Upload train_qwen3_vl.py with huggingface_hub

train_qwen3_vl.py

@@ -33,7 +33,7 @@ print(f"Train: {len(train_dataset)}, Eval: {len(eval_dataset)}")
 
 # Configure trainer with VL-specific settings
 trainer = SFTTrainer(
-    model="Qwen/Qwen3-VL-3B-Instruct",
+    model="Qwen/Qwen2.5-VL-3B-Instruct",
     train_dataset=train_dataset,
     eval_dataset=eval_dataset,
     peft_config=LoraConfig(
@@ -47,7 +47,7 @@ trainer = SFTTrainer(
     args=SFTConfig(
         output_dir="qwen3-vl-3b-llava-instruct",
         push_to_hub=True,
-        hub_model_id="merve/qwen3-vl-3b-llava-instruct",
+        hub_model_id="merve/qwen2.5-vl-3b-llava-instruct",
         num_train_epochs=3,
         per_device_train_batch_size=1,
         gradient_accumulation_steps=8,