Merge branch 'feature/llm-complete-updates' of https://github.com/zenml-io/zenml-projects into feature/llm-complete-updates

Author: strickvl

.github/workflows/run_complete_llm.yml

@@ -0,0 +1,49 @@
+name: Staging Trigger LLM-COMPLETE
+on:
+  pull_request:
+    types: [opened, synchronize]
+    branches: [staging, main]
+concurrency:
+  # New commit on branch cancels running workflows of the same branch
+  group: ${{ github.workflow }}-${{ github.ref }}
+  cancel-in-progress: true
+
+jobs:
+  run-staging-workflow:
+    runs-on: ubuntu-dind-runners
+    env:
+      ZENML_HOST: ${{ secrets.ZENML_HOST }}
+      ZENML_API_KEY: ${{ secrets.ZENML_API_KEY }}
+      ZENML_STAGING_STACK: 51a49786-b82a-4646-bde7-a460efb0a9c5
+      ZENML_GITHUB_SHA: ${{ github.event.pull_request.head.sha }}
+      ZENML_GITHUB_URL_PR: ${{ github.event.pull_request._links.html.href }}
+      ZENML_DEBUG: true
+      ZENML_ANALYTICS_OPT_IN: false
+      ZENML_LOGGING_VERBOSITY: INFO
+
+    steps:
+      - name: Check out repository code
+        uses: actions/checkout@v3
+
+      - uses: actions/setup-python@v4
+        with:
+          python-version: '3.11'
+
+      - name: Install requirements
+        run: |
+          pip3 install -r requirements.txt
+          zenml integration install gcp -y
+
+      - name: Connect to ZenML server
+        run: |
+          zenml connect --url $ZENML_HOST --api-key $ZENML_API_KEY
+
+      - name: Set stack (Staging)
+        if: ${{ github.base_ref == 'staging' }}
+        run: |
+          zenml stack set ${{ env.ZENML_STAGING_STACK }}
+
+      - name: Run pipeline (Staging)
+        if: ${{ github.base_ref == 'staging' }}
+        run: |
+          python run.py --rag --evaluation --no-cache

llm-complete-guide/README.md

@@ -43,7 +43,7 @@ environment and install the dependencies using the following command:
 pip install -r requirements.txt
 ```
 
-Depending on your setup you may run into some issues when running the pip install command with the
+Depending on your setup you may run into some issues when running the `pip install` command with the
 `flash_attn` package. In that case running `FLASH_ATTENTION_SKIP_CUDA_BUILD=TRUE pip install flash-attn --no-build-isolation` could help you.
 
 In order to use the default LLM for this query, you'll need an account and an

llm-complete-guide/configs/rag.yaml

@@ -1,6 +1,7 @@
 # environment configuration
 settings:
   docker:
+    parent_image: "zenmldocker/prepare-release:base-0.68.0"
     requirements:
       - unstructured
       - sentence-transformers>=3
@@ -10,3 +11,6 @@ settings:
       - numpy
       - psycopg2-binary
       - tiktoken
+    environment:
+      ZENML_ENABLE_RICH_TRACEBACK: FALSE
+      ZENML_LOGGING_VERBOSITY: INFO

llm-complete-guide/configs/rag_eval.yaml

@@ -3,6 +3,7 @@ enable_cache: False
 # environment configuration
 settings:
   docker:
+    parent_image: "zenmldocker/prepare-release:base-0.68.0"
     requirements:
       - unstructured
       - sentence-transformers>=3

llm-complete-guide/run.py

@@ -111,13 +111,6 @@
     default=False,
     help="Run the synthetic data pipeline.",
 )
-@click.option(
-    "--local",
-    "local",
-    is_flag=True,
-    default=False,
-    help="Uses a local LLM via Ollama.",
-)
 @click.option(
     "--embeddings",
     "embeddings",
@@ -132,13 +125,6 @@
     default=False,
     help="Uses Argilla annotations.",
 )
-@click.option(
-    "--dummyembeddings",
-    "dummyembeddings",
-    is_flag=True,
-    default=False,
-    help="Fine-tunes embeddings.",
-)
 @click.option(
     "--reranked",
     "reranked",
@@ -160,9 +146,7 @@ def main(
     model: str = OPENAI_MODEL,
     no_cache: bool = False,
     synthetic: bool = False,
-    local: bool = False,
     embeddings: bool = False,
-    dummyembeddings: bool = False,
     argilla: bool = False,
     reranked: bool = False,
     chunks: bool = False,
@@ -177,7 +161,6 @@ def main(
         no_cache (bool): If `True`, cache will be disabled.
         synthetic (bool): If `True`, the synthetic data pipeline will be run.
         local (bool): If `True`, the local LLM via Ollama will be used.
-        dummyembeddings (bool): If `True`, dummyembeddings will be used
         embeddings (bool): If `True`, the embeddings will be fine-tuned.
         argilla (bool): If `True`, the Argilla annotations will be used.
         chunks (bool): If `True`, the chunks pipeline will be run.
@@ -225,8 +208,6 @@ def main(
             os.path.dirname(os.path.realpath(__file__)), "configs", "embeddings.yaml"
         )
         finetune_embeddings.with_options(config_path=config_path, **embeddings_finetune_args)()
-    if dummyembeddings:
-        chunking_experiment.with_options(**pipeline_args)()
     if chunks:
         generate_chunk_questions.with_options(**pipeline_args)()