Train multi-node tasks including cycle multi-hop rag and trainable agent #306
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
…e, output graph as simplified version of the dynamic computation graph
|
Check out this pull request on See visual diffs & provide feedback on Jupyter Notebooks. Powered by ReviewNB |
…es the component trace, and make the backward engine with only eval mode and without output a parameter
…t all gradients correctly in moving batch
… and feedback will be present in sequential order
liyin2015
changed the title
liyin2015
changed the title
… graidents which makes the input args way too big, and add parameter and output parameter allowed_components
…n the debug report
…dient pass through grad component correction, the code follows the 3 cases of gradcomponent design now, still need to further optimize the agent
…ckward engine and optimizer, has made it 2% higher accuracy in step 3 than before
…nt clearly and make many previous component only a class
…ring which is quite high with gpt35
…roved the joined training experiemnt tracking
…d pass setup for the backward engine via the trainer.fit function
…gineering strategy
…ain multi-hop retriever
… instead of just pass-through gradient
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
, trainable, added component graph with cycl…e, output graph as simplified version of the dynamic computation graph
Auto-optimize an agent with planner.
The graph is from the text grad debug mode. And when click on a node, it will show all details on the gradients.
Big updates to the codebase:
Prompt,Parser,OutputParser. The visualization of these parts need to be done manually to whichever use it and does not come out of box. This is not important for now.