Skip to content

Commit eaefa0d

Browse files
AlexejPennerAlexejPenner
committed
Refactored configs
1 parent f82e6a1 commit eaefa0d

File tree

15 files changed

+361
-141
lines changed

15 files changed

+361
-141
lines changed

‎.github/workflows/production_run_complete_llm.yml‎

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -56,4 +56,4 @@ jobs:
5656
- name: Run pipeline, create pipeline, configure trigger (Production)
5757
working-directory: ./llm-complete-guide
5858
run: |
59-
python gh_action_rag.py --no-cache --create-template ----event-source-id --service-account-id ${{ env.ZENML_SERVICE_ACCOUNT_ID }} --action-id ${{ env.ZENML_ACTION_ID }} --config rag_gcp.yaml
59+
python gh_action_rag.py --no-cache --create-template ----event-source-id --service-account-id ${{ env.ZENML_SERVICE_ACCOUNT_ID }} --action-id ${{ env.ZENML_ACTION_ID }} --config production/rag.yaml

‎.github/workflows/staging_run_complete_llm.yml‎

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -52,4 +52,4 @@ jobs:
5252
- name: Run pipeline (Staging)
5353
working-directory: ./llm-complete-guide
5454
run: |
55-
python gh_action_rag.py --no-cache --config rag_local_dev.yaml
55+
python gh_action_rag.py --no-cache --config staging/rag.yaml

llm-complete-guide/configs/embeddings.yaml renamed to llm-complete-guide/configs/dev/embeddings.yaml

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -33,7 +33,7 @@ settings:
3333

3434
# configuration of the Model Control Plane
3535
model:
36-
name: finetuned-zenml-docs-embeddings
36+
name: dev_finetuned-zenml-docs-embeddings
3737
version: latest
3838
license: Apache 2.0
3939
description: Finetuned LLM on ZenML docs

llm-complete-guide/configs/rag_local_dev.yaml renamed to llm-complete-guide/configs/dev/rag.yaml

Lines changed: 2 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -23,12 +23,12 @@ settings:
2323

2424
# configuration of the Model Control Plane
2525
model:
26-
name: finetuned-zenml-docs-embeddings
26+
name: dev_finetuned-zenml-docs-embeddings
2727
license: Apache 2.0
2828
description: Finetuned LLM on ZenML docs
2929
tags: ["rag", "finetuned"]
3030

3131
steps:
3232
url_scraper:
3333
parameters:
34-
docs_url: https://docs.zenml.io
34+
docs_url: https://docs.zenml.io/stack-components/orchestrators

llm-complete-guide/configs/rag_eval.yaml renamed to llm-complete-guide/configs/dev/rag_eval.yaml

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -16,7 +16,7 @@ settings:
1616

1717
# configuration of the Model Control Plane
1818
model:
19-
name: finetuned-zenml-docs-embeddings
19+
name: dev_finetuned-zenml-docs-embeddings
2020
license: Apache 2.0
2121
description: Finetuned LLM on ZenML docs
2222
tags: ["rag", "finetuned"]

llm-complete-guide/configs/synthetic.yaml renamed to llm-complete-guide/configs/dev/synthetic.yaml

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -31,7 +31,7 @@ settings:
3131

3232
# configuration of the Model Control Plane
3333
model:
34-
name: finetuned-zenml-docs-embeddings
34+
name: dev_finetuned-zenml-docs-embeddings
3535
version: latest
3636
license: Apache 2.0
3737
description: Finetuned LLM on ZenML docs

‎llm-complete-guide/configs/production/embeddings.yaml‎

Lines changed: 48 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,48 @@
1+
# enable_cache: False
2+
3+
# environment configuration
4+
settings:
5+
docker:
6+
parent_image: "zenmldocker/prepare-release:base-0.68.1"
7+
requirements:
8+
- langchain-community
9+
- ratelimit
10+
- langchain>=0.0.325
11+
- langchain-openai
12+
- pgvector
13+
- psycopg2-binary
14+
- beautifulsoup4
15+
- unstructured
16+
- pandas
17+
- numpy
18+
- sentence-transformers>=3
19+
- transformers[torch]==4.43.1
20+
- litellm
21+
- ollama
22+
- tiktoken
23+
- umap-learn
24+
- matplotlib
25+
- pyarrow
26+
- rerankers[flashrank]
27+
- datasets
28+
- torch
29+
- pygithub
30+
environment:
31+
ZENML_PROJECT_SECRET_NAME: llm_complete
32+
33+
34+
# configuration of the Model Control Plane
35+
model:
36+
name: prod_finetuned-zenml-docs-embeddings
37+
version: latest
38+
license: Apache 2.0
39+
description: Finetuned LLM on ZenML docs
40+
tags: ["rag", "finetuned"]
41+
42+
steps:
43+
finetune:
44+
step_operator: "gcp_a100"
45+
settings:
46+
step_operator.vertex:
47+
accelerator_count: 1
48+
accelerator_type: NVIDIA_TESLA_A100

llm-complete-guide/configs/rag_gcp.yaml renamed to llm-complete-guide/configs/production/eval.yaml

