mlops-prototyping

Name: mlops-prototyping
Author: MLOps-Courses

19stars3forks

How to install mlops-prototyping?

mlops-courses/mlops-coding-skills/mlops-prototyping

$npx -y skills add MLOps-Courses/mlops-coding-skills --skill mlops-prototyping

Installs into the current project.

Summary

Guide to create structured, reproducible Jupyter notebooks for MLOps prototyping, emphasizing configuration management and pipeline integrity.

Files · 1

View on GitHub

SKILL.md

1	# MLOps Prototyping
2
3	## Goal
4
5	To create standardized, reproducible, and production-ready prototypes in Jupyter notebooks. This skill enforces a structured layout (Imports -> Configs -> Load -> EDA -> Modeling -> Eval) and robust engineering practices (Pipelines, Split-Verification) to prevent technical debt and data leakage.
6
7	## Prerequisites
8
9	- Language: Python
10	- Environment: `uv` managed project (`.venv`)
11	- Context: Executed within a `.ipynb` file or converting to one.
12
13	## Instructions
14
15	### 1. Notebook Structure
16
17	Enforce the following linear sections in every notebook to ensure readability and maintainability.
18
19	1. Title & Purpose: H1 Title and a brief description of the experiment goals.
20	2. Imports: Group standard libraries, third-party, and usage-specific imports.
21	3. Configs: Define Global Constants (paths, random seeds, hyperparameters) here. No magic numbers deeper in the code.
22	4. Datasets: Load, validate, and split data.
23	5. Analysis (EDA): Inspect target distributions and correlations.
24	6. Modeling: Define and train `sklearn.pipeline.Pipeline` objects.
25	7. Evaluations: Compute metrics and visualize performance on held-out data.
26
27	### 2. Configuration Standards
28
29	Expose all "knobs" at the top of the notebook for easy experimentation.
30
31	- Randomness: Define `RANDOM_STATE = 42` and use it in splits and model initialization.
32	- Paths: Use `pathlib` for robust path handling.
33
34	```python
35	from pathlib import Path
36	ROOT = Path("..")
37	DATA_PATH = ROOT / "data" / "input.parquet"
38	```
39
40	- Hyperparameters: Group model params (e.g., `N_ESTIMATORS`, `MAX_DEPTH`).
41	- Toggles: Use booleans for expensive operations (e.g., `USE_GPU = True`, `RUN_GRID_SEARCH = False`).
42
43	### 3. Data Management
44
45	Ensure data integrity and prevent leakage.
46
47	- Loading: Prefer `pd.read_parquet` for speed/types, or `pd.read_csv`.
48	- Splitting:
49	- Always split into `X_train`, `X_test`, `y_train`, `y_test` before any data-dependent transformations (imputation, scaling).
50	- Random Split: Use `sklearn.model_selection.train_test_split` with `stratify` for balanced classification.
51	- Time Series: Use `sklearn.model_selection.TimeSeriesSplit` if data has a temporal dimension (do NOT shuffle).
52	- Use `random_state=RANDOM_STATE`.
53
54	### 4. Pipeline Construction
55
56	Prohibit raw data transformations on the full dataset.
57
58	- Mandate: Use `sklearn.pipeline.Pipeline` or `ColumnTransformer`.
59	- Why: Automation of `fit` on train and `transform` on test prevents data leakage.
60	- Example:
61
62	```python
63	from sklearn.pipeline import Pipeline
64	from sklearn.preprocessing import StandardScaler, OneHotEncoder
65	from sklearn.impute import SimpleImputer
66	from sklearn.compose import ColumnTransformer
67
68	CACHE = "./.cache" # Define a cache directory
69
70	numeric_transformer = Pipeline(steps=[
71	('imputer', SimpleImputer(strategy='median')),
72	('scaler', StandardScaler())
73	])
74
75	preprocessor = ColumnTransformer(transformers=[
76	('num', numeric_transformer, numeric_features)
77	])
78
79	# Use 'memory' to cache transformer outputs, speeding up GridSearch
80	model = Pipeline(steps=[
81	('preprocessor', preprocessor),
82	('classifier', RandomForestClassifier())
83	], memory=CACHE)
84	```
85
86	### 5. Evaluation & Visualization
87
88	Go beyond accuracy/MSE.
89
90	- Metrics: Use `sklearn.metrics` appropriate for the task (F1, ROC-AUC, RMSE, MAE).
91	- Baselines: Compare against a "Dummy" model (mean/mode) to verify learning.
92	- Visualization:
93	- Regression: Residual plots, Actual vs Predicted.
94	- Classification: Confusion Matrix, ROC Curve, Precision-Recall.
95	- Feature Importance: Visualize `feature_importances_` or SHAP values.
96
97	### 6. Transition to Production
98
99	Facilitate the move from notebook to python package (`src/`).
100
101	- Function Refactoring: Once a block of code is stable (e.g., a complex data cleaning step), refactor it into a function within the notebook. This makes moving it to a `.py` file trivial later.
102	- Cell Tagging: Use tags like `parameters` (for Papermill) or `export` to mark cells that should be part of the final documentation or automated pipeline.
103	- Clean State: Ensure the notebook runs top-to-bottom (`Restart Kernel and Run All`) without errors before committing.
104
105	## Self-Correction Checklist
106
107	- [ ] No Magic Numbers: Are all parameters in the `Configs` section?
108	- [ ] No Data Leakage: Is `fit` called ONLY on `X_train`?
109	- [ ] Reproducibility: Is `random_state` set for all stochastic operations?
110	- [ ] Resilience: are paths defined relative to the project root?
111	- [ ] Clarity: Does the notebook read like a report (Markdown cells explaining the Why)?

