yuhan
yuhan
大概2分7秒跳一次这个，番遍了源码没找到127秒这个数据
> io.mycat.MycatServer > > heartbeatScheduler.scheduleAtFixedRate(updateTime(), 0L, TIME_UPDATE_PERIOD, TimeUnit.MILLISECONDS); > heartbeatScheduler.scheduleAtFixedRate(processorCheck(), 0L, system.getProcessorCheckPeriod(), TimeUnit.MILLISECONDS); > heartbeatScheduler.scheduleAtFixedRate(dataNodeConHeartBeatCheck(dataNodeIldeCheckPeriod), 0L, dataNodeIldeCheckPeriod, TimeUnit.MILLISECONDS); > heartbeatScheduler.scheduleAtFixedRate(dataNodeHeartbeat(), 0L, system.getDataNodeHeartbeatPeriod(), TimeUnit.MILLISECONDS); > heartbeatScheduler.scheduleAtFixedRate(dataSourceOldConsClear(), 0L, DEFAULT_OLD_CONNECTION_CLEAR_PERIOD, TimeUnit.MILLISECONDS); > heartbeatScheduler.scheduleAtFixedRate(dataNodeCalcActiveCons(),...
> 感觉有毒 我看了下源码，setDataNodeHeartbeatPeriod那里是灰的，应该是根本没有使用正常的设置方法，我猜也许是搞了个什么轮询或者设置缓存，导致修改完这个设置根本没得反应
> 我运行了代码库中的moss_api_demo.py，使用的是fnlp/moss-moon-003-sft-int4模型， 使用postman进行测试api的情况，看到返回的请奇怪，有时是Machine，有时是MOSS，有时还是Python，这个是什么原因？另外，返回的这个东西代表什么含义？ > > ``` > [2023-05-11 15:42:02] ", prompt:": 你平时如何锻炼？", response:"'\n: 我是一个机器人，没有身体，无法进行体育锻炼。不过我可以利用我的计算能力和智能算法帮助人类进行运动训练和健康管理。比如，我可以根据用户的身体数据和目标制定个性化的运动计划，或者提供健康饮食建议等。'" > INFO: 27.42.96.148:17606 - "POST / HTTP/1.1" 200 OK > [2023-05-11 15:42:57] ", prompt:": 什么是WASP？", response:"'\n:...
> Adding the parameter `strict=False` parameter in line 70 of `utils.model.model_utils.create_critic_model` lets me load the checkpoint: > > ``` > def create_critic_model(model_name_or_path, > tokenizer, > ds_config, > num_padding_at_beginning=0, > rlhf_training=False):...
> Thank you for your contributions. We are indeed interested in your multi-gpu implementation. As requested, please consider merging the original inference script and yours into a unified one, so...
> Is it possible to use one script for both single-gpu and multi-gpu inference (for instance, users can launch the script with `--single-gpu` (default) or `--multi-gpu`) ? so that we...
> Hi, > > let me try to understand your question: > > You are asking if the Alpaca-LoRA adapter weights (e.g. https://huggingface.co/tloen/alpaca-lora-7b) that were trained on top of the...
> I understand. Based on my experience and knowledge, I think that what you propose is not possible, as the weights of the original LlaMA model and the OpenLlaMA model...
我使用的体验下来，llama-based的这些llm都比glm效果好一些，不过见仁见智吧