Upload folder using huggingface_hub

data/data_collator.py

@@ -712,7 +712,8 @@ class MemecoinCollator:
             'labels': torch.stack([item['labels'] for item in batch]) if batch and 'labels' in batch[0] else None,
             'labels_mask': torch.stack([item['labels_mask'] for item in batch]) if batch and 'labels_mask' in batch[0] else None,
             # Debug info
-            'token_addresses': [item.get('token_address', 'unknown') for item in batch]
+            'token_addresses': [item.get('token_address', 'unknown') for item in batch],
+            't_cutoffs': [item.get('t_cutoff', 'unknown') for item in batch]
         }
 
         # Filter out None values (e.g., if no labels provided)

data/data_loader.py

@@ -1623,6 +1623,16 @@ class OracleDataset(Dataset):
             else:
                 break
         
+        # DEBUG: Log label computation details
+        print(f"  DEBUG LABELS: token={token_address[:12]}...")
+        print(f"    T_cutoff={T_cutoff.isoformat()}, cutoff_ts={cutoff_ts_val}")
+        print(f"    Successful trades count: {len(all_trades)}")
+        print(f"    current_price_idx={current_price_idx}, current_price={current_price}")
+        print(f"    last_trade_ts={last_trade_ts_val}, trades_after_cutoff={len(all_trades) - current_price_idx - 1}")
+        if current_price_idx >= 0 and current_price_idx + 1 < len(all_trades):
+            next_trade = all_trades[current_price_idx + 1]
+            print(f"    Next trade: ts={_timestamp_to_order_value(next_trade['timestamp'])}, price={next_trade.get('price_usd')}")
+        
         label_values = []
         mask_values = []
         
@@ -1661,6 +1671,7 @@ class OracleDataset(Dataset):
 
         return {
             'token_address': token_address,  # For debugging
+            't_cutoff': T_cutoff.isoformat() if T_cutoff else None,  # For debugging
             'event_sequence': event_sequence,
             'wallets': wallet_data, 
             'tokens': all_token_data, 

log.log

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:f1c2198c3ed6e249ddf7b7b017b99b2389e4611b8b0649c63d30c40c59e03ac1
-size 76001
+oid sha256:6c006d7598527d9f388edff81fcd301b87ad3698d315090426dec45751757798
+size 109185

train.py

@@ -156,6 +156,11 @@ def log_debug_batch_context(batch: Dict[str, Any], logger: logging.Logger, step:
         if token_addresses:
             logger.info(f"Token Address: {token_addresses[0]}")
         
+        # Log T_cutoff timestamp
+        t_cutoffs = batch.get('t_cutoffs', [])
+        if t_cutoffs:
+            logger.info(f"T_cutoff: {t_cutoffs[0]}")
+        
         context_str = ", ".join(events) 
         logger.info(f"Event Stream ({len(events)}): [{context_str}]")