turing/sm_75 (CUTLASS `Mma<16,8,16>` / CUDA 719)

TextGenerate crashes on Qwen3-VL 8B since #17361: linear(): argument 'weight' must be Tensor, not NoneType

[Bug] Process crashes with CUDA error on second prompt execution with Dynamic VRAM enabled (Qwen 3.x Text Encoder)

关于relayvsr后台LOG进度条兼容问题

[Bug] Silent crash / Application Error in aimdo.dll during video generation on RTX 50-series

Merge List of Tiles to Image corrupts borders with FP16 intermediates

Minimaxh3 Tiled Artifact

Unload leaks int8_convrot (comfy-kitchen) weights — VRAM/RAM never released, next job OOMs (ROCm; plain fp8/fp16 unloads cleanly)

Severe UI Regressions in v0.44: Volatile History Assets, Missing Durations, and Broken Native Video Nodes

Asset scan aborts with SQLITE_CANTOPEN on a transient WAL handle race (no retry in _write_in_batches)

help with trellis2

How to enable SSD->VRAM directly?

pid1.5 4k image upscale much slower than before

AMD portable build: hardcoded path D:\a\ComfyUI\... causes offload-arch error

Security hardening: use `folder_paths.is_within_directory` (`os.path.realpath`) and `os.path.basename` consistently in `server.py` / `user_manager.py`

[Windows/NVIDIA] Severe system-wide UI stutter during generation on 0.44.1 (comfy-aimdo 0.6.0) - fine on 0.42.3 (0.5.6), independent of model size

For Security Add "None" / Clear option to resource loader nodes (Load LoRA, Image, Video, Audio)

Text Encode Qwen Image 2.1 failed

ROCm gfx1100: Comfy model compiler + auto-detected fast_disk adds ~30 s per sampling step; either --disable-comfy-compiler or --disable-fast-disk fixes it

ROCm gfx1100: contiguous-k/v SDPA workaround is gfx1151-only, MiniMax H3 sampling 2.2x slower on RX 7900 XTX

Execution cache silently replays stale outputs when a reference image file changes (autogrow images inputs are not folded into downstream cache signatures)

AttributeError: 'ImageModelDescriptor' object has no attribute 'patcher'