Upload source/emergency_upload.py with huggingface_hub

source/emergency_upload.py

@@ -0,0 +1,53 @@
+import subprocess, os, time, glob
+
+REPO = "OpenTransformer/web-crawl-2026"
+files = []
+
+# Get all chunks - upload finished ones first, then active ones (copy-then-upload)
+for pattern in ["/workspace/scraped_data_rust/*.jsonl.gz", "/workspace/scraped_data_go/*.jsonl.gz", "/workspace/scraped_data/*.jsonl.gz"]:
+    files.extend(glob.glob(pattern))
+
+print(f"Found {len(files)} files to upload")
+
+for f in sorted(files, key=os.path.getsize, reverse=True):
+    size_mb = os.path.getsize(f) / 1024 / 1024
+    basename = os.path.basename(f)
+    
+    # For active files, copy first so we upload a consistent snapshot
+    mtime_age = time.time() - os.path.getmtime(f)
+    if mtime_age < 300:  # modified in last 5 min = still active
+        print(f"ACTIVE: {basename} ({size_mb:.0f}MB, {mtime_age:.0f}s ago) - copying snapshot...")
+        snap = f"/tmp/snap_{basename}"
+        os.system(f"cp {f} {snap}")
+        upload_path = snap
+    else:
+        print(f"DONE: {basename} ({size_mb:.0f}MB)")
+        upload_path = f
+    
+    # Determine HF subfolder
+    if "rust" in f:
+        hf_path = f"data/{basename}"
+    elif "go" in f:
+        hf_path = f"data/{basename}"
+    else:
+        hf_path = f"data/{basename}"
+    
+    print(f"  Uploading {hf_path}...")
+    from huggingface_hub import HfApi
+    api = HfApi()
+    try:
+        api.upload_file(
+            path_or_fileobj=upload_path,
+            path_in_repo=hf_path,
+            repo_id=REPO,
+            repo_type="dataset",
+        )
+        print(f"  OK: {hf_path}")
+    except Exception as e:
+        print(f"  FAIL: {e}")
+    
+    # Clean up snapshot
+    if upload_path.startswith("/tmp/snap_"):
+        os.remove(upload_path)
+
+print("ALL DONE")