Upload experiments/exp5_upload.py with huggingface_hub

experiments/exp5_upload.py

@@ -0,0 +1,97 @@
+#!/usr/bin/env python3
+"""
+Experiment 5: Upload all artifacts to HuggingFace.
+Uploads: 50K dataset, model weights, training logs, all experiment results.
+"""
+
+import os
+import json
+
+
+def upload_all(results_dir, token):
+    from huggingface_hub import HfApi, upload_file, upload_folder
+
+    api = HfApi()
+    repo_id = "vnmoorthy/pavo-bench"
+
+    parent_dir = os.path.dirname(results_dir)
+    outputs_dir = os.path.join(results_dir, "outputs")
+
+    print(f"  Uploading to {repo_id}...")
+
+    # 1. Upload 50K dataset files (train + test)
+    for fname in ["tier3_50k_train.jsonl", "tier3_50k_test.jsonl", "tier3_50k_summary.json"]:
+        fpath = os.path.join(parent_dir, fname)
+        if os.path.exists(fpath):
+            print(f"    Uploading {fname}...")
+            api.upload_file(
+                path_or_fileobj=fpath,
+                path_in_repo=f"data/{fname}",
+                repo_id=repo_id,
+                repo_type="dataset",
+                token=token,
+            )
+        else:
+            print(f"    WARNING: {fname} not found at {fpath}")
+
+    # 2. Upload original experiment results
+    for fname in os.listdir(parent_dir):
+        if fname.endswith(".json") and not fname.startswith("tier3_50k"):
+            fpath = os.path.join(parent_dir, fname)
+            print(f"    Uploading {fname}...")
+            api.upload_file(
+                path_or_fileobj=fpath,
+                path_in_repo=f"results/{fname}",
+                repo_id=repo_id,
+                repo_type="dataset",
+                token=token,
+            )
+
+    # 3. Upload new experiment outputs
+    if os.path.exists(outputs_dir):
+        for fname in os.listdir(outputs_dir):
+            fpath = os.path.join(outputs_dir, fname)
+            if os.path.isfile(fpath):
+                print(f"    Uploading outputs/{fname}...")
+                api.upload_file(
+                    path_or_fileobj=fpath,
+                    path_in_repo=f"outputs/{fname}",
+                    repo_id=repo_id,
+                    repo_type="dataset",
+                    token=token,
+                )
+
+    # 4. Upload model weights specifically to a models/ directory
+    for model_file in ["meta_controller.pt", "meta_controller_best.pt"]:
+        model_path = os.path.join(outputs_dir, model_file)
+        if os.path.exists(model_path):
+            print(f"    Uploading model: {model_file}...")
+            api.upload_file(
+                path_or_fileobj=model_path,
+                path_in_repo=f"models/{model_file}",
+                repo_id=repo_id,
+                repo_type="dataset",
+                token=token,
+            )
+
+    # 5. Upload training log
+    log_path = os.path.join(outputs_dir, "training_log.json")
+    if os.path.exists(log_path):
+        print(f"    Uploading training_log.json...")
+        api.upload_file(
+            path_or_fileobj=log_path,
+            path_in_repo=f"training/training_log.json",
+            repo_id=repo_id,
+            repo_type="dataset",
+            token=token,
+        )
+
+    print("  Upload complete!")
+    print(f"  View at: https://huggingface.co/datasets/{repo_id}")
+
+
+if __name__ == "__main__":
+    import sys
+    token = sys.argv[1] if len(sys.argv) > 1 else input("HuggingFace token: ").strip()
+    results_dir = os.path.dirname(os.path.abspath(__file__))
+    upload_all(results_dir, token)