Make transformers dependency explicit

README.md

@@ -64,6 +64,10 @@ You can follow [@NormalComputing](https://twitter.com/NormalComputing), [@remilo
 pip install outlines
 ```
 
+The dependencies needed to use models are not installed by default. You will need to run:
+
+- `pip install openai` to be able to use OpenAI [models](https://platform.openai.com/docs/api-reference).
+- `pip install transformers` to be able to use HuggingFace `transformers` [models](https://huggingface.co/models?pipeline_tag=text-generation).
 
 ## Guided generation
 
@@ -94,7 +98,11 @@ import outlines.models as models
 
 model = models.transformers("gpt2")
 
-prompt = labelling("Just awesome", examples)
+prompt = """You are a sentiment-labelling assistant.
+Is the following review positive or negative?
+
+Review: This restaurant is just awesome!
+"""
 answer = generate.choice(model, ["Positive", "Negative"])(prompt)
 ```
 

outlines/models/transformers.py

@@ -87,7 +87,12 @@ def convert_token_to_string(self, token: str) -> str:
 
 
 def transformers(model_name: str, device: Optional[str] = None, **model_kwargs):
-    from transformers import AutoModelForCausalLM
+    try:
+        from transformers import AutoModelForCausalLM
+    except ImportError:
+        raise ImportError(
+            "The `transformers` library needs to be installed in order to use `transformers` models."
+        )
 
     model = AutoModelForCausalLM.from_pretrained(model_name, **model_kwargs)
     tokenizer = TransformersTokenizer(model_name)

pyproject.toml

@@ -6,7 +6,7 @@ build-backend = "setuptools.build_meta"
 name = "outlines"
 authors= [{name = "Normal Computing", email = "[email protected]"}]
 description = "Probabilistic Generative Model Programming"
-requires-python = ">=3.7"
+requires-python = ">=3.10"
 keywords=[
     "normal computing",
     "machine learning",