南栀沁寒 发布的帖子

@第六神觉 检查一下权限和登录协议

在使用（remove reply）关闭某一个群的回复后这个群的词库被莫名其妙地完全清空了 这是正常的吗？

请给艾特bot回复的功能上一个开关 因为 我的@是用于另外一个回复插件的 但是我想用新版的cosmatch功能 另外我问问如果词库非常大 开启cosmatch功能是否会让程序回复速度变得很慢

@南栀沁寒 在 Molly-Mirai 接入茉莉云聊天机器人 中说：

佬 请问这是个什么情况

茉莉云的官网好像也打不开了 是不是挂了

佬 请问这是个什么情况

你好 请问可否实现只对一个人的消息进行单向转发?

你好 请问可不可以指定转发一个人的信息

建议是代币不够的时候返回个提示比较好

@Mitr-yuzr ok了！期待后续 不知道后面是打算怎样将训练好的模型接入到bot上呢 触发条件是什么（@?还是说概率回复的那种）

每一步都按部就班地完成了可是训练的时候只花了39秒就结束了好像么有训练上是什么情况 生成的文本也是默认的风格
2022-10-02 14:51:34.665015: W tensorflow/core/common_runtime/gpu/gpu_bfc_allocator.cc:39] Overriding allow_growth setting because the TF_FORCE_GPU_ALLOW_GROWTH environment variable is set. Original config value was 0.
args:
Namespace(batch_size=8, bpe_token=False, device='0,1,2,3', encoder_json='tokenizations/encoder.json', epochs=30, fp16=False, fp16_opt_level='O1', gradient_accumulation=1, log_step=1, lr=0.00015, max_grad_norm=1.0, min_length=128, model_config='/content/drive/MyDrive/mirai-gpt2/pretrain_model/config.json', num_pieces=31353, output_dir='/content/drive/MyDrive/mirai-gpt2/', pretrained_model='/content/drive/MyDrive/mirai-gpt2/pretrain_model', raw=True, raw_data_path='data/train.json', save_every=5, segment=False, start_epoch=0, stride=768, tokenized_data_path='data/tokenized/', tokenizer_path='/content/drive/MyDrive/mirai-gpt2/pretrain_model/vocab.txt', vocab_bpe='tokenizations/vocab.bpe', warmup_steps=2000, writer_dir='tensorboard_summary/')
config:
{
"activation_function": "gelu_new",
"architectures": [
"GPT2LMHeadModel"
],
"attn_pdrop": 0.1,
"embd_pdrop": 0.1,
"finetuning_task": null,
"gradient_checkpointing": false,
"initializer_range": 0.02,
"layer_norm_epsilon": 1e-05,
"model_type": "gpt2",
"n_ctx": 1024,
"n_embd": 768,
"n_head": 12,
"n_inner": null,
"n_layer": 6,
"n_positions": 1024,
"num_labels": 1,
"output_attentions": false,
"output_hidden_states": false,
"output_past": true,
"pruned_heads": {},
"resid_pdrop": 0.1,
"summary_activation": null,
"summary_first_dropout": 0.1,
"summary_proj_to_labels": true,
"summary_type": "cls_index",
"summary_use_proj": true,
"task_specific_params": {
"text-generation": {
"do_sample": true,
"max_length": 400
}
},
"tokenizer_class": "BertTokenizer",
"torchscript": false,
"use_bfloat16": false,
"vocab_size": 21128
}

