# pip install modelscope[framework]# 模型库，已经预安装pip install py-data-juicer[sci]# 数据处理库# pip install ms-swift[llm]# 训练库，已经预安装# pip install ms-swift[eval] # 评测库，已经预安装

from modelscope import MsDatasetimport jsonimport pandas as pd
# 下载数据ds =MsDataset.load('OmniData/Zhihu-KOL', cache_dir="data", split='train')
# 处理 metadatametadata = list(map(lambda x: json.loads(x), ds['METADATA']))
# 处理 upvotes vote_list = []for item in metadata:try:upvotes = item['upvotes'][3:]if not upvotes:votes = 0elif '万' in upvotes:votes = int(float(upvotes[:-2]) * 10000)else:votes = int(upvotes)except Exception as e:votes = 0vote_list.append(votes)
# 写入 jsonl 文件df = pd.DataFrame.from_dict({'query': ds['INSTRUCTION'],'response': ds['RESPONSE'],'upvotes': vote_list})
df.to_json("data/zhihu.jsonl", orient="records", lines=True, force_ascii=False)

{'INSTRUCTION': '怎么说服男朋友买烤箱？', 'METADATA': '{"question_id": 357137111.0, "answer_id": 914332816.0, "url": '"https://www.zhihu.com/question/357137111/answer/914332816", ' '"upvotes": "赞同 15", "answer_creation_time": ' '"2019-11-28T12:01:22.000Z"}', 'RESPONSE': 'emmmmm，首先想说的是，我买厨房用品一般是不用「说服」的，只是在厨房堆的满满当当的情况下会象征性的问一下我老公，他就会回答我说：你看看你还有地方放吗。然后我会思考一下，如果是特别想买的，就不会问他了。自己决定就好。' '比如，前几天我又买了两个盘子~~~~他还不知道。可以给题主看看我有多少的锅具：自家炒菜用什么锅好？各有什么优缺点？' '说回烤箱的问题，买的时候处于热恋期，我告诉他我有一个买烤箱的计划。虽然他基本不吃点心，也不喜欢烘焙，但那个时期的他欣然同意并热情洋溢的给我选烤箱。可能是他有憧憬我会给他做什么好吃的吧。又因为我是一个不怎么吃甜食的湖南人，烤箱在我家烘焙的使用率很低。' '但是！！你还是可以告诉他烤箱的作用是可以烤制各种肉类！！！我不相信有不喜欢吃肉的男生！！烤箱真的是可以烤一切的肉类，熟悉之后会觉得非常简单。' '我很久以前用烤箱做的最多的就是烤羊排和烤鸡翅，我老公不怎么吃羊肉和鸡翅。这个烤箱因为厨房放不下，被放在了餐厅，也就闲置了下来…… ' '要说的事是，烤箱真的能给你做出很多不一样的美食，尤其是来了客人，在你两个灶台忙不过来的时候，烤箱特别适合准备一个荤素搭配的豪华大菜。在烹饪其他需要爆炒的菜肴的空档去处理一下就可以了。' '总结来说理由如下：1、如果你家是你做饭多，那么为什么有这么多话说， 也不是他用，等着吃就好了。' '2、工欲善其事，必先利其器。没有好的工具怎么能吃到更好的美食。3、我要我喜欢，不要你喜欢。我还不能有个爱好吗？', 'SOURCE': 'Zhihu'}

# global parametersproject_name: 'zhihu-process'dataset_path: 'data/zhihu.jsonl'# path to your dataset directory or filenp: 16# number of subprocess to process your dataset
text_keys: 'response' # key of text in your dataset file
export_path: 'data/zhihu_refine.jsonl'# path to save processed dataset
# process schedule# a list of several process operators with their argumentsprocess:- specified_numeric_field_filter: # filter text with the specified numeric field info out of specific rangefield_key: 'upvotes'# the target key corresponding to multi-level field information need to be separated by '.'min_value: 500# the min filter value in SpecifiedNumericField op- text_length_filter: # filter text with the length out of specific rangemin_len: 100max_len: 2000
- clean_email_mapper: # remove emails from text.- clean_html_mapper:# remove html formats form text.- clean_ip_mapper:# remove ip addresses from text.- clean_links_mapper: # remove web links from text.- clean_copyright_mapper: # remove copyright comments.# fix unicode errors in text.
- language_id_score_filter: # filter text in specific language with language scores larger than a specific max valuelang: zhmin_score: 0.9- alphanumeric_filter:# filter text with alphabet/numeric ratio out of specific range.tokenization: falsemin_ratio: 0.72- flagged_words_filter: # filter text with the flagged-word ratio larger than a specific max valuelang: zhtokenization: falsemax_ratio: 0.0005- perplexity_filter:# filter text with perplexity score out of specific rangelang: zhmax_ppl: 4000- special_characters_filter:# filter text with special-char ratio out of specific rangemax_ratio: 0.4- document_simhash_deduplicator:# deduplicate texts with simhashtokenization: characterwindow_size: 5lowercase: falseignore_pattern: '\p{P}'num_blocks: 10hamming_distance: 6 # larger hamming distance threshold for short texts- topk_specified_field_selector:# selector to select top samples based on the sorted specified fieldfield_key: 'upvotes'# the target keys corresponding to multi-level field information need to be separated by '.'topk: 50000 # number of selected top sample      reverse: True                                           # determine the sorting rule, if reverse=True, then sort in descending order

