DSPy: The framework for programming—not prompting—language models
Python
updating mipro tests
klopsahlongd7bf3bcff510
diff --git a/tests/teleprompt/test_signature_opt_bayesian.py b/tests/teleprompt/test_signature_opt_bayesian.pyindex 0cf655784..38f560da4 100644--- a/tests/teleprompt/test_signature_opt_bayesian.py+++ b/tests/teleprompt/test_signature_opt_bayesian.py@@ -3,7 +3,7 @@ import pytestimport reimport dspyfrom dsp.modules import LM-from dspy.teleprompt.signature_opt_bayesian import BayesianSignatureOptimizer+from dspy.teleprompt.signature_opt_bayesian import MIPROfrom dspy.utils.dummies import DummyLMfrom dspy import Example@@ -41,7 +41,7 @@ class ConditionalLM(LM):def __init__(self):super().__init__("conditional-lm")- def basic_request(self, prompt, n=1, **kwargs):+ def basic_request(self, prompt, num_candidates=1, **kwargs):# If we are in the "optimization" stage, we don't say much.if prompt.endswith("Observations:"):answer = " (*silence*)"@@ -113,8 +113,8 @@ class ConditionalLM(LM):def test_bayesian_signature_optimizer_initialization():- optimizer = BayesianSignatureOptimizer(- metric=simple_metric, n=10, init_temperature=1.4, verbose=True, track_stats=True+ optimizer = MIPRO(+ metric=simple_metric, num_candidates=10, init_temperature=1.4, verbose=True, track_stats=True)assert optimizer.metric == simple_metric, "Metric not correctly initialized"assert optimizer.n == 10, "Incorrect 'n' parameter initialization"@@ -141,9 +141,9 @@ def test_signature_optimizer_optimization_process():student = SimpleModule(signature="input -> output")- optimizer = BayesianSignatureOptimizer(+ optimizer = MIPRO(metric=simple_metric,- n=10,+ num_candidates=10,init_temperature=1.4,verbose=False,track_stats=False,@@ -152,11 +152,12 @@ def test_signature_optimizer_optimization_process():# Adjustments: Include required parameters for the compile methodoptimized_student = optimizer.compile(student=student,- devset=trainset,- optuna_trials_num=10,+ trainset=trainset,+ num_trials=10,max_bootstrapped_demos=3,max_labeled_demos=5,eval_kwargs={"num_threads": 1, "display_progress": False},+ requires_permission_to_run=False,)assert len(optimized_student.predictor.demos) == 5@@ -167,9 +168,9 @@ def test_signature_optimizer_bad_lm():lm=DummyLM([f"Optimized instruction {i}" for i in range(30)]))student = SimpleModule(signature="input -> output")- optimizer = BayesianSignatureOptimizer(+ optimizer = MIPRO(metric=simple_metric,- n=10,+ num_candidates=10,init_temperature=1.4,verbose=False,track_stats=False,@@ -181,11 +182,12 @@ def test_signature_optimizer_bad_lm():with pytest.raises(ValueError):_optimized_student = optimizer.compile(student=student,- devset=trainset,- optuna_trials_num=10,+ trainset=trainset,+ num_trials=10,max_bootstrapped_demos=3,max_labeled_demos=5,eval_kwargs={"num_threads": 1, "display_progress": False},+ requires_permission_to_run=False,)@@ -195,9 +197,9 @@ def test_optimization_and_output_verification():lm = ConditionalLM()dspy.settings.configure(lm=lm)- optimizer = BayesianSignatureOptimizer(+ optimizer = MIPRO(metric=simple_metric,- n=10,+ num_candidates=10,init_temperature=1.4,verbose=False,track_stats=True,@@ -208,11 +210,12 @@ def test_optimization_and_output_verification():# Compile the student with the optimizeroptimized_student = optimizer.compile(student=student,- devset=trainset,- optuna_trials_num=4,+ trainset=trainset,+ num_trials=4,max_bootstrapped_demos=2,max_labeled_demos=3,eval_kwargs={"num_threads": 1, "display_progress": False},+ requires_permission_to_run=False,)# Simulate calling the optimized student with a new input