Tabular Regression
Transformers
Safetensors
yield-text-image
feature-extraction
crop-yield-estimation
digital-agriculture
multimodal
multi-temporal
regression
custom_code
Instructions to use ICICLE-AI/multimodal-yield-estimation with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use ICICLE-AI/multimodal-yield-estimation with Transformers:
# pip install -U transformers accelerate # Load model directly from transformers import AutoModel model = AutoModel.from_pretrained("ICICLE-AI/multimodal-yield-estimation", trust_remote_code=True, device_map="auto") - Notebooks
- Google Colab
- Kaggle
Download scripts/inference_batch_hf.py from ICICLE-AI/multimodal-yield-estimation: direct link, hf CLI and curl.
- Browser
- Download file 3.5 kB
-
https://huggingface.co/ICICLE-AI/multimodal-yield-estimation/resolve/main/scripts/inference_batch_hf.py
- Command line
-
hf download hf://ICICLE-AI/multimodal-yield-estimation/scripts/inference_batch_hf.py
-
curl -L -o inference_batch_hf.py https://huggingface.co/ICICLE-AI/multimodal-yield-estimation/resolve/main/scripts/inference_batch_hf.py
3.5 kB
| """Batch yield prediction from a JSONL file (one sample per line, same format as the single-sample JSON). | |
| Example: | |
| python scripts/inference_batch_hf.py --hf_model_dir checkpoints \ | |
| --input_jsonl examples/sample_batch.jsonl --output_json predictions.json [--output_csv predictions.csv] | |
| Each line: {"sample_id", "planting_date", "week", "texts": [...], "images": [{"path", "date"}, ...]} | |
| Use --week to predict every sample at the same week. | |
| """ | |
| import sys | |
| from pathlib import Path | |
| PROJECT_ROOT = Path(__file__).resolve().parents[1] | |
| sys.path.insert(0, str(PROJECT_ROOT)) | |
| import argparse | |
| import json | |
| import pandas as pd | |
| import torch | |
| from transformers import AutoModel | |
| from data.embeddings import ImageEmbedder, TextEmbedder, embed_samples | |
| from hf.auto import register_multimodal_autoclass | |
| def parse_args(): | |
| p = argparse.ArgumentParser() | |
| p.add_argument("--hf_model_dir", required=True, help="local checkpoint dir or Hub repo id") | |
| p.add_argument("--input_jsonl", required=True) | |
| p.add_argument("--output_json", default="inference_predictions.json") | |
| p.add_argument("--output_csv", default=None, help="optional: also save as CSV") | |
| p.add_argument("--week", type=int, default=None, help="overrides each sample's week (1..22)") | |
| p.add_argument("--batch_size", type=int, default=64) | |
| return p.parse_args() | |
| def main(): | |
| a = parse_args() | |
| register_multimodal_autoclass() | |
| device = torch.device("cuda" if torch.cuda.is_available() else "cpu") | |
| model = AutoModel.from_pretrained(a.hf_model_dir, trust_remote_code=True).to(device).eval() | |
| with open(a.input_jsonl) as f: | |
| samples = [json.loads(line) for line in f if line.strip()] | |
| print(f"{len(samples)} samples") | |
| x = embed_samples( | |
| samples, | |
| TextEmbedder(model.config.text_model, device), | |
| ImageEmbedder(model.config.image_model, device), | |
| week=a.week, | |
| ) | |
| # Samples can have different weeks: run each week group as its own batch. | |
| preds = [None] * len(samples) | |
| with torch.no_grad(): | |
| for week in sorted(set(x["week"].tolist())): | |
| idx = [i for i, w in enumerate(x["week"]) if w == week] | |
| for s in range(0, len(idx), a.batch_size): | |
| b = idx[s:s + a.batch_size] | |
| out = model( | |
| text=torch.from_numpy(x["text"][b]).to(device), | |
| image=torch.from_numpy(x["image"][b]).to(device), | |
| img_mask=torch.from_numpy(x["img_mask"][b]).to(device), | |
| img_age=torch.from_numpy(x["img_age"][b]).to(device), | |
| cutoff=week, | |
| ) | |
| for i, p in zip(b, out.predictions.cpu().tolist()): | |
| preds[i] = p | |
| results = [] | |
| for i, s in enumerate(samples): | |
| week = int(x["week"][i]) | |
| results.append({ | |
| "sample_id": s.get("sample_id", i), | |
| "week": week, | |
| "has_image_by_week": bool(x["img_mask"][i, week - 1] > 0), | |
| "image_age_weeks": int(x["img_age"][i, week - 1]), | |
| "yield_pred_bu_per_acre": float(preds[i]), | |
| }) | |
| with open(a.output_json, "w") as f: | |
| json.dump(results, f, indent=2) | |
| print(pd.DataFrame(results).head()) | |
| print(f"saved {len(results)} predictions to {a.output_json}") | |
| if a.output_csv: | |
| pd.DataFrame(results).to_csv(a.output_csv, index=False) | |
| print(f"saved {len(results)} predictions to {a.output_csv}") | |
| if __name__ == "__main__": | |
| main() | |