Dhar Rawal

How to design a system to provide long answers

6

I am not sure how to train DSPy to synthesize long, detailed answers such as those are required for "how to" questions. So far, I have tried training on long...

Per-folder favorites?

1

Instead of one global favorites file... - Allow favorites file per folder - Build favorites list by appending all favorite files in folder hierarchy all the way to $HOME (favorites...

Fixed issue #775, #824, #894 and refactored Ensemble optimizer

7

EnsembledProgram can now be loaded/saved ~~Inference of programs (forward function) is now concurrent~~ Added checks for size = 0 or size > number of programs

Dev issue - Ruff check succeeds but commit fails with ruff check message

2

It seems like if you touch the file, commit requires you to fix all the ruff issues in that file, regardless of whether your changes are related to those lines...

Crashes with ValueError exception in predict.py - "Required 'max_tokens' or 'max_output_tokens' not specified in settings." if you use Bedrock with llama2 or llama3

Mipro crash in mipro_optimizer.py

2

Notice in the watch window that there is nothing in the example_sets for i=3. Needs a check here? 

Optimizer issue: Using a pre-compiled module as a teacher causes an assertion crash in bootstrap.py

9

Assertion: Student and teacher must have the same number of predictors. Code sample to repro: optimizer = BootstrapFewShotWithRandomSearch( metric=MyModule.metric, teacher_settings=dict({"lm": teacher_lm}), ) teacher_module, _ = MyModule.load_compiled_model() for _ in range(2):...

Error loading program compiled using the Ensemble optimizer

4

The app crashes in below code with a keyerror: File "/home/drawal/miniconda3/envs/xray/lib/python3.11/site-packages/dspy/primitives/module.py", line 75, in load_state param.load_state(state[name]) This is because the load function expects the top level keys in the saved...