junte/ML
1
0
Fork 0
mirror of https://github.com/Hestia-Homes/ML.git synced 2026-06-08 11:17:25 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

medium model with scenario and upgraded autogluon

Browse source
This commit is contained in:
Michael Duong 2024-03-26 22:30:50 +00:00
parent ad2c4d6019
commit 8a9b5877b5
10 changed files with 250 additions and 46 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

4
.github/workflows/MLPipelinePullRequest.yml vendored
View file

@ -98,6 +98,10 @@ jobs:
git fetch --depth=1 origin ${TARGET_BRANCH}:${TARGET_BRANCH}
dvc metrics diff --md --all ${TARGET_BRANCH} >> report.md
echo "## Scenario metrics" >> report.md
cat metrics/scenario_table.md >> report.md
cml comment create report.md
# echo "## Residuals plot from model" >> report.md

125
modules/ml-pipeline/src/pipeline/5_generate_scenarios.py Normal file
View file

@ -0,0 +1,125 @@
"""
Fourth part of the pipeline:
After the model is built and metrics are generated,
we want to test this model against known scenarios
"""
import os
import pandas as pd
from core.interface.InterfaceModels import MLModel
from core.interface.InterfaceDataClient import DataClient
from configs.post_prediction_logic import post_prediction_logic
from core.DataClient import dataclient_factory
from core.MLModels import model_factory
from core.Logger import logger
from config import settings
logger.info(f"--- Initiate Parameters ---")
RUNTIME_ENVIRONMENT = os.environ.get("RUNTIME_ENVIRONMENT", "local")
client_params = settings.client
prepare_data_params = settings.prepare_data
build_model_params = settings.build_model
generate_predictions_params = settings.generate_predictions
generate_metrics_params = settings.generate_metrics
feature_process_params = settings.feature_processor
scenarios_params = settings.scenarios
model_filepath = build_model_params["model_save_filepath"]
target = feature_process_params["feature_processor_config"]["target"]
scenario_data_filepaths = scenarios_params["scenario_data_filepaths"]
predictions_column_name = generate_predictions_params["predictions_column_name"]
output_filepath = scenarios_params["output_filepath"]
logger.info(f"--- Initiate MLModel ---")
model = model_factory(build_model_params["model_type"])
logger.info(f"--- Initiate DataClient ---")
# Use data client for input and output, as we use dvc to cache later to the cloud
input_dataclient_type = scenarios_params["input_dataclient_type"]
input_dataclient = dataclient_factory(
dataclient_type=input_dataclient_type,
dataclient_config=client_params[input_dataclient_type],
)
output_dataclient_type = scenarios_params["output_dataclient_type"]
output_dataclient = dataclient_factory(
dataclient_type=output_dataclient_type,
dataclient_config=client_params[output_dataclient_type],
)
def generate_scenario_predictions(
input_dataclient: DataClient,
output_dataclient: DataClient,
model: MLModel,
model_filepath: str,
scenario_data_filepaths: list,
predictions_column_name: str,
output_filepath: str,
):
"""
Given the new model, we generate prediction for expected scenarios
"""
logger.info("--- Loading Scenario Data ---")
scenario_data = pd.DataFrame()
# Can have multiple scenario data files
for scenario_data_filepath in scenario_data_filepaths:
scenario_data = pd.concat(
[
scenario_data,
input_dataclient.load_data(scenario_data_filepath, load_config=None),
]
)
logger.info("--- Loading Model ---")
model.load_model(model_filepath)
logger.info("--- Generating Predictions ---")
predictions = model.predict(
data=scenario_data, post_prediction_logic=post_prediction_logic
)
logger.info("--- Generate Scenario Predicted Impact ---")
predictions_df = pd.DataFrame(predictions)
predictions_df.columns = [predictions_column_name]
scenario_data = pd.concat([scenario_data, predictions_df], axis=1)
scenario_data["predicted_impact"] = abs(
scenario_data[predictions_column_name] - scenario_data["sap_starting"]
)
logger.info("--- Save prediction into metrics ---")
output_df = scenario_data[["uprn", "id", "impact", "predicted_impact"]]
output_dataclient.save_data(
obj=output_df, location=output_filepath, save_config=None
)
if __name__ == "__main__":
logger.info(f"--- {__file__} - Start! ---")
logger.info(f"--- Generate Scenario Predictions ---")
generate_scenario_predictions(
input_dataclient=input_dataclient,
output_dataclient=output_dataclient,
model=model,
model_filepath=model_filepath,
scenario_data_filepaths=scenario_data_filepaths,
predictions_column_name=predictions_column_name,
output_filepath=output_filepath,
)
logger.info(f"--- {__file__} - Complete! ---")

