ComfyUI, a modular GUI for stable diffusion models, has experienced notable user-reported performance issues related to memory management and GPU utilization, despite active community involvement and development efforts.
The project, designed to facilitate complex workflows for stable diffusion models through a graphical interface, continues to evolve with contributions from a diverse development team. Recent activities have focused on addressing performance inconsistencies, particularly with Flux models, and enhancing the platform's usability and functionality.
Recent issues and pull requests (PRs) have predominantly centered around performance optimization and bug resolution. Notable issues include #4561, which highlights prolonged execution times with Flux models, and #4556, reporting a CUDA error. These issues collectively suggest underlying inefficiencies in resource allocation or model handling.
The development team has been actively working on bug fixes and feature enhancements. Key contributors include:
These efforts indicate a strong focus on improving stability and performance across different hardware configurations.
Overall, while ComfyUI's development is marked by active contributions and community involvement, addressing performance challenges remains critical to maintaining user satisfaction and software reliability.
| Timespan | Opened | Closed | Comments | Labeled | Milestones |
|---|---|---|---|---|---|
| 7 Days | 126 | 76 | 284 | 8 | 1 |
| 14 Days | 198 | 111 | 575 | 12 | 1 |
| 30 Days | 323 | 173 | 1178 | 20 | 1 |
| All Time | 2876 | 1242 | - | - | - |
Like all software activity quantification, these numbers are imperfect but sometimes useful. Comments, Labels, and Milestones refer to those issues opened in the timespan in question.
| Developer | Avatar | Branches | PRs | Commits | Files | Changes |
|---|---|---|---|---|---|---|
| Chenlei Hu |  |
1 | 11/9/2 | 9 | 122 | 283507 |
| comfyanonymous |  |
2 | 3/3/0 | 167 | 89 | 52904 |
| guill |  |
1 | 1/1/0 | 2 | 23 | 3164 |
| Robin Huang |  |
3 | 6/5/2 | 17 | 30 | 1590 |
| 来新璐 |  |
1 | 1/1/0 | 1 | 4 | 513 |
| pythongosssss |  |
1 | 0/1/1 | 1 | 8 | 258 |
| Alex "mcmonkey" Goodwin |  |
1 | 6/6/0 | 6 | 3 | 160 |
| PhilWun |  |
1 | 2/1/0 | 1 | 1 | 64 |
| Jaret Burkett |  |
1 | 1/1/0 | 1 | 1 | 53 |
| Ashen |  |
1 | 0/0/0 | 1 | 2 | 51 |
| Silver |  |
1 | 3/3/0 | 3 | 4 | 29 |
| Xrvk |  |
1 | 1/1/0 | 1 | 2 | 25 |
| Matthew Turnshek |  |
1 | 1/1/0 | 1 | 2 | 15 |
| bymyself |  |
1 | 3/3/0 | 3 | 3 | 14 |
| filtered (webfiltered) | 1 | 1/1/0 | 1 | 1 | 7 | |
| ljleb |  |
1 | 1/1/0 | 1 | 1 | 5 |
| Svein Ove Aas |  |
1 | 1/1/0 | 1 | 1 | 4 |
| Yoland Yan |  |
2 | 2/1/1 | 2 | 2 | 4 |
| fgdfgfthgr-fox |  |
1 | 1/1/0 | 1 | 2 | 4 |
| Vladimir Semyonov |  |
1 | 1/1/0 | 1 | 1 | 3 |
| David |  |
1 | 1/1/0 | 1 | 1 | 3 |
| Alexander Brown |  |
1 | 2/1/1 | 1 | 1 | 2 |
| a-One-Fan |  |
1 | 1/1/0 | 1 | 1 | 2 |
| TTPlanetPig |  |
1 | 1/1/0 | 1 | 1 | 2 |
| Jairo Correa |  |
1 | 1/1/0 | 1 | 1 | 1 |
| woolion (Le09) | 0 | 1/0/0 | 0 | 0 | 0 | |
| None (bash-j) | 0 | 1/0/0 | 0 | 0 | 0 | |
| Umut Polat (umutpo) | 0 | 1/0/1 | 0 | 0 | 0 | |
| .raw (crmbz0r) | 0 | 1/0/1 | 0 | 0 | 0 | |
| Kel Phillipson (kelfish) | 0 | 2/0/2 | 0 | 0 | 0 | |
| Mel Massadian (melMass) | 0 | 2/0/2 | 0 | 0 | 0 | |
| Njal Kuhn (njalooo) | 0 | 1/0/0 | 0 | 0 | 0 | |
| Yuta Hayashibe (shirayu) | 0 | 1/0/1 | 0 | 0 | 0 | |
| 311_code (311-code) | 0 | 1/0/1 | 0 | 0 | 0 | |
| Hanmo (Hanmo123) | 0 | 1/0/1 | 0 | 0 | 0 | |
| Harsh Bhardwaj (Harshsfd) | 0 | 1/0/1 | 0 | 0 | 0 | |
| Markus Hartung (hartmark) | 0 | 1/0/0 | 0 | 0 | 0 | |
| John Slegers (jslegers) | 0 | 2/0/0 | 0 | 0 | 0 | |
| Dr.Lt.Data (ltdrdata) | 0 | 1/0/0 | 0 | 0 | 0 | |
