SimpleBerry / LLaMA-O1

Large Reasoning Models
537 stars 30 forks source link

root_node.state contains solution, is it reasonable? #11

Open fenngMing opened 1 day ago

fenngMing commented 1 day ago

first, initial_state=self.envoirment.sample_initial_state() #here returns both initial state and solution then root_node = TreeNode(state=initial_state) #here root_node contans solution too third path_to_string here solution leaks to prompt f"{node.parent.index if node.parent else -1}{node.index}{node.state}{value_to_rating_token(node.value)}" for node in path[::-1]

trotsky1997 commented 1 day ago

Will be fix in next commit

trotsky1997 commented 1 day ago

initial_state, gt=self.envoirment.sample_initial_state()

return of sample_initial_state has been changed, but I have forget to change corresponding call, sorry