Complete the LLM deployment (#139)

* make deployed version work

* set 4o as the new default

* formatting

llm-complete-guide/constants.py

@@ -27,7 +27,7 @@
 RATE_LIMIT = 5  # Maximum number of requests per second
 
 # LLM Utils constants
-OPENAI_MODEL = "gpt-3.5-turbo"
+OPENAI_MODEL = "gpt-4o"
 EMBEDDINGS_MODEL = "sentence-transformers/all-MiniLM-L12-v2"
 MODEL_NAME_MAP = {
     "gpt4": "gpt-4",

llm-complete-guide/most_basic_eval.py

@@ -14,12 +14,10 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-import os
 import re
 import string
 
 from openai import OpenAI
-
 from utils.openai_utils import get_openai_api_key
 
 

llm-complete-guide/most_basic_rag_pipeline.py

@@ -15,12 +15,10 @@
 # limitations under the License.
 #
 
-import os
 import re
 import string
 
 from openai import OpenAI
-
 from utils.openai_utils import get_openai_api_key
 
 

llm-complete-guide/notebooks/argilla_embeddings.ipynb

@@ -68,10 +68,10 @@
    "source": [
     "import os\n",
     "\n",
-    "from distilabel.steps.tasks import GenerateSentencePair\n",
-    "from distilabel.llms import OpenAILLM, OllamaLLM\n",
-    "from distilabel.steps import LoadDataFromHub\n",
+    "from distilabel.llms import OpenAILLM\n",
     "from distilabel.pipeline import Pipeline\n",
+    "from distilabel.steps import LoadDataFromHub\n",
+    "from distilabel.steps.tasks import GenerateSentencePair\n",
     "\n",
     "# TODO: I think we might optimize this a bit more.\n",
     "\n",
@@ -254,8 +254,8 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "from sentence_transformers import SentenceTransformer\n",
     "import torch\n",
+    "from sentence_transformers import SentenceTransformer\n",
     "\n",
     "model_id = \"Snowflake/snowflake-arctic-embed-m\"  # Hugging Face model ID\n",
     "\n",
@@ -508,8 +508,8 @@
    "outputs": [],
    "source": [
     "import argilla as rg\n",
-    "from zenml.client import Client\n",
     "from argilla._exceptions import ConflictError\n",
+    "from zenml.client import Client\n",
     "\n",
     "zenml_client = Client()\n",
     "api_key = zenml_client.get_secret(\"argilla_hf\").secret_values[\"api_key\"]\n",
@@ -654,13 +654,13 @@
    "outputs": [],
    "source": [
     "import torch\n",
+    "from datasets import concatenate_datasets, load_dataset\n",
     "from sentence_transformers import SentenceTransformer\n",
     "from sentence_transformers.evaluation import (\n",
     "    InformationRetrievalEvaluator,\n",
     "    SequentialEvaluator,\n",
     ")\n",
     "from sentence_transformers.util import cos_sim\n",
-    "from datasets import load_dataset, concatenate_datasets\n",
     "\n",
     "model_id = \"Snowflake/snowflake-arctic-embed-m\"  # Hugging Face model ID\n",
     "matryoshka_dimensions = [384, 256, 128, 64]  # Important: large to small\n",
@@ -752,8 +752,8 @@
    "outputs": [],
    "source": [
     "from sentence_transformers import (\n",
-    "    SentenceTransformerModelCardData,\n",
     "    SentenceTransformer,\n",
+    "    SentenceTransformerModelCardData,\n",
     ")\n",
     "\n",
     "model_id = \"Snowflake/snowflake-arctic-embed-m\"\n",

llm-complete-guide/notebooks/finetune_embeddings.ipynb

@@ -64,7 +64,6 @@
    "outputs": [],
    "source": [
     "from datasets import Dataset, DatasetDict\n",
-    "from huggingface_hub import create_repo\n",
     "from zenml.client import Client\n",
     "\n",
     "# Assuming you have a Dataset object named 'dataset'\n",
@@ -174,9 +173,8 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "from torch.utils.data import DataLoader\n",
     "from sentence_transformers import losses\n",
-    "\n",
+    "from torch.utils.data import DataLoader\n",
     "\n",
     "train_dataloaderB = DataLoader(train_examplesB, shuffle=True, batch_size=64)\n",
     "train_lossB = losses.MultipleNegativesRankingLoss(model=modelB)\n",
@@ -218,8 +216,8 @@
     }
    ],
    "source": [
-    "from torch.utils.data import Dataset, DataLoader\n",
     "from sentence_transformers import InputExample\n",
+    "from torch.utils.data import DataLoader, Dataset\n",
     "\n",
     "\n",
     "class InputExampleDataset(Dataset):\n",

llm-complete-guide/notebooks/visualise_embeddings.ipynb

@@ -16,8 +16,8 @@
    ],
    "source": [
     "import matplotlib.pyplot as plt\n",
-    "from sklearn.manifold import TSNE\n",
     "import umap\n",
+    "from sklearn.manifold import TSNE\n",
     "from zenml.client import Client\n",
     "\n",
     "artifact = Client().get_artifact_version(\n",
@@ -32,8 +32,8 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "import numpy as np\n",
     "import matplotlib.pyplot as plt\n",
+    "import numpy as np\n",
     "from matplotlib.colors import ListedColormap\n",
     "\n",
     "embeddings = np.array([doc.embedding for doc in documents])\n",

llm-complete-guide/pipelines/llm_basic_rag.py

@@ -22,7 +22,7 @@
 )
 from steps.url_scraper import url_scraper
 from steps.web_url_loader import web_url_loader
-from zenml import pipeline, Model
+from zenml import pipeline
 
 
 @pipeline

llm-complete-guide/pipelines/rag_deployment.py

@@ -1,5 +1,5 @@
-from zenml import pipeline
 from steps.rag_deployment import gradio_rag_deployment
+from zenml import pipeline
 
 
 @pipeline(enable_cache=False)

llm-complete-guide/steps/distilabel_generate_queries.py

@@ -12,7 +12,6 @@
 #  or implied. See the License for the specific language governing
 #  permissions and limitations under the License.
 
