Replies: 8 comments 1 reply
-
Thanks for your interest in contributing. 🙂
No problem. No time pressure at all. 🙂
Do you mean https://github.com/GimelStudio/GimelStudio/blob/master/src/nodes/corenodes/input/string_node.py?
I must admit I forgot we only had an integer node (https://github.com/GimelStudio/GimelStudio/blob/master/src/nodes/corenodes/input/integer_node.py).
Use the There will be an announcement coming in the near-ish future. There is barely anything in the
I must admit I haven't done too much regarding machine learning, but it sounds like it would be nice. What you've listed seems to be relatively useful.
Go for it. 🙂 |
Beta Was this translation helpful? Give feedback.
-
Are we still working on this? |
Beta Was this translation helpful? Give feedback.
-
Life has gotten in the way a bit, but one approach I have considered is a
node that can detect and interact with something like Automatic1111 for
Stable Diffusion. This wouldn’t be the end goal of the nodes as it isn’t
low-code enough, but it’s a starting point that power users could take
advantage of
|
Beta Was this translation helpful? Give feedback.
-
My thought on this would be that at least the image generation would be out of scope for the Gimel Studio core nodes. The sd image generation and perhaps even the other ones relying on models could be a separate node that users could install, but maybe not something to include in the "core" nodes shipped with Gimel Studio. I'm not really aware of how this works on a practical level though. Is it an api that does this or do you have to have a copy of the ml model on your computer in order for it to work? |
Beta Was this translation helpful? Give feedback.
-
I imagine that we could give users the ability to add their own nodes (similar to blender addons). |
Beta Was this translation helpful? Give feedback.
-
Correct me if I'm wrong @Correct-Syntax, but support for custom nodes already exist. However, add-ons (as in, adding additional functionality to the app not including nodes) is not implemented. |
Beta Was this translation helpful? Give feedback.
-
If A1111 is the chosen route for ML nodes then this could actually go in as an 'OpenAPI'/ 'HTTP'/ 'Fetch' node instead. This change would decouple from having to download models at all and could bring in other remote services - it'd make sense as a core node? For ML itself, I still think having nodes capable of handling the downloads& execution is useful, and the recent Hugging Face Transformer Agents provide a useful API for that, on top of the existing Transformers library. See: https://huggingface.co/docs/transformers/transformers_agents These come with the caveat that image generation updates might then be better behind a user-triggered button, instead of on-change (due to time/memory/network use). Not ideal, but if nodes could have a 'forceFlowToManualGeneration' bool on them, the best of both could be achieved? |
Beta Was this translation helpful? Give feedback.
-
I'm unsure where GS is up to with multithreading but the Python 3.12 sub-interpreter features are exciting for true asynchronous tasks. It's still a few months out but this may well become necessary to use for Fetch and ML nodes. Release notes: https://docs.python.org/3.12/whatsnew/3.12.html |
Beta Was this translation helpful? Give feedback.
-
Hi, I'm new to the project and it's exactly what I'm looking for so I'm keen to start developing some custom machine learning based nodes (timeframe TBC...). I'm thinking about:
Possible ancillary nodes needed:
Questions:
Beta Was this translation helpful? Give feedback.
All reactions