Update the examples

examples/babyagi.py

@@ -4,14 +4,16 @@
 
 The original repo can be found at https://github.com/yoheinakajima/babyagi
 """
+
 from collections import deque
 from typing import Deque, List
 
+from openai import OpenAI
+
 import outlines
-import outlines.models as models
 
-model = models.openai("gpt-4o-mini")
-complete = outlines.generate.text(model)
+model = outlines.from_openai(OpenAI(), "gpt-4o-mini")
+complete = outlines.Generator(model)
 
 
 #################

examples/bentoml/service.py

@@ -58,11 +58,7 @@ def __init__(self) -> None:
 
         import outlines
 
-        self.model = outlines.models.transformers(
-            self.bento_model_ref.path,
-            device="cuda",
-            model_kwargs={"torch_dtype": torch.float16},
-        )
+        self.model = outlines.from_transformers()
 
     @bentoml.api
     async def generate(

examples/cerebrium/main.py

@@ -1,6 +1,11 @@
 import outlines
+from transformers import AutoModelForCausalLM, AutoTokenizer
+
+model = outlines.from_transformers(
+    AutoModelForCausalLM.from_pretrained("mistralai/Mistral-7B-Instruct-v0.2"),
+    AutoTokenizer.from_pretrained("mistralai/Mistral-7B-Instruct-v0.2"),
+)
 
-model = outlines.models.transformers("mistralai/Mistral-7B-Instruct-v0.2")
 
 schema = {
     "title": "Character",
@@ -29,14 +34,13 @@
     },
 }
 
-generator = outlines.generate.json(model, schema)
-
 
 def generate(
     prompt: str = "Amiri, a 53 year old warrior woman with a sword and leather armor.",
 ):
-    character = generator(
-        f"<s>[INST]Give me a character description. Describe {prompt}.[/INST]"
+    character = model(
+        f"<s>[INST]Give me a character description. Describe {prompt}.[/INST]",
+        outlines.JsonType(schema),
     )
 
     print(character)

examples/cfg.py

@@ -1,5 +1,7 @@
-import outlines.generate as generate
-import outlines.models as models
+from transformers import AutoModelForCausalLM, AutoTokenizer
+
+import outlines
+from outlines.types import Cfg
 
 nlamb_grammar = r"""
     start: sentence
@@ -75,11 +77,16 @@
     %ignore WS
 """
 
-model = models.transformers("hf-internal-testing/tiny-random-gpt2")
+model_name = "hf-internal-testing/tiny-random-gpt2"
+model = outlines.from_transformers(
+    AutoModelForCausalLM.from_pretrained(model_name),
+    AutoTokenizer.from_pretrained(model_name),
+)
+
 batch_size = 10
 for grammar in [nlamb_grammar, calc_grammar, dyck_grammar, json_grammar]:
-    generator = generate.cfg(model, grammar, max_tokens=model.model.config.n_positions)
-    sequences = generator([" "] * batch_size)
+    generator = outlines.Generator(model, Cfg(grammar))
+    sequences = generator([" "] * batch_size, max_tokens=model.model.config.n_positions)
     for seq in sequences:
         try:
             parse = generator.fsm.parser.parse(seq)

examples/dating_profile.py

@@ -6,7 +6,6 @@
 from pydantic import BaseModel, conlist
 
 import outlines
-from outlines import models
 
 
 class QuestionChoice(str, Enum):
@@ -103,25 +102,27 @@ def dating_profile_prompt(description: str, examples: list[Example]):
 # Below requires ~13GB of GPU memory
 # https://huggingface.co/mosaicml/mpt-7b-8k-instruct
 # Motivation: Reasonably large model that fits on a single GPU and has been fine-tuned for a larger context window
+model_name = "mosaicml/mpt-7b-8k-instruct"
 config = transformers.AutoConfig.from_pretrained(
     "mosaicml/mpt-7b-8k-instruct", trust_remote_code=True
 )
 config.init_device = "meta"
-model = models.transformers(
-    model_name="mosaicml/mpt-7b-8k-instruct",
-    device="cuda",
-    model_kwargs={
-        "config": config,
-        "trust_remote_code": True,
-        "torch_dtype": torch.bfloat16,
-        "device_map": {"": 0},
-    },
+model = outlines.from_transformers(
+    transformers.AutoModelForCausalLM.from_pretrained(
+        model_name,
+        device="cuda",
+        config=config,
+        trust_remote_code=True,
+        torch_dtype=torch.bfloat16,
+        device_map={"": 0},
+    ),
+    transformers.AutoTokenizer.from_pretrained(model_name),
 )
 
 new_description = "I'm a laid-back lawyer who spends a lot of his free-time gaming. I work in a corporate office, but ended up here after the start-up I cofounded got acquired, so still play ping pong with my cool coworkers every day. I have a bar at home where I make cocktails, which is great for entertaining friends. I secretly like to wear suits and get a new one tailored every few months. I also like weddings because I get to wear those suits, and it's a good excuse for a date. I watch the latest series because I'm paying, with my hard-earned money, for every streaming service."
 
