Synthetic data generation (#9759)

### Description

Implements synthetic data generation with the fields and preferences
given by the user. Adds showcase notebook.
Corresponding prompt was proposed for langchain-hub.

### Example

```
output = chain({"fields": {"colors": ["blue", "yellow"]}, "preferences": {"style": "Make it in a style of a weather forecast."}})
print(output)

# {'fields': {'colors': ['blue', 'yellow']},
 'preferences': {'style': 'Make it in a style of a weather forecast.'},
 'text': "Good morning! Today's weather forecast brings a beautiful combination of colors to the sky, with hues of blue and yellow gently blending together like a mesmerizing painting."}
```

### Twitter handle 

@deepsense_ai @matt_wosinski

---------

Co-authored-by: Bagatur <baskaryan@gmail.com>
This commit is contained in:
Mateusz Wosinski
2023-09-20 01:29:50 +02:00
committed by GitHub
parent c4a6de3fc9
commit a29cd89923
3 changed files with 503 additions and 0 deletions

View File

@@ -0,0 +1,51 @@
from typing import TYPE_CHECKING, Any, Dict, List, Optional
from langchain.chains.llm import LLMChain
from langchain_experimental.synthetic_data.prompts import SENTENCE_PROMPT
if TYPE_CHECKING:
from langchain.chains.base import Chain
from langchain.prompts import PromptTemplate
from langchain.schema.language_model import BaseLanguageModel
def create_data_generation_chain(
llm: BaseLanguageModel,
prompt: Optional[PromptTemplate] = None,
) -> Chain:
"""Creates a chain that generates synthetic sentences with
provided fields.
Args:
llm: The language model to use.
prompt: Prompt to feed the language model with.
If not provided, the default one will be used.
"""
prompt = prompt or SENTENCE_PROMPT
return LLMChain(
llm=llm,
prompt=prompt,
)
class DatasetGenerator:
"""Generates synthetic dataset with a given language model."""
def __init__(
self,
llm: BaseLanguageModel,
sentence_preferences: Optional[Dict[str, Any]] = None,
):
self.generator = create_data_generation_chain(llm)
self.sentence_preferences = sentence_preferences or {}
def __call__(self, fields_collection: List[List[Any]]) -> List[Dict[str, Any]]:
results: List[Dict[str, Any]] = []
for fields in fields_collection:
results.append(
self.generator(
{"fields": fields, "preferences": self.sentence_preferences}
)
)
return results

View File

@@ -0,0 +1,15 @@
from langchain.prompts.prompt import PromptTemplate
sentence_template = """Given the following fields, create a sentence about them.
Make the sentence detailed and interesting. Use every given field.
If any additional preferences are given, use them during sentence construction as well.
Fields:
{fields}
Preferences:
{preferences}
Sentence:
"""
SENTENCE_PROMPT = PromptTemplate(
template=sentence_template, input_variables=["fields", "preferences"]
)