Views
No views yet
config.json — pipeline provenance (l_max, grid, feature/model targets).feature_config.json + feature_state.safetensors — fitted feature pipeline.model_config.json + model.safetensors — torch model weights.mpinv version: 0.1.0l_max: 5scale_factor: 1000000.0mpinv.models.multi_head_mlp.MultiHeadMLPcomposite1from mpinv.pipeline import InversionPipeline
2pipeline = InversionPipeline.from_pretrained('user/repo') # or local path
3# Raw antenna power patterns MUST be scaled by pipeline.meta.scale_factor
4# before being fed to predict. `preprocess` does that explicitly:
5P_scaled = pipeline.preprocess(P_raw)
6out = pipeline.predict(P_scaled) # {'packed': ..., 'P_pred': ...}