opto.optimizers.textgrad.GRADIENT_OF_RESULTS_INSTRUCTION#

GRADIENT_OF_RESULTS_INSTRUCTION = 'For the search results from {engine_name} we got the following feedback:\n\n<FEEDBACK>{results_gradient}</FEEDBACK>\n\n'#

reduce / sum

Type:

Gradient accumulation