Michael Poluektov

@sushreebarsa Thanks for your reply! I'm working on a TFLite backend, and unfortunately I don't get to pick which operators our clients use. I hacked together a pass that does...

From the [official PyTorch implementation](https://github.com/juho-lee/set_transformer/blob/master/modules.py): ``` class PMA(nn.Module): def __init__(self, dim, num_heads, num_seeds, ln=False): super(PMA, self).__init__() self.S = nn.Parameter(torch.Tensor(1, num_seeds, dim)) nn.init.xavier_uniform_(self.S) self.mab = MAB(dim, dim, dim, num_heads, ln=ln) def...

I started an issue but no PR yet. Adding this requires quite a major refactor, and I'll most likely submit a draft this weekend. Issue tracked here: https://github.com/open-webui/open-webui/issues/4219

I like this. Then we can delete all of the unused stuff from requirements.txt and have an even slimmer image. I might be wrong but I don't think `langfuse`, `anthropic`,...

I would say "just use pipelines" but since so much of the functionality in functions is currently not supported by pipelines (`__event_emitter__`, `__event_call__`) I think this is a good option.

https://github.com/open-webui/open-webui/pull/4724

The way I would implement a Python tool would be to define a function that takes no external parameters other than `__messages__` which calls `eval` on the last snippet of...

The goal is eventually not to have tool outputs only be pre-pended on each query, but potentially called mid-way through the generation. Similar to https://openwebui.com/f/michaelpoluektov/openai_react/ (I'm waiting on Ollama to...

> I have working on lately and I have working version (surely still has bugs and needs some cleaning). If anyone is interested, it's here: > > https://github.com/smonux/open-webui/blob/dev-apitools/backend/open_webui/main.py > >...

Tested chatting with Ollama and made sure params were properly passed through. Didn't test Ollama V1 endpoints since I don't know where they're being called from anyway.