Upload finetune_coding.py with huggingface_hub

finetune_coding.py

@@ -8,7 +8,6 @@
 #     "accelerate>=0.24.0",
 #     "bitsandbytes>=0.41.0",
 #     "datasets>=2.0.0",
-#     "trackio",
 #     "jinja2>=3.0",
 # ]
 # ///
@@ -37,7 +36,6 @@ subprocess.run(
 )
 sys.path.insert(0, "/tmp/apertus-format")
 
-import trackio
 from datasets import load_dataset, concatenate_datasets
 from peft import LoraConfig
 from src import (  # noqa: E402 — inserted after sys.path patch
@@ -249,9 +247,9 @@ config = SFTConfig(
     eval_strategy="steps",
     eval_steps=100,
 
-    report_to="trackio",
-    project="apertus-coding-finetune",
-    run_name="apertus-8b-coding-v1",
+    # Trackio's on_push_begin callback crashes with a PyArrow struct bug (empty rank_pattern
+    # field). Disable it so checkpoint pushes to Hub proceed without the crash.
+    report_to="none",
 )
 
 bnb_config = BitsAndBytesConfig(
@@ -284,7 +282,5 @@ trainer.train()
 
 print("Pushing to Hub...")
 trainer.push_to_hub()
-trackio.finish()
 
 print(f"Done! Model at: https://huggingface.co/{OUTPUT_REPO}")
-print(f"Metrics at:    https://huggingface.co/spaces/Colby/trackio")