Upload train_survival.py with huggingface_hub

train_survival.py

@@ -0,0 +1,89 @@
+# /// script
+# dependencies = ["trl", "peft", "bitsandbytes", "datasets", "transformers"]
+# ///
+
+from datasets import load_dataset
+from peft import LoraConfig
+from trl import SFTTrainer, SFTConfig
+from transformers import AutoModelForCausalLM, AutoTokenizer, BitsAndBytesConfig
+import torch
+import os
+
+# Configuration
+MODEL_ID = "Qwen/Qwen2.5-3B-Instruct"
+DATASET_ID = "sunkencity/survival-instruct"
+OUTPUT_MODEL_ID = "sunkencity/survival-expert-3b"
+
+# Load Dataset
+dataset = load_dataset(DATASET_ID, split="train")
+
+# Load Model with Quantization (for efficiency)
+bnb_config = BitsAndBytesConfig(
+    load_in_4bit=True,
+    bnb_4bit_quant_type="nf4",
+    bnb_4bit_compute_dtype=torch.float16,
+)
+
+model = AutoModelForCausalLM.from_pretrained(
+    MODEL_ID,
+    quantization_config=bnb_config,
+    device_map="auto",
+    use_cache=False
+)
+tokenizer = AutoTokenizer.from_pretrained(MODEL_ID)
+tokenizer.pad_token = tokenizer.eos_token
+
+# LoRA Configuration
+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", "gate_proj", "up_proj", "down_proj"]
+)
+
+# Training Arguments
+training_args = SFTConfig(
+    output_dir="./results",
+    num_train_epochs=3,
+    per_device_train_batch_size=4,
+    gradient_accumulation_steps=4,
+    learning_rate=2e-4,
+    logging_steps=10,
+    push_to_hub=True,
+    hub_model_id=OUTPUT_MODEL_ID,
+    fp16=True,
+    max_seq_length=1024,
+    dataset_text_field="text", # We need to format the data first if it's not in 'text'
+    packing=False
+)
+
+# Formatting function for SFT (Chat format)
+def formatting_prompts_func(example):
+    output_texts = []
+    for i in range(len(example['instruction'])):
+        instruction = example['instruction'][i]
+        response = example['response'][i]
+        
+        # Qwen/Llama chat template format
+        text = f"<|im_start|>user\n{instruction}<|im_end|>\n<|im_start|>assistant\n{response}<|im_end|>"
+        output_texts.append(text)
+    return output_texts
+
+# Trainer
+trainer = SFTTrainer(
+    model=model,
+    train_dataset=dataset,
+    peft_config=peft_config,
+    formatting_func=formatting_prompts_func,
+    args=training_args,
+    tokenizer=tokenizer,
+)
+
+print("Starting training...")
+trainer.train()
+
+print("Pushing to hub...")
+trainer.push_to_hub()
+print("Done!")