1
00:00:00,800 --> 00:00:05,518
Before a model can help with an operational task, its training examples need careful review.

2
00:00:05,980 --> 00:00:09,340
Relay Lab brings that preparation into one workflow.

3
00:00:10,480 --> 00:00:14,860
Here, a manufacturing data lead is building a focused calculation set.

4
00:00:15,660 --> 00:00:20,040
Each question gives hourly production, working hours, and the rejection rate.

5
00:00:20,800 --> 00:00:23,440
The answer must state how many pieces are accepted.

6
00:00:24,520 --> 00:00:25,780
The request is precise.

7
00:00:26,280 --> 00:00:29,160
Include every quantity and return a whole number.

8
00:00:30,000 --> 00:00:36,020
Choose Distill, select the question and answer models, then set the example count and spending cap.

9
00:00:36,780 --> 00:00:39,820
This experiment requests 16 examples.

10
00:00:41,360 --> 00:00:42,642
Generate starts the work.

11
00:00:42,760 --> 00:00:46,760
You can follow the new pairs, progress, and spend reported by the application.

12
00:00:47,400 --> 00:00:49,360
The waiting time is shortened here.

13
00:00:51,460 --> 00:00:52,840
Now, open the output.

14
00:00:53,720 --> 00:00:58,980
In this example, 200 pieces an hour over eight hours produces 1,600 pieces.

15
00:00:59,480 --> 00:01:01,140
18% are rejected.

16
00:01:01,680 --> 00:01:07,180
The answer is 1,312 accepted pieces, matching an independent calculation.

17
00:01:10,100 --> 00:01:12,880
A second example changes the rate and rejection level.

18
00:01:13,580 --> 00:01:18,080
All 16 answers in this small set were checked before the dataset moved forward.

19
00:01:19,840 --> 00:01:21,320
Select Use in ModelForge.

20
00:01:22,440 --> 00:01:27,020
The same dataset name appears in the training form with 16 rows and three fields.

21
00:01:27,680 --> 00:01:30,603
The reviewed examples are now selected for the next step.

22
00:01:31,940 --> 00:01:33,180
Define the task.

23
00:01:33,780 --> 00:01:34,940
Create the examples.

24
00:01:35,560 --> 00:01:36,680
Inspect the result.

25
00:01:37,200 --> 00:01:38,636
Then continue into training.

26
00:01:39,240 --> 00:01:42,680
Start with a task your team can verify in Relay by Axeron AI.