feat: genetic algo based optimizer

[shahules786]

metric =AspectCritic(name="answer_correctness",definition="Given the user_input, reference and response. Is the response correct compared with the reference",llm=llm_4o)
metric.train("alignment_sample.json")

dummy data

[shahules786]

The progress bar for this now is not neat. I would like to have it as one single bar with a subprocess shown with progress. I'm not sure what the best way to do it is.

[jjmachan]

@shahules786 check out this

ragas/src/ragas/executor.py

Lines 146 to 181 in 99d1a47

	# With batching, show nested progress bars
	batches = batched(self.jobs, self.batch_size) # generator of job tuples
	n_batches = (len(self.jobs) + self.batch_size - 1) // self.batch_size
	with (
	tqdm(
	total=len(self.jobs),
	desc=self.desc,
	disable=not self.show_progress,
	position=1,
	leave=True,
	) as overall_pbar,
	tqdm(
	total=min(self.batch_size, len(self.jobs)),
	desc=f"Batch 1/{n_batches}",
	disable=not self.show_progress,
	position=0,
	leave=False,
	) as batch_pbar,
	):
	for i, batch in enumerate(batches, 1):
	batch_pbar.reset(total=len(batch))
	batch_pbar.set_description(f"Batch {i}/{n_batches}")
	# Create coroutines per batch
	coroutines = [
	afunc(*args, **kwargs) for afunc, args, kwargs, _ in batch
	]
	for future in await as_completed(coroutines, max_workers):
	result = await future
	results.append(result)
	overall_pbar.update(1)
	batch_pbar.update(1)
	return results

which does something similar for batching

[shahules786]

@jjmachan thanks, just fixed it. Please review

[jjmachan]

overall this works really well but I'm not sure about if we are using the executor correctly - will love to discuss this over a call before merging this in

also comments - the code is very hard to read since there are some new alogs here - proper variable names and comments will go a long way to improve the readability of the code