using device: cuda
building files
reading lines
100% 31353/31353 [00:01<00:00, 25351.73it/s]
finish
files built
number of parameters: 59541504
calculating total steps
100% 31353/31353 [00:00<00:00, 46655.81it/s]
total steps = 63
starting training
epoch 1
time: 2022-10-02 14:51:40.207784
epoch 1 finished
time: 2022-10-02 14:51:40.822857
time for one epoch: 0:00:00.615073
epoch 2
time: 2022-10-02 14:51:40.822894
epoch 2 finished
time: 2022-10-02 14:51:41.449992
time for one epoch: 0:00:00.627098
epoch 3
time: 2022-10-02 14:51:41.450048
epoch 3 finished
time: 2022-10-02 14:51:42.077769
time for one epoch: 0:00:00.627721
epoch 4
time: 2022-10-02 14:51:42.077812
epoch 4 finished
time: 2022-10-02 14:51:42.708398
time for one epoch: 0:00:00.630586
epoch 5
time: 2022-10-02 14:51:42.708441
saving model for epoch 5
epoch 5 finished
time: 2022-10-02 14:51:44.504427
time for one epoch: 0:00:01.795986
epoch 6
time: 2022-10-02 14:51:44.504479
epoch 6 finished
time: 2022-10-02 14:51:45.159980
time for one epoch: 0:00:00.655501
epoch 7
time: 2022-10-02 14:51:45.160022
epoch 7 finished
time: 2022-10-02 14:51:45.775230
time for one epoch: 0:00:00.615208
epoch 8
time: 2022-10-02 14:51:45.775269
epoch 8 finished
time: 2022-10-02 14:51:46.400834
time for one epoch: 0:00:00.625565
epoch 9
time: 2022-10-02 14:51:46.400874
epoch 9 finished
time: 2022-10-02 14:51:47.023521
time for one epoch: 0:00:00.622647
epoch 10
time: 2022-10-02 14:51:47.023573
saving model for epoch 10
epoch 10 finished
time: 2022-10-02 14:51:49.037504
time for one epoch: 0:00:02.013931
epoch 11
time: 2022-10-02 14:51:49.038147
epoch 11 finished
time: 2022-10-02 14:51:49.708357
time for one epoch: 0:00:00.670210
epoch 12
time: 2022-10-02 14:51:49.708414
epoch 12 finished
time: 2022-10-02 14:51:50.342309
time for one epoch: 0:00:00.633895
epoch 13
time: 2022-10-02 14:51:50.342346
epoch 13 finished
time: 2022-10-02 14:51:51.120483
time for one epoch: 0:00:00.778137
epoch 14
time: 2022-10-02 14:51:51.120527
epoch 14 finished
time: 2022-10-02 14:51:51.818518
time for one epoch: 0:00:00.697991
epoch 15
time: 2022-10-02 14:51:51.818555
saving model for epoch 15
epoch 15 finished
time: 2022-10-02 14:51:53.887909
time for one epoch: 0:00:02.069354
epoch 16
time: 2022-10-02 14:51:53.887966
epoch 16 finished
time: 2022-10-02 14:51:54.638956
time for one epoch: 0:00:00.750990
epoch 17
time: 2022-10-02 14:51:54.638997
epoch 17 finished
time: 2022-10-02 14:51:55.352443
time for one epoch: 0:00:00.713446
epoch 18
time: 2022-10-02 14:51:55.352490
epoch 18 finished
time: 2022-10-02 14:51:56.041411
time for one epoch: 0:00:00.688921
epoch 19
time: 2022-10-02 14:51:56.041461
epoch 19 finished
time: 2022-10-02 14:51:56.683434
time for one epoch: 0:00:00.641973
epoch 20
time: 2022-10-02 14:51:56.683474
saving model for epoch 20
epoch 20 finished
time: 2022-10-02 14:51:58.132586
time for one epoch: 0:00:01.449112
epoch 21
time: 2022-10-02 14:51:58.132646
epoch 21 finished
time: 2022-10-02 14:51:58.956908
time for one epoch: 0:00:00.824262
epoch 22
time: 2022-10-02 14:51:58.956947
epoch 22 finished
time: 2022-10-02 14:51:59.660956
time for one epoch: 0:00:00.704009
epoch 23
time: 2022-10-02 14:51:59.660999
epoch 23 finished
time: 2022-10-02 14:52:00.359305
time for one epoch: 0:00:00.698306
epoch 24
time: 2022-10-02 14:52:00.359344
epoch 24 finished
time: 2022-10-02 14:52:01.096733
time for one epoch: 0:00:00.737389
epoch 25
time: 2022-10-02 14:52:01.096784
saving model for epoch 25
epoch 25 finished
time: 2022-10-02 14:52:02.663749
time for one epoch: 0:00:01.566965
epoch 26
time: 2022-10-02 14:52:02.663813
epoch 26 finished
time: 2022-10-02 14:52:03.444945
time for one epoch: 0:00:00.781132
epoch 27
time: 2022-10-02 14:52:03.444996
epoch 27 finished
time: 2022-10-02 14:52:04.191383
time for one epoch: 0:00:00.746387
epoch 28
time: 2022-10-02 14:52:04.191432
epoch 28 finished
time: 2022-10-02 14:52:04.841133
time for one epoch: 0:00:00.649701
epoch 29
time: 2022-10-02 14:52:04.841188
epoch 29 finished
time: 2022-10-02 14:52:05.454232
time for one epoch: 0:00:00.613044
epoch 30
time: 2022-10-02 14:52:05.454272
saving model for epoch 30
epoch 30 finished
time: 2022-10-02 14:52:06.953251
time for one epoch: 0:00:01.498979
training finished

@Mitr-yuzr 催更催更 这几天已经收集到将近三万条数据了

@Mitr-yuzr 感谢

可不可以实现对指定的对象说的话进行过滤 不收集这个人发的句子
或者导出时过滤掉关键词
/NLPHelper exportBySQL "SELECT * FROM NLPH WHERE content NOT LIKE '%妈%';" gpt2
这个句子如果想添加多个过滤词该怎么写?

@yuanbolin 我发现的机制是 如果你在另外一个群里@机器人 那么bot会在所有群回应接下来的发言

等一个正则

支持良性竞争

太棒了 期待后续

用过chatleaning以后我也一直有这种想法 感觉会很有意思 奈何本人技术力不够没法实现（哭）

@Nekoer 谢谢 已解决

你好 请问这是什么情况 之前可以用 今天突然用不了了