Created
June 3, 2024 20:54
-
-
Save juvi21/6b3a059686e189b293b2e047ec63e3b8 to your computer and use it in GitHub Desktop.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| from transformers import AutoModelForCausalLM, AutoTokenizer | |
| from deepeval.models.base_model import DeepEvalBaseLLM | |
| from deepeval.benchmarks import MMLU, GSM8K | |
| import pandas as pd | |
| # Define custom model class | |
| class Hermes2ThetaLlama3_8B(DeepEvalBaseLLM): | |
| def __init__(self, model, tokenizer): | |
| self.model = model | |
| self.tokenizer = tokenizer | |
| def load_model(self): | |
| return self.model | |
| def generate(self, prompt: str) -> str: | |
| model = self.load_model() | |
| device = "cuda" # Use GPU | |
| model_inputs = self.tokenizer([prompt], return_tensors="pt").to(device) | |
| model.to(device) | |
| generated_ids = model.generate(**model_inputs, max_new_tokens=100, do_sample=True) | |
| return self.tokenizer.batch_decode(generated_ids)[0] | |
| async def a_generate(self, prompt: str) -> str: | |
| return self.generate(prompt) | |
| def batch_generate(self, prompts: list) -> list: | |
| model = self.load_model() | |
| device = "cuda" | |
| model_inputs = self.tokenizer(prompts, return_tensors="pt").to(device) | |
| model.to(device) | |
| generated_ids = model.generate(**model_inputs, max_new_tokens=100, do_sample=True) | |
| return self.tokenizer.batch_decode(generated_ids) | |
| def get_model_name(self): | |
| return "Hermes-2-Theta-Llama-3-8B" | |
| # Load model and tokenizer | |
| model_name = "NousResearch/Hermes-2-Theta-Llama-3-8B" | |
| tokenizer = AutoTokenizer.from_pretrained(model_name) | |
| model = AutoModelForCausalLM.from_pretrained(model_name) | |
| # Create custom model instance | |
| hermes_2_theta_llama = Hermes2ThetaLlama3_8B(model=model, tokenizer=tokenizer) | |
| # Benchmark on MMLU | |
| mmlu_benchmark = MMLU() # 0-shot evaluation | |
| mmlu_results = mmlu_benchmark.evaluate(model=hermes_2_theta_llama) | |
| print("MMLU Overall Score:", mmlu_results.overall_score) | |
| # Benchmark on GSM8K | |
| gsm8k_benchmark = GSM8K(n_problems=1319, n_shots=0, enable_cot=False) # 0-shot evaluation | |
| gsm8k_results = gsm8k_benchmark.evaluate(model=hermes_2_theta_llama) | |
| print("GSM8K Overall Score:", gsm8k_results.overall_score) | |
| # Save results to files | |
| mmlu_task_scores_df = pd.DataFrame(mmlu_benchmark.task_scores) | |
| gsm8k_task_scores_df = pd.DataFrame(gsm8k_benchmark.task_scores) | |
| mmlu_predictions_df = pd.DataFrame(mmlu_benchmark.predictions) | |
| gsm8k_predictions_df = pd.DataFrame(gsm8k_benchmark.predictions) | |
| mmlu_task_scores_df.to_csv("mmlu_task_scores.csv", index=False) | |
| gsm8k_task_scores_df.to_csv("gsm8k_task_scores.csv", index=False) | |
| mmlu_predictions_df.to_csv("mmlu_predictions.csv", index=False) | |
| gsm8k_predictions_df.to_csv("gsm8k_predictions.csv", index=False) | |
| # Print detailed scores | |
| print("MMLU Task-specific Scores: ", mmlu_benchmark.task_scores) | |
| print("GSM8K Task-specific Scores: ", gsm8k_benchmark.task_scores) | |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment