import random from internbootcamp.bootcamp.base import Basebootcamp from internbootcamp.libs.chemStructure2Property.ChemStructureGenerator import InChIGenerator from internbootcamp.bootcamp.ChemStructure2Property.utils import last_boxed_only_string, remove_boxed from rdkit import Chem from rdkit.Chem import Crippen class InChI2logPbootcamp(Basebootcamp): def __init__(self, max_atoms=15, min_atoms=3, elements=None, seed=None): # super.__init__() self.max_atoms = max_atoms self.min_atoms = min_atoms # self.InChIGenerator = InChIGenerator(max_atoms=max_atoms, min_atoms=min_atoms, elements=elements, seed=seed) # self.tolerance_factor = tolerance_factor # 1 for 1% error consider true, 0.1 for 0.1% error true, 10 for 10% error def case_generator(self) -> str: """ 生成一组数字和目标值。 """ self.InChIGenerator = InChIGenerator(max_atoms=self.max_atoms, min_atoms=self.min_atoms, elements=None, seed=None) inchis = self.InChIGenerator.generate_n_valid_inchi(10) # print(inchis) n = random.randint(0, 9) # print(n) return inchis[n] def prompt_func(self, InChI) -> str: instruction = f"Given the InChI, determine the lipophilicity (logP) value of the material. The InChI is: {InChI}" instruction_following = """Let's think step by step and output the final answer within \\boxed{}.The final answer should be one float number. For example "Final Answer: \\boxed{afloat}".""" prompt = instruction + '\n' + instruction_following return prompt @staticmethod def extract_output(output): """ Extract the output from the solution. Args: output: Model output to be processed. Returns: The processed output. """ output = last_boxed_only_string(output) if output is None: return None return remove_boxed(output) @classmethod def _verify_correction(cls, solution, InChI) -> float: """ Verify the correction of the solution and return a score between 0 and 1. The score is based on the relative error with respect to a maximum relative error of 0.1. """ mol = Chem.MolFromInchi(InChI) true_logp = Crippen.MolLogP(mol) solution_float = float(solution) # Handle case where true_logp is 0 if true_logp == 0: # If true_logp is 0, we check how close the solution is to 0 relative_error = abs(solution_float) else: # Calculate the relative error relative_error = abs(true_logp - solution_float) / abs(true_logp) # Define the maximum allowed relative error max_relative_error = 0.1 # Calculate the score based on the relative error if relative_error >= max_relative_error: return 0.0 # Error is too large, score is 0 else: # Linear interpolation: score decreases linearly from 1 to 0 as error goes from 0 to max_relative_error return 1.0 return 1 - (relative_error / max_relative_error) * 0.5 ## For RL if __name__ == "__main__": bootcamp = InChI2logPbootcamp() while True: case = bootcamp.case_generator() print('case') print(case) input()