People who liked this also viewed

6 picks

Type

find-skillsHelps users discover and install agent skills when they ask questions like "how do I do X", "find a skill for X", "is there a skill that can...", or express interest in extending capabilities. This skSkillsJul 20262.6M27k
grill-meA relentless interview to sharpen a plan or design.SkillsJul 2026633k182k
vercel-react-best-practicesVercel's official collection of agent skillsSkillsJul 2026573k29k
improve-codebase-architectureScan a codebase for deepening opportunities, present them as a visual HTML report, then grill through whichever one you pick.SkillsJul 2026522k182k
setup-matt-pocock-skillsConfigure this repo for the engineering skills — set up its issue tracker, triage label vocabulary, and domain doc layout. Run once before first use of the other engineering skills.SkillsJul 2026454k182k
triageMove issues and external PRs through a state machine of triage roles — categorise, verify, grill if needed, and write agent-ready briefs.SkillsJul 2026424k182k

mlops-prototyping

19stars3forks

View on GitHub

How to install mlops-prototyping?

mlops-courses/mlops-coding-skills/mlops-prototyping

$npx -y skills add MLOps-Courses/mlops-coding-skills --skill mlops-prototyping

Installs into the current project.

Summary

Guide to create structured, reproducible Jupyter notebooks for MLOps prototyping, emphasizing configuration management and pipeline integrity.