1
modules/ml-pipeline/src/pipeline/config.py
View file

@ -7,6 +7,7 @@ settings = Dynaconf(
"./configs/settings.yaml",
"./configs/build_model.yaml",
"./configs/analysis.yaml",
"./configs/scenarios.yaml",
],
)

4
modules/ml-pipeline/src/pipeline/configs/build_model.yaml
View file

@ -15,8 +15,8 @@ default:
problem_type: regression
eval_metric: mean_squared_error #mean_absolute_error
time_limit: 1800
presets: good_quality
excluded_model_types: ['RF', 'FASTAI', 'CAT', 'NN_TORCH', 'KNN', 'XT']
presets: medium_quality
excluded_model_types: ['RF', 'CAT', 'NN_TORCH', 'KNN', 'XT']
infer_limit: 0.05
infer_limit_batch_size: 10000
ag_args_ensemble: {'num_folds_parallel': 2}

8
modules/ml-pipeline/src/pipeline/configs/scenarios.yaml Normal file
View file

@ -0,0 +1,8 @@
default:
scenarios:
input_dataclient_type: aws-s3
output_dataclient_type: local
scenario_data_filepaths:
# - s3://retrofit-data-dev/scenario_data/22-03-2024-19-20-09/recommendations_scoring_data.parquet
- s3://retrofit-data-dev/scenario_data/24-03-2024-20-23-25/recommendations_scoring_data.parquet
output_filepath: ./metrics/scenario_table.md

37
modules/ml-pipeline/src/pipeline/configs/settings.yaml
View file

@ -18,12 +18,7 @@ default:
prepare_data:
input_dataclient_type: aws-s3
output_dataclient_type: local
# data_filepath: s3://retrofit-data-dev/sap_change_model/dataset_with_differencing.parquet
# data_filepath: s3://retrofit-data-dev/sap_change_model/floor_area_clean_test.parquet
# data_filepath: s3://retrofit-data-dev/sap_change_model/dataset_without_differencing.parquet
# data_filepath: s3://retrofit-data-dev/sap_change_model/dataset_test.parquet
data_filepath: s3://retrofit-data-dev/sap_change_model/dataset.parquet
# data_filepath: s3://retrofit-datalake-dev/dataset_with0perm_all.parquet
data_filepath: s3://retrofit-data-dev/sap_change_model/2024-03-22-18-56-53/dataset_rooms.parquet
train_proportion: 0.9
output_train_filepath: ./data/prepared_data/train.parquet
output_test_filepath: ./data/prepared_data/test.parquet
@ -35,9 +30,35 @@ default:
subsample_seed: 0
target: sap_ending
identifier_columns: ["uprn"]
drop_columns: ["heat_demand_change", "carbon_change", "rdsap_change", "heat_demand_ending", "carbon_ending", "days_to_starting", "days_to_ending"]
# retain_features: ["SAP_STARTING", "TOTAL_FLOOR_AREA_DIFF"]
# drop_columns: ["heat_demand_change", "carbon_change", "rdsap_change", "heat_demand_ending", "carbon_ending", "days_to_starting", "days_to_ending"]
drop_columns: [
"heat_demand_change", "carbon_change", "rdsap_change", "heat_demand_ending", "carbon_ending", "days_to_starting", "days_to_ending",
'number_habitable_rooms_starting', 'number_habitable_rooms_ending', 'number_heated_rooms_starting', 'number_heated_rooms_ending',
'number_habitable_rooms', 'number_heated_rooms']
retain_features: null
# retain_features: ['uprn', 'sap_starting', 'hot_water_energy_eff_ending',
# 'mainheat_energy_eff_ending', 'constituency', 'roof_energy_eff_ending',
# 'walls_energy_eff_ending', 'secondheat_description_ending',
# 'property_type', 'mainheatc_energy_eff_ending', 'built_form',
# 'walls_insulation_thickness_ending', 'potential_energy_efficiency',
# 'transaction_type_ending',
# 'floor_thermal_transmittance_ending',
# 'low_energy_lighting_ending', 'heat_demand_starting',
# 'photo_supply_ending', 'carbon_starting',
# 'walls_thermal_transmittance_ending',
# 'roof_insulation_thickness_ending',
# 'total_floor_area_ending', 'number_open_fireplaces_ending',
# 'windows_energy_eff_ending',
# 'floor_height_ending',
# 'extension_count_ending',
# 'has_air_source_heat_pump_ending',
# 'charging_system_ending', 'construction_age_band', 'glazed_type_ending',
# 'roof_thermal_transmittance_ending',
# 'floor_insulation_thickness_ending', 'has_mains_gas_ending',
# 'estimated_perimeter_starting', 'energy_consumption_potential',
# 'environment_impact_potential', 'heater_type_ending',
# 'multi_glaze_proportion_ending',
# 'lighting_energy_eff_ending', 'fixed_lighting_outlets_count']
generate_predictions:
input_dataclient_type: local

