subtitles/zh-CN/33_the-push-to-hub-api-(pytorch).srt

1
00:00:00,321 --> 00:00:01,497
（空气呼啸）
(air whooshing)

2
00:00:01,497 --> 00:00:02,330
（笑脸弹出）
(smiley face popping)

3
00:00:02,330 --> 00:00:05,130
（空气呼啸）
(air whooshing)

4
00:00:05,130 --> 00:00:06,830
- [Instructor] push_to_hub API。
- [Instructor] So push_to_hub API.

5
00:00:08,310 --> 00:00:10,533
让我们看一下 push_to_hub API。
Let's have a look at the push_to_hub API.

6
00:00:11,730 --> 00:00:14,640
你需要使用你的 Hugging Face 帐户登录
You will need to be logged in with your Hugging Face account

7
00:00:14,640 --> 00:00:17,400
你可以通过执行第一个单元格里的操作来登录，
which you can do by executing this first cell,

8
00:00:17,400 --> 00:00:21,123
或者在 terminal 中输入 huggingface-cli login。
or by typing huggingface-cli login in a terminal.

9
00:00:21,990 --> 00:00:26,640
只需输入你的用户名和密码，然后点击登录，
Just enter you username and password, then click login,

10
00:00:26,640 --> 00:00:28,620
登录后将存储一个授权 token
this will store a authentication token

11
00:00:28,620 --> 00:00:30,670
到你正在使用的机器的缓存中。
in the cache of the machine you're using.

12
00:00:31,890 --> 00:00:35,790
现在，让我们基于 GLUE COLA 数据集
Now, let's launch a fine tuning of a BERT model

13
00:00:35,790 --> 00:00:37,920
对 BERT 模型进行微调。
on the GLUE COLA dataset.

14
00:00:37,920 --> 00:00:39,600
我们不会深入探讨微调代码
We won't go over the fine tuning code

15
00:00:39,600 --> 00:00:42,270
你可以在任何 transformer 教程
because you can find it in any transformer tutorial,

16
00:00:42,270 --> 00:00:44,670
或查看下面的视频链接找到相关参考。
or by looking at the videos link below.

17
00:00:44,670 --> 00:00:46,470
我们感兴趣的是
What interests us here is

18
00:00:46,470 --> 00:00:48,970
如何在训练期间利用 model hub。
how we can leverage the model hub during training.

19
00:00:49,860 --> 00:00:52,980
这是通过 “push_to_hub=true” 参数完成的
This is done with the "push_to_hub=true" argument

20
00:00:52,980 --> 00:00:55,530
将该参数添加到你的 TrainingArguments。
passed in your TrainingArguments.

21
00:00:55,530 --> 00:00:57,240
每次保存时将自动上传
This will automatically upload your model

22
00:00:57,240 --> 00:00:59,400
你的模型到 Hub，在我们的示例中，
to the Hub each time it is saved,

23
00:00:59,400 --> 00:01:01,323
每个 epoch 都会如此操作。
so every epoch in our case.

24
00:01:02,280 --> 00:01:04,860
如果当前的被打断
This allows you to resume training from a different machine

25
00:01:04,860 --> 00:01:06,873
这允许你从不同的机器恢复训练之前的训练。
if the current one gets interrupted.

26
00:01:08,220 --> 00:01:10,440
该模型将使用
The model will be updated in your namespace

27
00:01:10,440 --> 00:01:14,640
你默认选择的输出目录的名称在你的 namespace 中更新。
with the name of the output directory you picked by default.

28
00:01:14,640 --> 00:01:16,020
你可以通过将其
You can choose another name

29
00:01:16,020 --> 00:01:19,113
传递给 hub_model_id 参数选择其他名称。
by passing it to the hub_model_id argument.

30
00:01:20,070 --> 00:01:23,370
你还可以通过传递完整的仓库名称
You can also push inside an organization you are a member of

31
00:01:23,370 --> 00:01:25,740
将模型 push 到你所属的组织内部，
by passing a full repository name,

32
00:01:25,740 --> 00:01:28,933
使用 organization/ 的形式,
with the name of the organization/,

33
00:01:28,933 --> 00:01:30,433
再加上你所选的 model ID。
the model ID you want to pick.

34
00:01:32,250 --> 00:01:34,650
完成后，我们就可以开始训练了，
With that done, we can just launch training,

35
00:01:34,650 --> 00:01:36,093
稍等一下。
and wait a little bit.