 prompt = dating_profile_prompt(description=new_description, examples=samples)
-profile = outlines.generate.json(model, DatingProfile)(prompt)  # type: ignore
+profile = model(prompt, outlines.JsonType(DatingProfile))  # type: ignore
 print(profile)
 
 # Sample generated profiles

examples/llamacpp_example.py

@@ -1,6 +1,7 @@
 from enum import Enum
 
 from pydantic import BaseModel, constr
+from llama_cpp import Llama
 
 import outlines
 
@@ -30,10 +31,10 @@ class Character(BaseModel):
 
 if __name__ == "__main__":
     # curl -L -o mistral-7b-instruct-v0.2.Q5_K_M.gguf https://huggingface.co/TheBloke/Mistral-7B-Instruct-v0.2-GGUF/resolve/main/mistral-7b-instruct-v0.2.Q5_K_M.gguf
-    model = outlines.models.llamacpp("./mistral-7b-instruct-v0.2.Q5_K_M.gguf")
+    model = outlines.from_llamacpp(Llama("./mistral-7b-instruct-v0.2.Q5_K_M.gguf"))
 
     # Construct structured sequence generator
-    generator = outlines.generate.json(model, Character)
+    generator = outlines.Generator(model, outlines.JsonType(Character))
 
     # Draw a sample
     seed = 789005

examples/math_generate_code.py

@@ -1,6 +1,7 @@
 """Example from https://dust.tt/spolu/a/d12ac33169"""
+
 import outlines
-import outlines.models as models
+import openai
 
 examples = [
     {"question": "What is 37593 * 67?", "code": "37593 * 67"},
@@ -35,7 +36,7 @@ def execute_code(code):
 
 
 prompt = answer_with_code_prompt(question, examples)
-model = models.openai("gpt-4o-mini")
-answer = outlines.generate.text(model)(prompt)
+model = outlines.from_openai(openai.OpenAI(), "gpt-4o-mini")
+answer = model(prompt)
 result = execute_code(answer)
 print(f"It takes Carla {result:.0f} minutes to download the file.")

examples/meta_prompting.py

@@ -9,10 +9,14 @@
        https://arxiv.org/abs/2102.07350.
 
 """
+
 import argparse
 
+import openai
 import outlines
-import outlines.models as models
+
+
+client = openai.OpenAI()
 
 
 def split_into_steps(question, model_name: str):
@@ -22,16 +26,15 @@ def solve(question):
         Rephrase : : as a true or false statement, identify an Object, relationship and subject
         """
 
-    model = models.openai(model_name)
-    generator = outlines.generate.text(model)
+    model = outlines.from_openai(client, model_name)
 
     prompt = solve(question)
-    answer = generator(prompt, 500)
+    answer = model(prompt, 500)
     prompt += (
         answer
         + "\n what is the only option that displays the same type of relationship as : :?"
     )
-    answer = generator(prompt, 500)
+    answer = model(prompt, 500)
     completed = prompt + answer
 
     return completed
@@ -49,13 +52,12 @@ def determine_goal(question):
     def solve(memory):
         """{{memory}}. Let's begin."""
 
-    model = models.openai(model_name)
-    generator = outlines.generate.text(model)
+    model = outlines.from_openai(client, model_name)
 
     prompt = determine_goal(question)
-    answer = generator(prompt, stop_at=["."])
+    answer = model(prompt, stop_at=["."])
     prompt = solve(prompt + answer)
-    answer = generator(prompt, max_tokens=500)
+    answer = model(prompt, max_tokens=500)
     completed = prompt + answer
 
     return completed
@@ -89,13 +91,12 @@ def get_answer(question, expert, memory):
         {{question}}
         """
 
-    model = models.openai(model_name)
-    generator = outlines.generate.text(model)
+    model = outlines.from_openai(client, model_name)
 
     prompt = find_expert(question)
-    expert = generator(prompt, stop_at=['"'])
+    expert = model(prompt, stop_at=['"'])
     prompt = get_answer(question, expert, prompt + expert)
-    answer = generator(prompt, max_tokens=500)
+    answer = model(prompt, max_tokens=500)
     completed = prompt + answer
 
     return completed
@@ -117,13 +118,12 @@ def get_answer(expert, memory):
         For instance, {{expert}} would answer
         """
 