10
modules/ml-pipeline/src/pipeline/core/DataClient.py
View file

@ -245,7 +245,8 @@ class LocalClient:
save_methods = {
".parquet": self._save_parquet,
".json": self._save_json
".json": self._save_json,
".md": self._save_md,
# "": _save_directory(**save_config),
# ADD MORE save_methods HERE
}
@ -294,3 +295,10 @@ class LocalClient:
# Write the contents of the buffer to the local file
with open(location, "wb") as f:
f.write(buffer.getvalue())
def _save_md(self, obj: pd.DataFrame, location: str, save_config: dict):
"""
Save object as markdown
"""
obj.to_markdown(location, **save_config)

96
modules/ml-pipeline/src/pipeline/dvc.lock
View file

@ -1,5 +1,16 @@
schema: '2.0'
stages:
startup_cleanup:
cmd: python 0_startup_cleanup.py
deps:
- path: 0_startup_cleanup.py
hash: md5
md5: b1b12f6b6393fbf8b83d23684df0a3d4
size: 1220
params:
configs/settings.yaml:
default.startup_cleanup.artefacts: ./data
default.startup_cleanup.metrics: ./metrics
prepare_data:
cmd: python 1_prepare_data.py
deps:
@ -17,12 +28,19 @@ stages:
- carbon_ending
- days_to_starting
- days_to_ending
- number_habitable_rooms_starting
- number_habitable_rooms_ending
- number_heated_rooms_starting
- number_heated_rooms_ending
- number_habitable_rooms
- number_heated_rooms
default.feature_processor.feature_processor_config.retain_features:
default.feature_processor.feature_processor_config.subsample_amount:
default.feature_processor.feature_processor_config.subsample_seed: 0
default.feature_processor.feature_processor_config.target: sap_ending
default.feature_processor.feature_processor_type: dataframe
default.prepare_data.data_filepath: s3://retrofit-data-dev/sap_change_model/dataset.parquet
default.prepare_data.data_filepath:
s3://retrofit-data-dev/sap_change_model/2024-03-22-18-56-53/dataset_rooms.parquet
default.prepare_data.input_dataclient_type: aws-s3
default.prepare_data.output_dataclient_type: local
default.prepare_data.output_test_filepath: ./data/prepared_data/test.parquet
@ -31,8 +49,8 @@ stages:
outs:
- path: data/prepared_data/
hash: md5
md5: 3d1144848fce4ce50f6abfaec5235552.dir
size: 46392840
md5: efa416abea618ae6220a0c3d597603cf.dir
size: 44750997
nfiles: 2
build_model:
cmd: python 2_build_model.py
@ -43,8 +61,8 @@ stages:
size: 4820
- path: data/prepared_data
hash: md5
md5: 3d1144848fce4ce50f6abfaec5235552.dir
size: 46392840
md5: efa416abea618ae6220a0c3d597603cf.dir
size: 44750997
nfiles: 2
params:
configs/build_model.yaml:
@ -62,10 +80,9 @@ stages:
problem_type: regression
eval_metric: mean_squared_error
time_limit: 1800
presets: good_quality
presets: medium_quality
excluded_model_types:
- RF
- FASTAI
- CAT
- NN_TORCH
- KNN
@ -77,18 +94,18 @@ stages:
outs:
- path: data/fit_predictions/
hash: md5
md5: 346b6611afbf2070e038bf945249a86e.dir
size: 3384302
md5: de46250d454c4d713ab580b10ff3fd31.dir
size: 3349318
nfiles: 1
- path: data/model/