36
00:01:36,960 --> 00:01:39,033
视频中会跳过等待的过程。
I'll cut the waiting time from the video.

37
00:01:43,260 --> 00:01:46,350
请注意，模型是异步 push 的，
Note that the model is pushed asynchronously,

38
00:01:46,350 --> 00:01:47,730
意味着当你的模型上传到 hub 时，
meaning that the training continues

39
00:01:47,730 --> 00:01:49,730
训练将继续进行。
while your model is uploaded to the hub.

40
00:01:51,060 --> 00:01:52,950
当你的第一次 commit 完成时，
When your first commit is finished,

41
00:01:52,950 --> 00:01:55,650
你可以通过查看你的 namespace 
you can go inspect your model on the Hub

42
00:01:55,650 --> 00:01:57,960
去 Hub 检查你的模型，
by looking inside your namespace,

43
00:01:57,960 --> 00:01:59,943
你会在最上面找到它。
and you'll find it at the very top.

44
00:02:01,980 --> 00:02:04,200
你甚至可以在继续训练的同时
You can even start playing with its inference widget

45
00:02:04,200 --> 00:02:06,630
开始使用它的 inference 小部件。
while it's continuing the training.

46
00:02:06,630 --> 00:02:09,270
Cola 数据集让模型确定
The Cola dataset tasks the model with determining

47
00:02:09,270 --> 00:02:11,970
句子在语法上是否是正确的。
if the sentence is grammatically correct on that.

48
00:02:11,970 --> 00:02:15,510
所以我们选择一个错误句子的例子来测试它。
So we pick an example of incorrect sentence to test it.

49
00:02:15,510 --> 00:02:16,950
请注意，第一次尝试使用它时，
Note that it'll take a bit of time

50
00:02:16,950 --> 00:02:18,750
这需要一些时间才能在 inference API 中
to load your model inside the inference APIs,

51
00:02:18,750 --> 00:02:20,880
完成模型加载。
so first time you try to use it.

52
00:02:20,880 --> 00:02:23,280
我们将根据时间从视频中删掉。
We'll cut by time from the video.

53
00:02:23,280 --> 00:02:24,870
标签有点问题，
There is something wrong with the labels,

54
00:02:24,870 --> 00:02:27,360
我们稍后会在本视频中修复它。
but we'll fix it later in this video.

55
00:02:27,360 --> 00:02:29,520
一旦你的训练完成，
Once your training is finished,

56
00:02:29,520 --> 00:02:31,770
你应该使用 trainer.push_to_hub 方法
you should do one last push with the trainer

57
00:02:31,770 --> 00:02:33,840
最后再提交一次。
that pushed to a method.

58
00:02:33,840 --> 00:02:35,430
这其中有两个原因。
This is for two reason.

59
00:02:35,430 --> 00:02:36,750
首先，若你尚未完成
First, this will make sure

60
00:02:36,750 --> 00:02:39,180
这将确保你正在预测模型的
you are predicting the final version of your model

61
00:02:39,180 --> 00:02:40,680
最终版本。
if you didn't already.

62
00:02:40,680 --> 00:02:42,480
例如，如果你曾经是在每一步保存
For instance, if you used to save

63
00:02:42,480 --> 00:02:46,980
而不是每秒保存，
every in step strategy instead of every second,

64
00:02:46,980 --> 00:02:48,180
这将创建一个 model card
this will draft a model card

65
00:02:48,180 --> 00:02:51,120
那将是你的 model repo 的最初始页面。
that will be the landing page of your model repo.

66
00:02:51,120 --> 00:02:52,260
commit 完成后，
Once the commit is done,

67
00:02:52,260 --> 00:02:54,810
让我们回到我们的 model 页面并刷新。
let's go back on our model page and refresh.

68
00:02:54,810 --> 00:02:56,820
我们可以看到 model card 的草稿
We can see the drafters model card

69
00:02:56,820 --> 00:02:58,080
其中包括信息，
which includes information,

70
00:02:58,080 --> 00:03:00,381
以及哪个模型被调整过。
and which one model we find tuned.

71
00:03:00,381 --> 00:03:03,570
接下来是最终的评估 loss 和 metric，
So final evaluation loss and metric,

72
00:03:03,570 --> 00:03:06,300
使用过的训练超参数，
the training hyperparameter used,

73
00:03:06,300 --> 00:03:08,670
中间训练结果，
the intermediate training results,

74
00:03:08,670 --> 00:03:10,320
以及我们使用的框架版本
and the framework versions we used

75
00:03:10,320 --> 00:03:13,173
以便其他人可以轻松地重现我们的结果。
so that other people can easily reproduce our results.

