Upload train_qwen3_vl.py with huggingface_hub

train_qwen3_vl.py

@@ -0,0 +1,79 @@
+# /// script
+# dependencies = [
+#   "trl>=0.12.0",
+#   "peft>=0.7.0",
+#   "trackio",
+#   "transformers>=4.45.0",
+#   "torch",
+#   "datasets",
+#   "pillow",
+#   "qwen-vl-utils"
+# ]
+# ///
+
+from datasets import load_dataset
+from peft import LoraConfig
+from trl import SFTTrainer, SFTConfig
+from transformers import AutoProcessor, Qwen2VLForConditionalGeneration
+import trackio
+import torch
+
+# Load 1% of the train split
+print("Loading dataset...")
+dataset = load_dataset("trl-lib/llava-instruct-mix", split="train[:1%]")
+
+print(f"Dataset size: {len(dataset)} examples")
+
+# Create a small eval split (10% of the 1%)
+dataset_split = dataset.train_test_split(test_size=0.1, seed=42)
+train_dataset = dataset_split["train"]
+eval_dataset = dataset_split["test"]
+
+print(f"Train: {len(train_dataset)}, Eval: {len(eval_dataset)}")
+
+# Configure trainer with VL-specific settings
+trainer = SFTTrainer(
+    model="Qwen/Qwen3-VL-3B-Instruct",
+    train_dataset=train_dataset,
+    eval_dataset=eval_dataset,
+    peft_config=LoraConfig(
+        r=16,
+        lora_alpha=32,
+        target_modules=["q_proj", "k_proj", "v_proj", "o_proj"],
+        lora_dropout=0.05,
+        bias="none",
+        task_type="CAUSAL_LM"
+    ),
+    args=SFTConfig(
+        output_dir="qwen3-vl-3b-llava-instruct",
+        push_to_hub=True,
+        hub_model_id="merve/qwen3-vl-3b-llava-instruct",
+        num_train_epochs=3,
+        per_device_train_batch_size=1,
+        gradient_accumulation_steps=8,
+        gradient_checkpointing=True,
+        learning_rate=2e-4,
+        warmup_steps=100,
+        logging_steps=10,
+        eval_strategy="steps",
+        eval_steps=50,
+        save_strategy="steps",
+        save_steps=100,
+        save_total_limit=2,
+        bf16=True,
+        report_to="trackio",
+        project="qwen3-vl-finetuning",
+        run_name="qwen3-vl-3b-llava-1pct",
+        max_length=None,  # Important for VL models - don't truncate image tokens
+        hub_strategy="every_save",
+        remove_unused_columns=False,  # Keep all columns for VL processing
+    )
+)
+
+print("Starting training...")
+trainer.train()
+
+print("Pushing final model to Hub...")
+trainer.push_to_hub()
+
+print("Training complete!")