8000 Optimizer: new API by dsblank · Pull Request #2050 · comet-ml/opik · GitHub
[go: up one dir, main page]
More Web Proxy on the site http://driver.im/
Skip to content

Optimizer: new API #2050

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 7 commits into from
May 6, 2025
Merged

Optimizer: new API #2050

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
7 commits
Select commit Hold shift + click to select a range
f731460
Optimizer: new API
May 6, 2025
6ce07fd
Optimizer: MIPRO new API done
May 6, 2025
361c309
Optimizer: Merged API + Optimizer-id
May 6, 2025
3c5fea1
Optimizer: leftover merge conflict fixes
May 6, 2025
44096f5
Optimizer: FewShot new API
May 6, 2025
c7e81a2
Optimizer: metaprompt new API
May 6, 2025
3cc81ce
Optimizer: metaprompt, don't divide by zero
May 6, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
37 changes: 0 additions & 37 deletions sdks/opik_optimizer/benchmarks/dspy-mipro-hotpot-500.py
View file
Open in desktop

This file was deleted.

Oops, something went wrong.
55 changes: 0 additions & 55 deletions sdks/opik_optimizer/benchmarks/opik-mipro-hotpot-500.py
View file
Open in desktop

This file was deleted.

61 changes: 16 additions & 45 deletions sdks/opik_optimizer/scripts/few_shot_bayesian_halueval_chat_prompt_example.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -4,7 +4,6 @@
from opik_optimizer.demo import get_or_create_dataset

from opik_optimizer import (
OptimizationConfig,
MetricConfig,
PromptTaskConfig,
from_dataset_field,
Expand Down Expand Up @@ -52,7 +51,6 @@ def score(
Answer with just one word: 'yes' if there is a hallucination and 'no' if there is not.
"""


optimizer = FewShotBayesianOptimizer(
model="gpt-4o-mini",
project_name=project_name,
Expand All @@ -62,54 +60,27 @@ def score(
seed=42,
)


optimization_config = OptimizationConfig(
dataset=halu_eval_dataset,
objective=MetricConfig(
metric=halu_eval_accuracy,
inputs={
"output": from_llm_response_text(),
"ground_truth": from_dataset_field(name="expected_hallucination_label"),
},
),
task=PromptTaskConfig(
instruction_prompt=prompt_instruction,
input_dataset_fields=["input", "llm_output"],
output_dataset_field="expected_hallucination_label",
use_chat_prompt=True,
),
)

initial_prompt_no_examples = [
{"role": "system", "content": prompt_instruction},
{
"role": "user",
"content": "The user input: {{input}} \n\n The llm output: {{llm_output}}",
metric_config = MetricConfig(
metric=halu_eval_accuracy,
inputs={
"output": from_llm_response_text(),
"ground_truth": from_dataset_field(name="expected_hallucination_label"),
},
]

initial_score = optimizer.evaluate_prompt(
dataset=halu_eval_dataset,
metric_config=optimization_config.objective,
prompt=initial_prompt_no_examples,
num_test=100,
)

print("Initial score:", initial_score)

result = optimizer.optimize_prompt(
optimization_config,
n_trials=5,
num_test=100,
task_config = PromptTaskConfig(
instruction_prompt=prompt_instruction,
input_dataset_fields=["input", "llm_output"],
output_dataset_field="expected_hallucination_label",
use_chat_prompt=True,
)

print("Final prompt:", result.prompt)

final_score = optimizer.evaluate_prompt(
result = optimizer.optimize_prompt(
dataset=halu_eval_dataset,
metric_config=optimization_config.objective,
prompt=result.prompt,
num_test=100,
metric_config=metric_config,
task_config=task_config,
n_trials=5,
n_samples=10,
)

print("Final score:", final_score)
print("Result:", result)
54 changes: 18 additions & 36 deletions sdks/opik_optimizer/scripts/few_shot_bayesian_hotpot_chat_prompt_example.py
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,9 +1,8 @@
from opik.evaluation.metrics import LevenshteinRatio
from opik.evaluation.metrics import Equals
from opik_optimizer.few_shot_bayesian_optimizer import FewShotBayesianOptimizer
from opik_optimizer.demo import get_or_create_dataset