76
00:03:15,270 --> 00:03:16,860
在所有这些信息之上，
On top of all that information,

77
00:03:16,860 --> 00:03:19,740
trainer 还包括一些 metadata，它可以
the trainer also included some metadata that is interpreted

78
00:03:19,740 --> 00:03:22,650
通过 model cloud 上的 HuggingFace 网站解析。
by the Hugging Face website in the model cloud.

79
00:03:22,650 --> 00:03:26,010
你将会得到一个漂亮的 widget 所返回的相关指标数值
You get the value of the metrics reported in a nice widget

80
00:03:26,010 --> 00:03:29,640
以及一个链接指向 leaderboard(Paper with Code)。
as well as a link to a leaderboard with paper with code.

81
00:03:29,640 --> 00:03:32,550
并且 Tensorboard 的运行结果也包含
So the Tensorboard runs have also been pushed

82
00:03:32,550 --> 00:03:34,560
在这份报告中，我们可以在 Model Hub 中
to this report, and we can look at them

83
00:03:34,560 --> 00:03:36,000
通过点击子菜单中的 
directly from the model hub

84
00:03:36,000 --> 00:03:38,850
Training metrics 查看报告。
by clicking on the training metrics sub menu.

85
00:03:38,850 --> 00:03:39,795
如果你不使用 Trainer API
If you are not using the Trainer API

86
00:03:39,795 --> 00:03:42,510
微调你的模型，
to fine-tune your model,

87
00:03:42,510 --> 00:03:43,770
你可以在模型上直接
you can use a push_to_hub method

88
00:03:43,770 --> 00:03:46,427
使用 push_to_hub 方法和分词器。
on the model, and tokenizer directly.

89
00:03:46,427 --> 00:03:50,160
让我们测试一下以修复 inference widget 中的所有标签。
Let's test this to fix all labels in the inference widget.

90
00:03:50,160 --> 00:03:52,740
inference widget 使用不同的标签名称
The inference widget was using different names for labels

91
00:03:52,740 --> 00:03:54,810
因为我们没有在整数和标签名称之间
because we did not indicate the correspondence

92
00:03:54,810 --> 00:03:57,030
注明关联性。
between integer and label names.

93
00:03:57,030 --> 00:03:58,740
当推送模型配置到 hub 时，  
We can fix this in the configuration

94
00:03:58,740 --> 00:04:01,350
我们可以通过将 label2id 
by setting the label2id,

95
00:04:01,350 --> 00:04:04,170
和 id2label 字段设置为合适的值
and id2label fields through the proper values

96
00:04:04,170 --> 00:04:06,933
在配置中解决这个问题。
when pushing the model config to the hub.

97
00:04:07,950 --> 00:04:10,620
完成后，我们可以在网站上查看，
Once this is done, we can check on the website,

98
00:04:10,620 --> 00:04:13,380
模型现在显示正确的标签。
and the model is now showing the proper label.

99
00:04:13,380 --> 00:04:15,240
现在模型在 hub 上，
Now that the model is on the hub,

100
00:04:15,240 --> 00:04:17,370
我们可以在任何地方使用它
we can use it from anywhere

101
00:04:17,370 --> 00:04:19,920
就像我们对任何其他 Transformer 模型
as we would any other Transformer model

102
00:04:19,920 --> 00:04:21,113
使用 from_pretrained 方法
with the from_pretrained method

103
00:04:21,113 --> 00:04:22,923
或者使用 pipeline 函数。
or with the pipeline function.

104
00:04:34,350 --> 00:04:36,780
我们只需要使用 hub 的标识符，
We just have to use the identifier from the hub,

105
00:04:36,780 --> 00:04:39,450
我们可以看到模型配置和权重
and we can see that the model configuration and weights

106
00:04:39,450 --> 00:04:42,483
以及分词处理后的文件会自动下载。
as well as the tokenized files are automatically downloaded.

107
00:04:53,880 --> 00:04:55,950
在下一次训练中尝试 push_to_hub API
Try the push_to_hub API in the next training

108
00:04:55,950 --> 00:04:58,650
轻松与世界其他地方分享你的模型。
to easily share your model with the rest of the world.

109
00:05:01,151 --> 00:05:03,818
（空气呼啸）
(air whooshing)