Download ai/train_sensor_model.py from m4az19/saef-api: direct link, hf CLI and curl.
- Browser
- Download file 2.09 kB
-
https://huggingface.co/spaces/m4az19/saef-api/resolve/main/ai/train_sensor_model.py
- Command line
-
hf download hf://spaces/m4az19/saef-api/ai/train_sensor_model.py
-
curl -L -o train_sensor_model.py https://huggingface.co/spaces/m4az19/saef-api/resolve/main/ai/train_sensor_model.py
2.09 kB
| """Explicit training entry point: --synthetic OR --csv, never auto-trains on startup.""" | |
| import argparse | |
| import csv | |
| import hashlib | |
| import json | |
| from pathlib import Path | |
| from ai.sensor_model import ROOT, synthetic_sessions, train, export, infer | |
| def main(): | |
| p=argparse.ArgumentParser();g=p.add_mutually_exclusive_group(required=True) | |
| g.add_argument('--synthetic',action='store_true');g.add_argument('--csv',type=Path) | |
| p.add_argument('--output',type=Path,default=ROOT/'models/onboard/sensor_model.json') | |
| p.add_argument('--header',type=Path,default=ROOT/'firmware/include/sensor_model_data.h') | |
| args=p.parse_args() | |
| if args.csv: | |
| with args.csv.open(newline='') as f:rows=list(csv.DictReader(f)) | |
| rows.sort(key=lambda r:(r['timestamp'],r['session_id'])) | |
| # Remap sessions in chronological order; lexical user IDs are not time order. | |
| ordered=list(dict.fromkeys(r['session_id'] for r in rows)) | |
| for row in rows:row['session_id']=f'{ordered.index(row["session_id"]):06d}' | |
| else:rows=synthetic_sessions() | |
| model=train(rows,'synthetic_bootstrap' if args.synthetic else 'device_recordings') | |
| model['dataset_sha256']=hashlib.sha256(json.dumps(rows,sort_keys=True).encode()).hexdigest() | |
| export(model,args.output,args.header) | |
| cases=[] | |
| for index in (130,650,1600,2500): | |
| r=rows[index%len(rows)] | |
| sensors={k:float(r[k]) for k in ('battery_voltage_v','solar_voltage_v','solar_current_a','temperature_c','ldr_fraction','battery_current_a')} | |
| sensors.update(accel_g=[0.,0.,float(r['accel_norm_g'])],gyro_dps=[0.,0.,float(r['gyro_norm_dps'])]) | |
| cases.append({'sensors':sensors,'expected':infer(sensors,model)}) | |
| bad={**cases[0]['sensors'],'temperature_c':85.} | |
| cases.append({'sensors':bad,'expected':infer(bad,model)}) | |
| (args.output.parent/'known_vectors.json').write_text(json.dumps(cases,indent=2)+'\n') | |
| print(json.dumps({'model':str(args.output),'header':str(args.header),'bytes':args.output.stat().st_size,'evaluation':model['evaluation']},indent=2)) | |
| if __name__=='__main__':main() | |