-    model = models.openai(model_name)
-    generator = outlines.generate.text(model)
+    model = outlines.from_openai(client, model_name)
 
     prompt = find_expert(question)
-    expert = generator(prompt, stop_at=["\n", "."])
+    expert = model(prompt, stop_at=["\n", "."])
     prompt = get_answer(expert, prompt + expert)
-    answer = generator(prompt, max_tokens=500)
+    answer = model(prompt, max_tokens=500)
     completed = prompt + answer
 
     return completed

examples/modal_example.py

@@ -4,17 +4,19 @@
 
 
 outlines_image = modal.Image.debian_slim(python_version="3.11").pip_install(
-    "outlines==0.0.37",
+    "outlines==1.0.0",
     "transformers==4.38.2",
     "datasets==2.18.0",
     "accelerate==0.27.2",
 )
 
 
 def import_model():
-    import outlines
+    from transformers import AutoModelForCausalLM, AutoTokenizer
 
-    outlines.models.transformers("mistralai/Mistral-7B-Instruct-v0.2")
+    model_id = "mistralai/Mistral-7B-Instruct-v0.2"
+    _ = AutoTokenizer.from_pretrained(model_id)
+    _ = AutoModelForCausalLM.from_pretrained(model_id)
 
 
 outlines_image = outlines_image.run_function(import_model)
@@ -63,12 +65,17 @@ def generate(
     prompt: str = "Amiri, a 53 year old warrior woman with a sword and leather armor.",
 ):
     import outlines
+    from transformers import AutoModelForCausalLM, AutoTokenizer
 
-    model = outlines.models.transformers("mistralai/Mistral-7B-v0.1", device="cuda")
+    model_id = "mistralai/Mistral-7B-Instruct-v0.2"
+    model = outlines.from_transformers(
+        tokenizer=AutoTokenizer.from_pretrained(model_id),
+        model=AutoModelForCausalLM.from_pretrained(model_id, device="cuda"),
+    )
 
-    generator = outlines.generate.json(model, schema)
-    character = generator(
-        f"<s>[INST]Give me a character description. Describe {prompt}.[/INST]"
+    character = model(
+        f"<s>[INST]Give me a character description. Describe {prompt}.[/INST]",
+        outlines.JsonType(schema),
     )
 
     print(character)

examples/parsing.py

@@ -1,4 +1,5 @@
 """An example illustrating parser-based masking."""
+
 import math
 import time
 from copy import copy

examples/pick_odd_one_out.py

@@ -9,8 +9,9 @@
        arXiv preprint arXiv:2212.06094.
 
 """
+
+import openai
 import outlines
-import outlines.models as models
 
 
 @outlines.prompt
@@ -31,7 +32,7 @@ def build_ooo_prompt(options):
 
 options = ["sea", "mountains", "plains", "sock"]
 
-model = models.openai("gpt-4o-mini")
+model = outlines.from_openai(openai.OpenAI(), "gpt-4o-mini")
 gen_text = outlines.generate.text(model)
 gen_choice = outlines.generate.choice(model, options)
 

examples/react.py

@@ -10,11 +10,12 @@
 .. [2] Yao, S., Zhao, J., Yu, D., Du, N., Shafran, I., Narasimhan, K., & Cao, Y. (2022). React: Synergizing reasoning and acting in language models. arXiv preprint arXiv:2210.03629.
 
 """
+
+from openai import OpenAI
 import requests  # type: ignore
 
 import outlines
-import outlines.generate as generate
-import outlines.models as models
+from outlines import Generator, Choice
 
 
 @outlines.prompt
@@ -46,11 +47,11 @@ def search_wikipedia(query: str):
 
 
 prompt = build_reAct_prompt("Where is Apple Computers headquarted? ")
-model = models.openai("gpt-4o-mini")
+model = outlines.from_openai(OpenAI(), "gpt-4o-mini")
 
-mode_generator = generate.choice(model, choices=["Tho", "Act"])
-action_generator = generate.choice(model, choices=["Search", "Finish"])
-text_generator = generate.text(model)
+mode_generator = Generator(model, Choice(["Tho", "Act"]))
+action_generator = Generator(model, Choice(["Search", "Finish"]))
+text_generator = Generator(model)
 
 for i in range(1, 10):
     mode = mode_generator(prompt, max_tokens=128)