brittle parser in tgd_optimizer

[mrdrprofuroboros]

Describe the bug

TypeError                                 Traceback (most recent call last)
Cell In[7], line 1
----> 1 trainer.fit(
      2     train_dataset=filtered_dataset.train,
      3     val_dataset=filtered_dataset.val,
      4     test_dataset=filtered_dataset.test,
      5     resume_from_ckpt=\"/Users/imatas/.adalflow/ckpt/TransitionCorrectnessAdalComponent/constrained_max_steps_3_72e3f_run_1.json\",
      6 )

File ~/opt/anaconda3/envs/aimon/lib/python3.11/site-packages/adalflow/optim/trainer/trainer.py:479, in Trainer.fit(self, adaltask, train_loader, train_dataset, val_dataset, test_dataset, debug, save_traces, raw_shots, bootstrap_shots, resume_from_ckpt)
    477     starting_step += self.max_steps
    478 elif self.strategy == \"constrained\":
--> 479     trainer_results = self._fit_text_grad_constraint(
    480         train_loader,
    481         val_dataset,
    482         test_dataset,
    483         trainer_results=trainer_results,
    484         starting_step=starting_step,
    485     )
    486     starting_step += self.max_steps
    487 else:

File ~/opt/anaconda3/envs/aimon/lib/python3.11/site-packages/adalflow/optim/trainer/trainer.py:1781, in Trainer._fit_text_grad_constraint(self, train_loader, val_dataset, test_dataset, trainer_results, starting_step)
   1775 all_losses.extend(losses)
   1776 all_y_preds.extend(
   1777     [y.full_response for y in y_preds if isinstance(y, Parameter)]
   1778 )
   1780 all_samples, all_losses, all_y_preds = (
-> 1781     self._text_grad_constraint_propose_step(
   1782         steps=steps,
   1783         all_samples=all_samples,
   1784         all_losses=all_losses,
   1785         all_y_preds=all_y_preds,
   1786     )
   1787 )
   1789 # check optimizer stages to see if the proposal was accepted so far
   1790 if not self._check_optimizer_proposal():

File ~/opt/anaconda3/envs/aimon/lib/python3.11/site-packages/adalflow/optim/trainer/trainer.py:1659, in Trainer._text_grad_constraint_propose_step(self, steps, all_samples, all_losses, all_y_preds, include_demo_optimizers)
   1654 tdqm_loader = tqdm(range(self.max_proposals_per_step), desc=\"Proposing\")
   1655 for i in tdqm_loader:
   1656 
   1657     # print(f\"Proposing step: {i}\")
   1658     # self.optimizer.propose()
-> 1659     self._propose_text_optimizers()  # new prompts
   1660     if include_demo_optimizers:
   1661         self._demo_optimizers_propose()

File ~/opt/anaconda3/envs/aimon/lib/python3.11/site-packages/adalflow/optim/trainer/trainer.py:859, in Trainer._propose_text_optimizers(self)
    857 def _propose_text_optimizers(self):
    858     for text_optimizer in self.text_optimizers:
--> 859         text_optimizer.propose()

File ~/opt/anaconda3/envs/aimon/lib/python3.11/site-packages/adalflow/optim/text_grad/tgd_optimizer.py:334, in TGDOptimizer.propose(self)
    331 log.info(f\"Response from the optimizer: {response}\")
    332 # extract the improved variable from the response
    333 # TODO: make it more robust
--> 334 improved_variable = extract_new_variable(proposed_data)
    335 param.propose_data(improved_variable)
    336 if self.do_gradient_memory:

File ~/opt/anaconda3/envs/aimon/lib/python3.11/site-packages/adalflow/optim/text_grad/tgd_optimizer.py:129, in extract_new_variable(text)
    126 pattern = re.compile(r\"<VARIABLE>(.*?)</VARIABLE>\", re.DOTALL)
    128 # Find all matches
--> 129 matches = pattern.findall(text)
    131 if len(matches) == 0:
    132     return text.strip()

TypeError: expected string or bytes-like object, got 'NoneType'"

To Reproduce
trainer.fit with openai gpt-4o

Expected behavior
graceful handling

[mrdrprofuroboros]

ah, found the core issue

Error code: 429 - {'error': {'message': 'Request too large for gpt-4o in organization ... on tokens per min (TPM): Limit 30000, Requested 37640. The input or output tokens must be reduced in order to run successfully. Visit https://platform.openai.com/account/rate-limits to learn more.', 'type': 'tokens', 'param': None, 'code': 'rate_limit_exceeded'}}

shall we propogate that somehow to user? might not worth the effort. reducing the optimization prompt size would be a good idea though :)