bophades-mistral-math-DPO-7B

bophades-v2-mistral-7B finetuned on kyujinpy/orca_math_dpo.

Method

Finetuned using an A100 on Google Colab. 🙏

Fine-tune a Mistral-7b model with Direct Preference Optimization - Maxime Labonne

Configuration

LoRA, model, and training settings:

# LoRA configuration
peft_config = LoraConfig(
 r=16,
 lora_alpha=16,
 lora_dropout=0.05,
 bias="none",
 task_type="CAUSAL_LM",
 target_modules=['k_proj', 'gate_proj', 'v_proj', 'up_proj', 'q_proj', 'o_proj', 'down_proj']
)

# Model to fine-tune
model = AutoModelForCausalLM.from_pretrained(
 model_name,
 torch_dtype=torch.bfloat16,
 load_in_4bit=True
)
model.config.use_cache = False

# Reference model
ref_model = AutoModelForCausalLM.from_pretrained(
 model_name,
 torch_dtype=torch.bfloat16,
 load_in_4bit=True
)

# Training arguments
training_args = TrainingArguments(
 per_device_train_batch_size=2,
 gradient_accumulation_steps=2,
 gradient_checkpointing=True,
 learning_rate=2e-5,
 lr_scheduler_type="cosine",
 max_steps=420,
 save_strategy="no",
 logging_steps=1,
 output_dir=new_model,
 optim="paged_adamw_32bit",
 warmup_steps=100,
 bf16=True,
 report_to="wandb",
)

# Create DPO trainer
dpo_trainer = DPOTrainer(
 model,
 ref_model,
 args=training_args,
 train_dataset=dataset,
 tokenizer=tokenizer,
 peft_config=peft_config,
 beta=0.1,
 max_prompt_length=1024,
 max_length=1536,
 force_use_ref_model=True
)

# Fine-tune model with DPO
dpo_trainer.train()