| Richard (xuhuaren) | 0 | 1/0/2 | 0 | 0 | 0 | |
| None (chaObserv) | 0 | 1/0/0 | 0 | 0 | 0 | |
| Pam (pamparamm) | 0 | 1/0/0 | 0 | 0 | 0 | |
| Zelin P Peter (PeterTF656) | 0 | 1/0/1 | 0 | 0 | 0 | |
| zhumingdaiyue (Yuanpei-Wu) | 0 | 1/0/1 | 0 | 0 | 0 | |
| Ruixi Fan (rfan-debug) | 0 | 1/0/1 | 0 | 0 | 0 | |
| KarryCharon (KarryCharon) | 0 | 1/0/1 | 0 | 0 | 0 | |
| Alistor (UrzasLegacy) | 0 | 1/0/1 | 0 | 0 | 0 | |
| jiaqianjing (jiaqianjing) | 0 | 1/0/0 | 0 | 0 | 0 | |
| None (shawnington) | 0 | 2/0/1 | 0 | 0 | 0 | |
| None (Extraltodeus) | 0 | 1/0/0 | 0 | 0 | 0 | |
| None (chrisgoringe) | 0 | 1/0/1 | 0 | 0 | 0 | |
| Shenghang Tsai (jackalcooper) | 0 | 1/0/1 | 0 | 0 | 0 | |
| Yaohui (ZhuYaoHui1998) | 0 | 1/0/1 | 0 | 0 | 0 | |
| Mr.G (RazvanManolache) | 0 | 1/0/1 | 0 | 0 | 0 | |
| Rahul Vadisetty (RahulVadisetty91) | 0 | 1/0/1 | 0 | 0 | 0 | |
| None (RandomGitUser321) | 0 | 1/0/1 | 0 | 0 | 0 |
PRs: created by that dev and opened/merged/closed-unmerged during the period
The ComfyUI project has seen significant recent activity, with 1,634 open issues currently logged. A notable trend is the emergence of various user support requests and bug reports related to model loading, performance issues, and compatibility with different hardware configurations. Many users report performance degradation following updates, particularly concerning memory management and GPU utilization. The community appears to be actively discussing workarounds and fixes for these issues, indicating a robust engagement with the project.
Several issues highlight a common theme of performance inconsistencies when switching between workflows or models, especially with the Flux models. Users are experiencing increased memory usage and slower processing times after updates, which could indicate underlying problems in the model handling or resource allocation mechanisms.
Issue #4561: flux-dev, executed in long time.
Issue #4560: How to determine whether a custom node is using GPU or CPU.
Issue #4557: When there is an error in obtaining a new frontend, an empty file with the latest version number of the new frontend will be created locally.
Issue #4556: CUDA error: CUBLAS_STATUS_NOT_SUPPORTED.
Issue #4555: ERROR Lora diffusion_model.
Issue #4552: Undo feature for textareas.
Issue #4550: Broken prompt weighting.
Issue #4549: Comfyui + Flux + Lora = Noise after only a few render.
Issue #4547: Add an option to define the IDs used in the API workflow.
Issue #4538: After updating to the latest version, flux nf4 becomes very slow.
Overall, the data reflects a vibrant user community engaged in troubleshooting and enhancing their experience with ComfyUI while also highlighting areas where further development may be necessary to improve stability and performance across diverse environments.
The dataset comprises a collection of pull requests (PRs) from the ComfyUI project, highlighting both open and closed PRs. The total number of open PRs is 151, with a variety of features, bug fixes, and enhancements being proposed or implemented.
#4558: Add denoise percentage to BetaSamplingScheduler node. This PR introduces a new feature to adjust the denoise percentage, which could enhance user control over image generation.
#4553: Added SamplerHeunCFGpp. This PR implements a new sampling method that aims to improve the efficiency of the CFG scale in image generation.
#4518: Fix nondeterministic results when add_noise==disable. This addresses a bug related to generating images consistently without noise.
#4504: Added minimal Dockerfiles and docker-compose.yaml. This PR facilitates easier deployment via Docker, although it lacks testing for Nvidia setups.
