Upload train_unesco_tagger.py with huggingface_hub

train_unesco_tagger.py

@@ -0,0 +1,69 @@
+# /// script
+# dependencies = [
+#     "trl>=0.12.0",
+#     "peft>=0.7.0",
+#     "transformers>=4.36.0",
+#     "accelerate>=0.24.0",
+#     "trackio",
+# ]
+# ///
+
+from datasets import load_dataset
+from peft import LoraConfig
+from trl import SFTTrainer, SFTConfig
+
+print("Loading dataset...")
+dataset = load_dataset("unesco-data-ai/unesco-thesaurus-sft")
+train_dataset = dataset["train"]
+eval_dataset = dataset["validation"]
+
+print(f"Train: {len(train_dataset)}, Eval: {len(eval_dataset)}")
+
+config = SFTConfig(
+    output_dir="qwen2.5-3b-unesco-tagger",
+    push_to_hub=True,
+    hub_model_id="unesco-data-ai/qwen2.5-3b-unesco-tagger-v1",
+    hub_strategy="every_save",
+    num_train_epochs=3,
+    per_device_train_batch_size=2,
+    gradient_accumulation_steps=8,
+    learning_rate=2e-5,
+    max_length=2048,
+    logging_steps=10,
+    save_strategy="steps",
+    save_steps=200,
+    save_total_limit=2,
+    eval_strategy="steps",
+    eval_steps=200,
+    warmup_ratio=0.1,
+    lr_scheduler_type="cosine",
+    report_to="trackio",
+    project="unesco-keyword-extraction",
+    run_name="qwen2.5-3b-sft-v1",
+)
+
+peft_config = LoraConfig(
+    r=16,
+    lora_alpha=32,
+    lora_dropout=0.05,
+    bias="none",
+    task_type="CAUSAL_LM",
+    target_modules=["q_proj", "k_proj", "v_proj", "o_proj"],
+)
+
+print("Initializing trainer...")
+trainer = SFTTrainer(
+    model="Qwen/Qwen2.5-3B-Instruct",
+    train_dataset=train_dataset,
+    eval_dataset=eval_dataset,
+    args=config,
+    peft_config=peft_config,
+)
+
+print("Starting training...")
+trainer.train()
+
+print("Pushing to Hub...")
+trainer.push_to_hub()
+
+print("Complete!")