Files · 1

View on GitHub

SKILL.md

1	# MLOps Prototyping
2
3	## Goal
4
5	To create standardized, reproducible, and production-ready prototypes in Jupyter notebooks. This skill enforces a structured layout (Imports -> Configs -> Load -> EDA -> Modeling -> Eval) and robust engineering practices (Pipelines, Split-Verification) to prevent technical debt and data leakage.
6
7	## Prerequisites
8
9	- Language: Python
10	- Environment: `uv` managed project (`.venv`)
11	- Context: Executed within a `.ipynb` file or converting to one.
12
13	## Instructions
14
15	### 1. Notebook Structure
16
17	Enforce the following linear sections in every notebook to ensure readability and maintainability.
18
19	1. Title & Purpose: H1 Title and a brief description of the experiment goals.
20	2. Imports: Group standard libraries, third-party, and usage-specific imports.
21	3. Configs: Define Global Constants (paths, random seeds, hyperparameters) here. No magic numbers deeper in the code.
22	4. Datasets: Load, validate, and split data.
23	5. Analysis (EDA): Inspect target distributions and correlations.
24	6. Modeling: Define and train `sklearn.pipeline.Pipeline` objects.
25	7. Evaluations: Compute metrics and visualize performance on held-out data.
26
27	### 2. Configuration Standards
28
29	Expose all "knobs" at the top of the notebook for easy experimentation.
30
31	- Randomness: Define `RANDOM_STATE = 42` and use it in splits and model initialization.
32	- Paths: Use `pathlib` for robust path handling.
33
34	```python
35	from pathlib import Path
36	ROOT = Path("..")
37	DATA_PATH = ROOT / "data" / "input.parquet"
38	```
39
40	- Hyperparameters: Group model params (e.g., `N_ESTIMATORS`, `MAX_DEPTH`).
41	- Toggles: Use booleans for expensive operations (e.g., `USE_GPU = True`, `RUN_GRID_SEARCH = False`).
42
43	### 3. Data Management
44
45	Ensure data integrity and prevent leakage.
46
47	- Loading: Prefer `pd.read_parquet` for speed/types, or `pd.read_csv`.
48	- Splitting:
49	- Always split into `X_train`, `X_test`, `y_train`, `y_test` before any data-dependent transformations (imputation, scaling).
50	- Random Split: Use `sklearn.model_selection.train_test_split` with `stratify` for balanced classification.
51	- Time Series: Use `sklearn.model_selection.TimeSeriesSplit` if data has a temporal dimension (do NOT shuffle).
52	- Use `random_state=RANDOM_STATE`.
53
54	### 4. Pipeline Construction
55
56	Prohibit raw data transformations on the full dataset.
57
58	- Mandate: Use `sklearn.pipeline.Pipeline` or `ColumnTransformer`.
59	- Why: Automation of `fit` on train and `transform` on test prevents data leakage.
60	- Example:
61
62	```python
63	from sklearn.pipeline import Pipeline
64	from sklearn.preprocessing import StandardScaler, OneHotEncoder
65	from sklearn.impute import SimpleImputer
66	from sklearn.compose import ColumnTransformer
67
68	CACHE = "./.cache" # Define a cache directory
69
70	numeric_transformer = Pipeline(steps=[
71	('imputer', SimpleImputer(strategy='median')),
72	('scaler', StandardScaler())
73	])
74
75	preprocessor = ColumnTransformer(transformers=[
76	('num', numeric_transformer, numeric_features)
77	])
78
79	# Use 'memory' to cache transformer outputs, speeding up GridSearch
80	model = Pipeline(steps=[
81	('preprocessor', preprocessor),
82	('classifier', RandomForestClassifier())
83	], memory=CACHE)
84	```
85
86	### 5. Evaluation & Visualization
87
88	Go beyond accuracy/MSE.
89
90	- Metrics: Use `sklearn.metrics` appropriate for the task (F1, ROC-AUC, RMSE, MAE).
91	- Baselines: Compare against a "Dummy" model (mean/mode) to verify learning.
92	- Visualization:
93	- Regression: Residual plots, Actual vs Predicted.
94	- Classification: Confusion Matrix, ROC Curve, Precision-Recall.
95	- Feature Importance: Visualize `feature_importances_` or SHAP values.
96
97	### 6. Transition to Production
98
99	Facilitate the move from notebook to python package (`src/`).
100
101	- Function Refactoring: Once a block of code is stable (e.g., a complex data cleaning step), refactor it into a function within the notebook. This makes moving it to a `.py` file trivial later.
102	- Cell Tagging: Use tags like `parameters` (for Papermill) or `export` to mark cells that should be part of the final documentation or automated pipeline.
103	- Clean State: Ensure the notebook runs top-to-bottom (`Restart Kernel and Run All`) without errors before committing.
104
105	## Self-Correction Checklist
106
107	- [ ] No Magic Numbers: Are all parameters in the `Configs` section?
108	- [ ] No Data Leakage: Is `fit` called ONLY on `X_train`?
109	- [ ] Reproducibility: Is `random_state` set for all stochastic operations?
110	- [ ] Resilience: are paths defined relative to the project root?
111	- [ ] Clarity: Does the notebook read like a report (Markdown cells explaining the Why)?

People who liked this also viewed

6 picks

Type