dj-analyze --config zhihu-bot.yaml

dj-process --config zhihu-bot.yaml

import pandas as pd
data = pd.read_json("data/zhihu_refine.jsonl", lines=True)
def split_data(data, save=False, suffix=''):# split data into train and test, 9: 1train_data = data.sample(frac=0.9, random_state=42)test_data = data.drop(train_data.index)
if suffix:suffix = '_' + suffixif save:train_data.to_json(f"data/zhihu_train{suffix}.jsonl", orient='records', lines=True, force_ascii=False)test_data.to_json(f"data/zhihu_test{suffix}.jsonl", orient='records', lines=True,force_ascii=False)return train_data, test_data
train_data, test_data = split_data(data, save=True)

pip install flash-attn --no-build-isolation

CUDA_VISIBLE_DEVICES=0 swift sft \--sft_type lora \--model_type qwen2-7b-instruct \--model_id_or_path qwen/Qwen2-7B-Instruct \--dataset data/zhihu_train.jsonl#4000 alpaca-zh#2000 \ --system "你是一个真实的人类。回答用户的问题，并在聊天回复中展现你的情绪，包括开心，幸福，愤怒，伤心，阴阳怪气等" \--dataset_test_ratio 0.01 \--output_dir output \--lora_target_modules ALL \--lora_rank 8 \--dtype bf16 \--seed 42 \--learning_rate 1e-4 \--warmup_ratio 0.05 \--max_length 2048 \--batch_size 4 \--eval_batch_size 4 \--num_train_epochs 1 \--gradient_accumulation_steps 4 \--save_total_limit 10 \--eval_steps 100 \--save_steps 100

./train.sh

{"query":"微信头像会影响第一印象吗？","response":"不行了！我实在是忍不住要回答这个问题了！这是我之前的头像 然后通知群老师发消息 哈哈哈哈哈哈哈哈哈我发完之后 就没有人敢说话了哈哈哈哈哈哈哈哈哈 这个头像真的是一脸“竟有此事！”  然后 然后我跟朋友吐槽这个事    原图给你们安排上了：5.28更新：今天突然发现已经两千赞了，谢谢大家喜欢这个回答！补一个情侣头像：写在最后："}

[{"name": "custom_general_qa","pattern": "general_qa","dataset": "data","subset_list": ["zhihu_test"]}]

CUDA_VISIBLE_DEVICES=0 swift eval \--ckpt_dir output/qwen2-7b-instruct/v1-20240819-150005/checkpoint-371 \--eval_dataset no \--infer_backend pt \--eval_backend Native \--eval_limit 10 \--seed 42 \--eval_batch_size 8 \--custom_eval_config custom_eval_config.json \--temperature 0.7 \--top_k 20 \--top_p 0.9

{"result": {"data": {"rouge-1-r": 0.1366327464084804, "rouge-1-p": 0.3397212949722054, "rouge-1-f": 0.1453481684882953, "rouge-2-r": 0.03827942419095308, "rouge-2-p": 0.11396557995638323, "rouge-2-f": 0.03626899512109694, "rouge-l-r": 0.1234295688857564, "rouge-l-p": 0.15583028795014991, "rouge-l-f": 0.08378730853798907, "bleu-1": 0.055066495373721956, "bleu-2": 0.01267421096081624, "bleu-3": 0.0009279523752259867, "bleu-4": 1.1801272718452154e-308}}, "model": "qwen2-7b-instruct", "time": "20240819_153042"}

import osos.environ['CUDA_VISIBLE_DEVICES'] = '0'import pandas as pd
from swift.llm import (get_model_tokenizer, get_template, inference, ModelType, get_default_template_type,)from swift.utils import seed_everythingfrom swift.tuners import Swiftimport torch
seed_everything(42)
def infer_querys(model, template, querys):if type(querys) == str:querys = [querys]responses = []for query in querys:response, history = inference(model, template, query)response = response.replace("\n", "\t")responses.append(response)print(f'response: {response}')return responses
def load_model(ckpt_dir):model_type = ModelType.qwen2_7b_instructtemplate_type = get_default_template_type(model_type)
model, tokenizer = get_model_tokenizer(model_type, model_kwargs={'device_map': 'auto'})model.generation_config.max_new_tokens = 500model.generation_config.temperature = 0.7model.generation_config.top_p = 0.9model.generation_config.top_k = 20if ckpt_dir:model = Swift.from_pretrained(model, ckpt_dir, inference_mode=True)system_prompt = "你是一个真实的人类。回答用户的问题，并在聊天回复中展现你的情绪，包括开心，幸福，愤怒，伤心，阴阳怪气等"template = get_template(template_type, tokenizer, default_system=system_prompt)return model, template
querys = pd.read_json("data/zhihu_test.jsonl", lines=True)["query"].sample(10, random_state=42).tolist()querys = ["你是谁?"] + querys
print(querys)
ckpt_dict = {'origin': None,'lora': 'output/qwen2-7b-instruct/v1-20240819-150005/checkpoint-371',}model = Nonemodel_responses = {}for ckpt_name, ckpt_dir in ckpt_dict.items():if model:del modeltorch.cuda.empty_cache()model, template = load_model(ckpt_dir)model_responses[ckpt_name] = infer_querys(model, template, querys)df = pd.DataFrame.from_dict(model_responses)df.index = querysdf.to_markdown("output.md")

from modelscope.hub.api import HubApi
YOUR_ACCESS_TOKEN = '请从ModelScope个人中心->访问令牌获取'
api = HubApi()api.login(YOUR_ACCESS_TOKEN)api.push_model(model_id="AlexEz/zhihu_bot_lora", # 用户名/模型仓库名称model_dir="output/qwen2-7b-instruct/v1-20240819-150005/checkpoint-371" # 本地模型目录，要求目录中必须包含configuration.json)