File size: 4,246 Bytes
7155cf2 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 |
# Copyright 2024 Bytedance Ltd. and/or its affiliates
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
from collections import defaultdict
from typing import Callable, Dict, List, Tuple, TypedDict
import torch
from transformers import PreTrainedTokenizer
from ...protocol import DataProto
from ...utils.reward_score import math_compute_score, r1v_compute_score, reason_with_in_limit_compute_score
from .config import RewardConfig
class RewardScore(TypedDict):
overall: float
format: float
accuracy: float
class CustomRewardManager:
def __init__(self, tokenizer: PreTrainedTokenizer, config: RewardConfig):
self.config = config
self.tokenizer = tokenizer
if config.score_function == "math":
self.compute_score: Callable[[str, str], RewardScore] = math_compute_score
elif config.score_function == "r1v":
self.compute_score: Callable[[str, str], RewardScore] = r1v_compute_score
elif config.score_function == "reason_with_in_limit":
self.reason_with_in_limit_compute_score: Callable[[str, str, int, int], RewardScore] = reason_with_in_limit_compute_score
else:
raise NotImplementedError(f"Unknown score function {config.score_function}.")
def __call__(self, data: DataProto) -> Tuple[torch.Tensor, Dict[str, List[float]]]:
reward_tensor = torch.zeros_like(data.batch["responses"], dtype=torch.float32)
reward_metrics = defaultdict(list)
for i in range(len(data)):
data_item = data[i]
# g 奖励函数所在处+dataitem数据读取用法
# DataProtoItem
response_ids = data_item.batch["responses"]
response_mask = data_item.batch["response_mask"]
budget_and_tokens = data_item.batch["budget_and_tokens"]
origin_response_length = data_item.batch["origin_response_length"]
prompt_ids = data_item.batch['prompts']
# print("*" * 100 + "budget_and_tokens.shape", budget_and_tokens.shape, "*" * 100)
# print("*" * 100 + "origin_response_length.shape", origin_response_length.shape, "*"* 100)
valid_response_length = response_mask.sum()
valid_response_ids = response_ids[:valid_response_length]
valid_prompt_ids = prompt_ids
response_str = self.tokenizer.decode(
valid_response_ids, skip_special_tokens=self.config.skip_special_tokens
)
raw_response_str = self.tokenizer.decode(
valid_response_ids, skip_special_tokens=True
)
# print(f"raw_response_str: {raw_response_str}")
prompt_str = self.tokenizer.decode(
valid_prompt_ids, skip_special_tokens=True
)
budget = data_item.non_tensor_batch["budget"]
current_epoch = data_item.non_tensor_batch["current_epoch"]
ground_truth = data_item.non_tensor_batch["ground_truth"]
if self.config.score_function == "math" or self.config.score_function == "r1v":
score = self.compute_score(response_str, ground_truth)
elif self.config.score_function == "reason_with_in_limit":
score = self.reason_with_in_limit_compute_score(response_str, ground_truth, current_length=origin_response_length, budget=budget_and_tokens, current_epoch=current_epoch, prompt_str=prompt_str, raw_response_str=raw_response_str)
reward_tensor[i, valid_response_length - 1] = score["overall"]
for key, value in score.items():
reward_metrics[key].append(value)
return reward_tensor, reward_metrics
|