Neural Krishna DPO
Fine-tuning + lnegth(choose)
- Training Args:
# 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.float16,
load_in_4bit=True
)
model.config.use_cache = False
# Training arguments
training_args = TrainingArguments(
per_device_train_batch_size=4,
gradient_accumulation_steps=4,
gradient_checkpointing=True,
learning_rate=5e-5,
lr_scheduler_type="cosine",
max_steps=120,
save_strategy="no",
logging_steps=1,
output_dir=new_model,
optim="paged_adamw_32bit",
warmup_steps=50,
bf16=True,
report_to="wandb",
)
# Create DPO trainer
dpo_trainer = DPOTrainer(
model,
args=training_args,
train_dataset=dataset,
tokenizer=tokenizer,
peft_config=peft_config,
beta=0.1,
max_prompt_length=1024,
max_length=1536,
)
# Fine-tune model with DPO
dpo_trainer.train()
Open LLM Leaderboard Evaluation Results
Detailed results can be found here
Metric | Value |
---|---|
Avg. | 76.00 |
AI2 Reasoning Challenge (25-Shot) | 74.06 |
HellaSwag (10-Shot) | 88.97 |
MMLU (5-Shot) | 64.41 |
TruthfulQA (0-shot) | 76.19 |
Winogrande (5-shot) | 84.29 |
GSM8k (5-shot) | 68.08 |
- Downloads last month
- 86
This model does not have enough activity to be deployed to Inference API (serverless) yet. Increase its social
visibility and check back later, or deploy to Inference Endpoints (dedicated)
instead.
Model tree for Kukedlc/NeuralKrishna-7B-V2-DPO
Spaces using Kukedlc/NeuralKrishna-7B-V2-DPO 6
Evaluation results
- normalized accuracy on AI2 Reasoning Challenge (25-Shot)test set Open LLM Leaderboard74.060
- normalized accuracy on HellaSwag (10-Shot)validation set Open LLM Leaderboard88.970
- accuracy on MMLU (5-Shot)test set Open LLM Leaderboard64.410
- mc2 on TruthfulQA (0-shot)validation set Open LLM Leaderboard76.190
- accuracy on Winogrande (5-shot)validation set Open LLM Leaderboard84.290
- accuracy on GSM8k (5-shot)test set Open LLM Leaderboard68.080