#4454: Add SA-Solver Sampler. A new sampler is introduced, which could enhance the model's performance based on stochastic processes.
#4385: Save checkpoint VAE optional. This PR makes the VAE optional during checkpoint saving, improving flexibility in model management.
#4374: Add type hints to main.py and server.py. This enhances code readability and maintainability by adding type annotations.
#4369: Add one-key update plugins script by command tools for Linux. This adds convenience for users managing plugins on Linux systems.
#4365: Improve extra_model_paths template by adding missing folder paths. This addresses user confusion regarding model paths.
#4551: Fix task.status.status_str caused by #2666. This PR resolves an issue where task statuses were incorrectly reported due to changes in the underlying codebase.
#4548: Generalize MacOS version check for force-upcast-attention. This PR improves compatibility with various MacOS versions, addressing issues with image generation.
#4535: Fix a bug where cached outputs affected IS_CHANGED. This ensures that cached outputs do not interfere with workflow execution logic.
#4522: Update add unet path for extra model path template. This improves documentation clarity regarding model paths.
The analysis of the current state of pull requests reveals several key themes and observations:
A significant number of open pull requests focus on enhancing existing features or introducing new functionalities to ComfyUI. For instance, PRs like #4558 (denoise percentage) and #4553 (SamplerHeunCFGpp) indicate ongoing efforts to refine user experience and improve image generation capabilities. These enhancements are crucial as they directly impact user satisfaction and the overall effectiveness of the tool in generating high-quality images.
Conversely, many pull requests also address bugs, such as #4518 (nondeterministic results) and #4385 (optional VAE), showcasing an active commitment to maintaining software reliability. The presence of labels like "Bug" on several PRs indicates that developers are responsive to issues raised by users, which is vital for community trust and engagement.
The discussions within the pull requests reflect a vibrant community actively contributing to the project. For example, comments on PR #4454 regarding the SA-Solver Sampler highlight collaborative efforts among contributors to optimize performance based on research findings. The back-and-forth dialogue in these discussions suggests that contributors are not only focused on code but also on sharing knowledge and improving project outcomes collectively.
Several pull requests aim to improve documentation or usability aspects of ComfyUI, such as #4504 (Docker support) and #4365 (extra_model_paths template). These improvements are essential for onboarding new users and ensuring that existing users can leverage all features effectively without confusion. The emphasis on enhancing documentation indicates an understanding that clear guidance can significantly reduce barriers to entry for potential users.
While there is a healthy number of open pull requests, some have been open for several days without merges or significant updates (e.g., PRs created 11 days ago). This may suggest resource constraints or prioritization challenges within the development team. It raises concerns about whether all contributions are being adequately reviewed and integrated into the main codebase in a timely manner.
Some pull requests have raised concerns about potential bugs or design flaws, such as those noted in PR #4442 regarding Flux model support. The discussions surrounding these anomalies highlight the importance of thorough testing before merging changes into the main branch. Ensuring that new features do not introduce regressions is critical for maintaining software stability.
In summary, while there is a robust flow of contributions aimed at enhancing ComfyUI's functionality and usability, attention must be given to ensuring timely reviews and merges of pull requests to maintain momentum in development efforts. Additionally, ongoing communication within the community will be vital in addressing any emerging challenges or concerns effectively.
Chenlei Hu (huchenlei)
comfyanonymous
David Benton (davidbenton)
Guill
Baughn
Alex "mcmonkey" Goodwin (mcmonkey4eva)
Robin Huang (robinjhuang)
Yoland Yan (yoland68)
Ashen
Christian Byrne (bymyself)
Eero Heikkinen (jaretburkett)
Matthew Turnshek (mturnshek)
Pythongosssss
Collaboration: Most contributions appear to be individual efforts without explicit collaboration noted among team members, suggesting a distributed work style where developers focus on their assigned tasks or features independently.
Focus on Bug Fixes and Optimizations: A significant portion of recent commits is dedicated to fixing bugs, particularly related to memory management and performance issues across different GPU configurations, indicating a priority on stability and efficiency.
Feature Enhancements: The team is actively adding new features such as API routes and shortcuts that enhance usability and functionality of the ComfyUI platform, reflecting an ongoing commitment to improving user experience.
Code Maintenance: Regular code cleanups and refactoring efforts are evident, which are essential for maintaining code quality as the project evolves.
Diversity of Contributions: The variety of contributions—from UI improvements to backend optimizations—demonstrates a well-rounded development effort aimed at enhancing both user interface and underlying architecture of ComfyUI.
Overall, the development team is engaged in a robust cycle of feature enhancement, bug fixing, and code maintenance, contributing to the ongoing evolution of the ComfyUI project while addressing user needs effectively.
Summary
Quantified Reports
Detailed Reports