-import os
 from typing import Annotated, Tuple
 
 from constants import (
@@ -22,12 +21,11 @@
 )
 from datasets import Dataset
 from distilabel.llms import OpenAILLM
+from distilabel.pipeline import Pipeline
 from distilabel.steps import LoadDataFromHub
 from distilabel.steps.tasks import GenerateSentencePair
-from distilabel.pipeline import Pipeline
-from zenml import step
-
 from utils.openai_utils import get_openai_api_key
+from zenml import step
 
 synthetic_generation_context = """
 The text is a chunk from technical documentation of ZenML.

llm-complete-guide/steps/eval_e2e.py

@@ -23,9 +23,8 @@
 from pydantic import BaseModel, conint
 from structures import TestResult
 from utils.llm_utils import process_input_with_retrieval
-from zenml import step
-
 from utils.openai_utils import get_openai_api_key
+from zenml import step
 
 logging.getLogger().setLevel(logging.WARNING)
 

llm-complete-guide/steps/generate_questions.py

@@ -20,11 +20,10 @@
 from litellm.exceptions import APIConnectionError, Timeout
 from rich import print
 from structures import Document
+from utils.openai_utils import get_openai_api_key
 from zenml import log_artifact_metadata, step
 from zenml.logger import get_logger
 
-from utils.openai_utils import get_openai_api_key
-
 logger = get_logger(__name__)
 LOCAL_MODEL = "ollama/mixtral"
 

llm-complete-guide/steps/huggingface_dataset_upload.py

@@ -13,13 +13,12 @@
 #  permissions and limitations under the License.
 
 import polars as pl
+from constants import SECRET_NAME
 from datasets import Dataset
 from huggingface_hub import create_repo
 from zenml import step
 from zenml.client import Client
 
-from constants import SECRET_NAME
-
 
 @step
 def upload_chunks_dataset_to_huggingface(

llm-complete-guide/steps/push_to_hf.py

@@ -11,11 +11,10 @@
 #  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express
 #  or implied. See the License for the specific language governing
 #  permissions and limitations under the License.
-from zenml.client import Client
-
 from constants import DATASET_NAME_DISTILABEL, SECRET_NAME
 from datasets import Dataset, DatasetDict
 from zenml import step
+from zenml.client import Client
 
 
 @step

llm-complete-guide/steps/rag_deployment.py

@@ -1,16 +1,27 @@
-from zenml import step
 import time
-import gradio as gr
 
+import gradio as gr
+from utils.llm_utils import process_input_with_retrieval
+from zenml import step
 
-def slow_echo(message, history):
-    for i in range(len(message)):
-        time.sleep(0.05)
-        yield "You typed: " + message[: i + 1]
 
+def predict(message, history):
+    return process_input_with_retrieval(
+        input=message,
+        n_items_retrieved=20,
+        use_reranking=True,
+    )
 
 
 @step
-def gradio_rag_deployment():
-    demo = gr.ChatInterface(slow_echo, type="messages")
-    demo.launch()
+def gradio_rag_deployment() -> None:
+    """Launches a Gradio chat interface with the slow echo demo.
+
+    Starts a web server with a chat interface that echoes back user messages.
+    The server runs indefinitely until manually stopped.
+    """
+    demo = gr.ChatInterface(predict, type="messages")
+    demo.launch(share=True, inbrowser=True)
+    # Keep the step running
+    while True:
+        time.sleep(1)

llm-complete-guide/steps/synthetic_data.py

@@ -19,17 +19,15 @@
 from typing import Annotated
 
 import pandas as pd
+from constants import SECRET_NAME
 from datasets import Dataset
 from huggingface_hub import create_repo
 from litellm import completion
-
-from constants import SECRET_NAME
 from structures import Document
+from utils.openai_utils import get_openai_api_key
 from zenml import ArtifactConfig, step
 from zenml.client import Client
 
-from utils.openai_utils import get_openai_api_key
-
 logger = logging.getLogger(__name__)
 
 LOCAL_MODEL = "ollama/mixtral"

llm-complete-guide/utils/llm_utils.py

@@ -21,7 +21,6 @@
 
 import logging
 
-from zenml.cli import secret
 from zenml.client import Client
 
 from utils.openai_utils import get_openai_api_key
@@ -35,7 +34,6 @@
 # Configure the logging level for the root logger
 logging.getLogger().setLevel(logging.ERROR)
 
-import os
 import re
 from typing import List, Tuple
 

llm-complete-guide/utils/openai_utils.py

@@ -1,8 +1,5 @@
-import os
-
-from zenml.client import Client
-
 from constants import SECRET_NAME
+from zenml.client import Client
 
 
 def get_openai_api_key():