wehub-resource-sync
106 lines
3.4 KiB
Python
106 lines
3.4 KiB
Python
from mlflow.exceptions import MlflowException
|
|
from mlflow.genai.optimize.optimize import optimize_prompts
|
|
from mlflow.genai.optimize.optimizers import (
|
|
BasePromptOptimizer,
|
|
GepaPromptOptimizer,
|
|
MetaPromptOptimizer,
|
|
)
|
|
from mlflow.genai.optimize.types import (
|
|
LLMParams,
|
|
OptimizerConfig,
|
|
PromptOptimizationResult,
|
|
PromptOptimizerOutput,
|
|
)
|
|
|
|
_MIGRATION_GUIDE = """
|
|
Migration guide:
|
|
The ``optimize_prompt()`` API has been replaced by
|
|
:py:func:`mlflow.genai.optimize_prompts()`, which provides more flexible
|
|
optimization capabilities with a joint optimization of prompts in an arbitrary function.
|
|
|
|
**Old API (removed):**
|
|
|
|
.. code-block:: python
|
|
|
|
from mlflow.genai import optimize_prompt
|
|
from mlflow.genai.optimize.types import OptimizerConfig, LLMParams
|
|
|
|
result = optimize_prompt(
|
|
target_llm=LLMParams(model_name="openai:/gpt-4o"),
|
|
prompt="prompts:/my-prompt/1",
|
|
train_data=dataset,
|
|
optimizer_config=OptimizerConfig(num_instruction_candidates=10),
|
|
)
|
|
|
|
**New API:**
|
|
|
|
.. code-block:: python
|
|
|
|
import mlflow
|
|
import openai
|
|
from mlflow.genai import optimize_prompts
|
|
from mlflow.genai.optimize.optimizers import GepaPromptOptimizer
|
|
from mlflow.genai.scorers import Correctness
|
|
|
|
|
|
# Define a predict function that uses the prompt and LLM
|
|
def predict_fn(inputs: dict[str, Any]) -> str:
|
|
prompt = mlflow.genai.load_prompt("prompts:/my-prompt/1")
|
|
formatted_prompt = prompt.format(**inputs)
|
|
completion = openai.OpenAI().chat.completions.create(
|
|
model="gpt-4o-mini",
|
|
messages=[{"role": "user", "content": formatted_prompt}],
|
|
)
|
|
return completion.choices[0].message.content
|
|
|
|
|
|
result = optimize_prompts(
|
|
predict_fn=predict_fn,
|
|
train_data=dataset,
|
|
prompt_uris=["prompts:/my-prompt/1"],
|
|
optimizer=GepaPromptOptimizer(
|
|
reflection_model="openai:/gpt-4o",
|
|
max_metric_calls=100,
|
|
),
|
|
scorers=[Correctness(model="openai:/gpt-4o")],
|
|
)
|
|
|
|
Key differences:
|
|
- Use ``optimize_prompts()`` (plural) instead of ``optimize_prompt()``
|
|
- Provide a predict function ``predict_fn`` instead of a prompt uri ``prompt``
|
|
- Use ``prompt_uris`` to reference registered prompts
|
|
- Specify an ``optimizer`` instance (e.g., ``GepaPromptOptimizer``)
|
|
|
|
For more details, see the documentation:
|
|
https://mlflow.org/docs/latest/genai/prompt-registry/optimize-prompts.html
|
|
"""
|
|
|
|
|
|
def optimize_prompt(*args, **kwargs):
|
|
f"""
|
|
Optimize a LLM prompt using the given dataset and evaluation metrics.
|
|
This function has been removed. Use mlflow.genai.optimize_prompts() instead.
|
|
|
|
{_MIGRATION_GUIDE}
|
|
"""
|
|
|
|
raise MlflowException(
|
|
f"""
|
|
The optimize_prompt() function has been removed in MLflow 3.5.0.
|
|
Please use mlflow.genai.optimize_prompts() instead.
|
|
{_MIGRATION_GUIDE}"""
|
|
)
|
|
|
|
|
|
__all__ = [
|
|
"optimize_prompts",
|
|
"optimize_prompt",
|
|
"LLMParams",
|
|
"OptimizerConfig",
|
|
"BasePromptOptimizer",
|
|
"GepaPromptOptimizer",
|
|
"MetaPromptOptimizer",
|
|
"PromptOptimizerOutput",
|
|
"PromptOptimizationResult",
|
|
]
|