hash: md5
md5: 8e37f21728cd092660bafa8c32dc109f.dir
size: 423840922
nfiles: 118
md5: 18bd7a93ece75a65d3a950b7dfdab4fb.dir
size: 735951861
nfiles: 35
- path: metrics/fit_metrics.json
hash: md5
md5: d63e1a8d31503055835ac35149554e41
size: 223
md5: 8a952a5e884c268e6059357a627b9251
size: 224
generate_predictions:
cmd: python 3_generate_predictions.py
deps:
@ -98,13 +115,13 @@ stages:
size: 2464
- path: data/model
hash: md5
md5: 8e37f21728cd092660bafa8c32dc109f.dir
size: 423840922
nfiles: 118
md5: 18bd7a93ece75a65d3a950b7dfdab4fb.dir
size: 735951861
nfiles: 35
- path: data/prepared_data
hash: md5
md5: 3d1144848fce4ce50f6abfaec5235552.dir
size: 46392840
md5: efa416abea618ae6220a0c3d597603cf.dir
size: 44750997
nfiles: 2
params:
configs/settings.yaml:
@ -116,8 +133,8 @@ stages:
outs:
- path: data/predictions/
hash: md5
md5: d148baf508140353d62c16d6ab0fb6b7.dir
size: 469224
md5: 07ef721a0dc94a52e3ba7a70ac45b8ff.dir
size: 463563
nfiles: 1
generate_metrics:
cmd: python 4_generate_metrics.py
@ -128,13 +145,13 @@ stages:
size: 3484
- path: data/predictions
hash: md5
md5: d148baf508140353d62c16d6ab0fb6b7.dir
size: 469224
md5: 07ef721a0dc94a52e3ba7a70ac45b8ff.dir
size: 463563
nfiles: 1
- path: data/prepared_data
hash: md5
md5: 3d1144848fce4ce50f6abfaec5235552.dir
size: 46392840
md5: efa416abea618ae6220a0c3d597603cf.dir
size: 44750997
nfiles: 2
params:
configs/settings.yaml:
@ -144,16 +161,25 @@ stages:
outs:
- path: metrics/metrics.json
hash: md5
md5: 196232f94b563ac525cf65ee5cc6d639
size: 222
startup_cleanup:
cmd: python 0_startup_cleanup.py
md5: 9f863f47799d42c101eba3b03a179455
size: 224
generate_scenerio_metrics:
cmd: python 5_generate_scenarios.py
deps:
- path: 0_startup_cleanup.py
- path: 5_generate_scenarios.py
hash: md5
md5: b1b12f6b6393fbf8b83d23684df0a3d4
size: 1220
md5: 30f80ffeb6ee50c5f7b82943a4dc7702
size: 4014
params:
configs/settings.yaml:
default.startup_cleanup.artefacts: ./data
default.startup_cleanup.metrics: ./metrics
configs/scenarios.yaml:
default.scenarios:
input_dataclient_type: aws-s3
output_dataclient_type: local
scenario_data_filepaths:
- s3://retrofit-data-dev/scenario_data/24-03-2024-20-23-25/recommendations_scoring_data.parquet
output_filepath: ./metrics/scenario_table.md
outs:
- path: metrics/scenario_table.md
hash: md5
md5: 54856c66fca8b2ebd1fa4dea2d25734a
size: 2133

10
modules/ml-pipeline/src/pipeline/dvc.yaml
View file

@ -71,6 +71,16 @@ stages:
outs:
- metrics/metrics.json
always_changed: true
generate_scenerio_metrics:
cmd: python 5_generate_scenarios.py
deps:
- 5_generate_scenarios.py
params:
- configs/scenarios.yaml:
- default.scenarios
outs:
- metrics/scenario_table.md
always_changed: true
metrics:
- metrics/metrics.json
- metrics/fit_metrics.json

1
modules/ml-pipeline/src/pipeline/metrics/.gitignore vendored
View file

@ -1,2 +1,3 @@
/fit_metrics.json
/metrics.json
/scenario_table.md