Upload train_brand_voice.py with huggingface_hub

train_brand_voice.py

@@ -0,0 +1,65 @@
+# /// script
+# dependencies = ["trl>=0.12.0", "peft>=0.7.0", "trackio", "datasets", "transformers", "accelerate", "bitsandbytes"]
+# ///
+
+from datasets import load_dataset
+from peft import LoraConfig
+from trl import SFTTrainer, SFTConfig
+import trackio
+import os
+
+# Get HF token from environment
+hf_token = os.environ.get("HF_TOKEN")
+
+print("Loading SkyyRose brand voice dataset...")
+dataset = load_dataset("damBruh/skyyrose-brand-voice-training", split="train", token=hf_token)
+print(f"Dataset loaded: {len(dataset)} examples")
+
+# LoRA configuration for efficient fine-tuning
+peft_config = LoraConfig(
+    r=16,
+    lora_alpha=32,
+    lora_dropout=0.05,
+    target_modules=["q_proj", "k_proj", "v_proj", "o_proj", "gate_proj", "up_proj", "down_proj"],
+    bias="none",
+    task_type="CAUSAL_LM",
+)
+
+# Training configuration
+training_args = SFTConfig(
+    output_dir="skyyrose-brand-voice-llm",
+    num_train_epochs=3,
+    per_device_train_batch_size=2,
+    gradient_accumulation_steps=4,
+    learning_rate=2e-4,
+    warmup_ratio=0.1,
+    weight_decay=0.01,
+    max_grad_norm=1.0,
+    logging_steps=1,
+    save_strategy="epoch",
+    save_total_limit=2,
+    gradient_checkpointing=True,
+    push_to_hub=True,
+    hub_model_id="damBruh/skyyrose-brand-voice-llm",
+    hub_token=hf_token,
+    report_to="trackio",
+    run_name="skyyrose-brand-voice-qwen2.5",
+    seed=42,
+)
+
+print("Initializing SFTTrainer with Qwen/Qwen2.5-1.5B-Instruct...")
+trainer = SFTTrainer(
+    model="Qwen/Qwen2.5-1.5B-Instruct",
+    train_dataset=dataset,
+    peft_config=peft_config,
+    args=training_args,
+)
+
+print("Starting training...")
+trainer.train()
+
+print("Saving and pushing to Hub...")
+trainer.save_model()
+trainer.push_to_hub()
+
+print("SUCCESS: Model available at huggingface.co/damBruh/skyyrose-brand-voice-llm")