Zachary Huang

i still have this error: (base) jupyter@cuda-11-0-20230324-133223:~/llama.cpp/build$ cmake --build . --config Release [ 5%] Built target ggml [ 16%] Built target llama [ 22%] Linking CXX executable ../bin/test-quantize /usr/bin/ld: ../libllama.a(ggml.c.o):...

> Thanks! I reinstall CUDA and torch, update tutel to the latest version, and it works! Thanks for your patience, that really helps me a lot. Can you share your...

Hey thank you for the question! > So I am trying to use params to make a flow_id variable. Sub flow will be the next step, but I am having...

> It could also be nice for a node to know where it is in a Flow or a subFlow like print(self.Flow.node_history) and find parent Flow if we are in...

> It seems that rework2_Flow doesn't get the param rework2_flow_min_count if rework_Flow was initialized before. I don't fully follow the issue here. I tried to run your codes with a...

I've created an example function that tracks call stack for nodes: https://the-pocket.github.io/PocketFlow/viz.html Minimal examples that will work for your example: ```python from pocketflow import * import os import inspect def...

Will fix it this weekend, or do you mind submit a PR?

Many different designs. You can have an agent that has access to both tools: https://github.com/The-Pocket/PocketFlow/tree/main/cookbook/pocketflow-agent And decide the next tool iteratively, e.g., based on the tool A's result Or if...

Thank you! For parallelization, Python’s GIL prevents true CPU-bound parallelism. But LLM calls are I/O-bound, so we can exploit async/await to overlap requests: https://github.com/openai/openai-python?tab=readme-ov-file#async-usage Below is a dummy example using...

通常需要一个declaratively well defined 的 shared data structure e.g., https://github.com/The-Pocket/PocketFlow-Tutorial-Codebase-Knowledge/blob/main/docs/design.md#shared-store 然后所有的nodes都遵守这个design 来prep & post 我发现如果shared store is well define, adapt to a new structure完全可以来copilot (like Claude code my current favorite)轻松实现