Lines changed: 9 additions & 18 deletions
Original file line numberDiff line numberDiff line change
@@ -1,3 +1,5 @@
1+
enable_cache: False
2+
13
# environment configuration
24
settings:
35
docker:
@@ -11,31 +13,20 @@ settings:
1113
- psycopg2-binary
1214
- tiktoken
1315
- ratelimit
14-
- rerankers
16+
- rerankers[flashrank]
17+
- matplotlib
18+
- pillow
1519
- pygithub
1620
environment:
1721
ZENML_PROJECT_SECRET_NAME: llm_complete
1822
ZENML_ENABLE_RICH_TRACEBACK: FALSE
1923
ZENML_LOGGING_VERBOSITY: INFO
2024

21-
steps:
22-
url_scraper:
23-
parameters:
24-
docs_url: https://docs.zenml.io
25-
repo_url: https://github.com/zenml-io/zenml
26-
website_url: https://zenml.io
27-
28-
# generate_embeddings:
29-
# step_operator: "terraform-gcp-6c0fd52233ca"
30-
# settings:
31-
# step_operator.vertex:
32-
# accelerator_type: "NVIDIA_TESLA_P100"
33-
# accelerator_count: 1
34-
# machine_type: "n1-standard-8"
35-
3625
# configuration of the Model Control Plane
3726
model:
38-
name: finetuned-zenml-docs-embeddings
27+
name: prod_finetuned-zenml-docs-embeddings
28+
version: latest
3929
license: Apache 2.0
4030
description: Finetuned LLM on ZenML docs
41-
tags: ["rag", "finetuned"]
31+
tags: ["rag", "finetuned"]
32+
limitations: "Only works for ZenML documentation. Not generalizable to other domains. Entirely build with synthetic data. The data is also quite noisy on account of how the chunks were split."

‎llm-complete-guide/configs/production/rag.yaml‎

Lines changed: 44 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,44 @@
1+
enable_cache: True
2+
3+
# environment configuration
4+
settings:
5+
docker:
6+
requirements:
7+
- unstructured
8+
- sentence-transformers>=3
9+
- pgvector
10+
- datasets
11+
- litellm
12+
- numpy
13+
- psycopg2-binary
14+
- tiktoken
15+
- ratelimit
16+
- rerankers
17+
- pygithub
18+
environment:
19+
ZENML_PROJECT_SECRET_NAME: llm_complete
20+
ZENML_ENABLE_RICH_TRACEBACK: FALSE
21+
ZENML_LOGGING_VERBOSITY: INFO
22+
23+
24+
# configuration of the Model Control Plane
25+
model:
26+
name: prod_finetuned-zenml-docs-embeddings
27+
license: Apache 2.0
28+
description: A fine-tuned embeddings model for ZenML documentation. Used for RAG retrieval.
29+
tags: ["rag", "finetuned"]
30+
limitations: Only works for ZenML documentation. Not generalizable to other domains. Entirely build with synthetic data. The data is also quite noisy on account of how the chunks were split.
31+
trade_offs: Focused on a specific RAG retrieval use case. Not generalizable to other domains.
32+
audience: ZenML users
33+
use_cases: RAG retrieval
34+
35+
steps:
36+
url_scraper:
37+
parameters:
38+
docs_url: https://docs.zenml.io
39+
generate_embeddings:
40+
step_operator: "gcp_a100"
41+
settings:
42+
step_operator.vertex:
43+
accelerator_count: 1
44+
accelerator_type: NVIDIA_TESLA_A100

‎llm-complete-guide/configs/production/synthetic.yaml‎

Lines changed: 39 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,39 @@
1+
# environment configuration
2+
settings:
3+
docker:
4+
requirements:
5+
- langchain-community
6+
- ratelimit
7+
- langchain>=0.0.325
8+
- langchain-openai
9+
- pgvector
10+
- psycopg2-binary
11+
- beautifulsoup4
12+
- unstructured
13+
- pandas
14+
- numpy
15+
- sentence-transformers>=3
16+
- transformers==4.43.1
17+
- litellm
18+
- ollama
19+
- tiktoken
20+
- umap-learn
21+
- matplotlib
22+
- pyarrow
23+
- rerankers[flashrank]
24+
- datasets
25+
- torch
26+
- distilabel
27+
- argilla
28+
- pygithub
29+
environment:
30+
ZENML_PROJECT_SECRET_NAME: llm_complete
31+
32+
33+
# configuration of the Model Control Plane
34+
model:
35+
name: prod_finetuned-zenml-docs-embeddings
36+
version: latest
37+
license: Apache 2.0
38+
description: Finetuned LLM on ZenML docs
39+
tags: ["rag", "finetuned"]

0 commit comments

Comments
 (0)