from opik_optimizer import (
OptimizationConfig,
MetricConfig,
PromptTaskConfig,
from_dataset_field,
Expand All @@ -17,10 +16,6 @@
Answer the question.
"""
project_name = "optimize-few-shot-bayesian-hotpot"
initial_prompt_no_examples = [
{"role": "system", "content": prompt_instruction},
{"role": "user", "content": "{{question}}"},
]

optimizer = FewShotBayesianOptimizer(
model="gpt-4o-mini",
Expand All @@ -31,40 +26,27 @@
seed=42,
)

optimization_config = OptimizationConfig(
dataset=hot_pot_dataset,
objective=MetricConfig(
metric=LevenshteinRatio(project_name=project_name),
inputs={
"output": from_llm_response_text(),
"reference": from_dataset_field(name="answer"),
},
),
task=PromptTaskConfig(
instruction_prompt=prompt_instruction,
input_dataset_fields=["question"],
output_dataset_field="answer",
use_chat_prompt=True,
),
metric_config = MetricConfig(
metric=Equals(project_name=project_name),
inputs={
"output": from_llm_response_text(),
"reference": from_dataset_field(name="answer"),
},
)

initial_score = optimizer.evaluate_prompt(
dataset=hot_pot_dataset,
metric_config=optimization_config.objective,
prompt=initial_prompt_no_examples,
task_config = PromptTaskConfig(
instruction_prompt=prompt_instruction,
input_dataset_fields=["question"],
output_dataset_field="answer",
use_chat_prompt=True,
)

print("Initial prompt:", initial_prompt_no_examples)
print("Initial score:", initial_score)

result = optimizer.optimize_prompt(optimization_config, n_trials=10)

print("Final prompt:", result.prompt)

final_score = optimizer.evaluate_prompt(
result = optimizer.optimize_prompt(
dataset=hot_pot_dataset,
metric_config=optimization_config.objective,
prompt=result.prompt,
metric_config=metric_config,
task_config=task_config,
n_trials=10,
n_samples=100,
)

print("Final score:", final_score)
print("Result:", result)
56 changes: 17 additions & 39 deletions sdks/opik_optimizer/scripts/meta_prompt_hotpot_example.py
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,9 +1,8 @@
from opik_optimizer import MetaPromptOptimizer
from opik.evaluation.metrics import LevenshteinRatio
from opik.evaluation.metrics import Equals
from opik_optimizer.demo import get_or_create_dataset

from opik_optimizer import (
OptimizationConfig,
MetricConfig,
PromptTaskConfig,
from_dataset_field,
Expand All @@ -30,49 +29,28 @@
)

# Create the optimization configuration
optimization_config = OptimizationConfig(
dataset=hotpot_dataset,
objective=MetricConfig(
metric=LevenshteinRatio(project_name=project_name),
inputs={
"output": from_llm_response_text(),
"reference": from_dataset_field(name="answer"),
},
),
task=PromptTaskConfig(
instruction_prompt=initial_prompt,
input_dataset_fields=["question"],
output_dataset_field="answer",
),
)

# Evaluate the initial prompt
initial_score = optimizer.evaluate_prompt(
dataset=hotpot_dataset,
metric_config=optimization_config.objective,
task_config=optimization_config.task,
prompt=initial_prompt,
num_test=100,
metric_config = MetricConfig(
metric=Equals(project_name=project_name),
inputs={
"output": from_llm_response_text(),
"reference": from_dataset_field(name="answer"),
},
)

print("Initial prompt:", initial_prompt)
print("Initial score:", initial_score)
task_config = PromptTaskConfig(
instruction_prompt=initial_prompt,
input_dataset_fields=["question"],
output_dataset_field="answer",
)

# Optimize the prompt using the optimization config
result = optimizer.optimize_prompt(
config=optimization_config,
num_test=100,
)

print(result)

# Evaluate the final optimized prompt
final_score = optimizer.evaluate_prompt(
dataset=hotpot_dataset,
metric_config=optimization_config.objective,
task_config=optimization_config.task,
prompt=result.prompt,
num_test=100,
metric_config=metric_config,
task_config=task_config,
auto_continue=False,
n_samples=10,
)

print("Final score:", final_score)
print(result)
Loading
0