Documentation

reference

Training Parameters and Value Ranges

Table 1 describes the names, types, value ranges, default values, and descriptions of training parameters of the reference model.

Table 1 Training parameters (wafer_mindspore/reference/reference_train.py)

Parameter

Type

Value Range

Default Value

Description

--rgb_dir

String

-

./xxx/

Path that stores RGB images.

--batch_size

Integer

[1, 128]

16

Training batch size.

--epochs

Integer

[1, 100]

50

Number of training epochs.

--train_output_path

String

-

'./ckpt'

Path for saving the checkpoint.

--chip_id

String

-

""

Processor ID.

--setupid

String

-

""

Processor layer.

--klarf_file

String

-

"xxx"

Name of a klarf file, which must be stored in rgb_dir.

--gray_file

String

-

"xxx"

Name of a grayscale image file, which must be stored in rgb_dir.

--device_num

Integer

1,2,4,8

1

Number of devices used for training.

--device_id

Integer

[0, 7]

0

Device ID for single-device training.

--rank_table_file

String

-

None

rank_table_file path used for multi-device training.

A new_image folder must be created in rgb_dir to store the images to be registered.

klarf_file must be a TXT file, which describes information about images photographed at a machine and must comply with the following rules:

  • Each line corresponds to an image.
  • Each line contains 12 fields.
  • Fields are separated by a space.

For details about the fields, see Table 2. The following is a compilation example.

Create a data.txt file, which contains the following three lines of character strings:

img_1.jpg 001 M1 11224.435 12925.229 0 wafer_id1 01-08-22-13:09:22; device_id1 L21510106 model1 20
img_2.jpg 001 M1 13817.899 17822.569 0 wafer_id2 12-29-21-08:17:05; device_id1 L21510104 model2 20
img_3.jpg 001 M1 970.857 24809.092 0 wafer_id3 12-21-21-06:02:08; device_id3 L21510206 model3 20
Table 2 Field description

Index

Type

Value Range

Description

0

String

-

Image path

1

String

-

Processor ID.

2

String

-

Processor layer.

3

Float

[0, 100000]

X coordinate of an image

4

Float

[0, 100000]

Y coordinate of an image

5

Integer

0 or 2

Image label

6

String

-

Wafer ID

7

String

-

Timestamp

8

String

-

Device ID

9

String

-

Batch number

10

String

-

Device model

11

Integer

[1, 1000]

Magnification ratio

Training Command Reference

The command for training the reference model is as follows:

python3 reference_train.py --rgb_dir=<rgb_dir>  --epochs=31 --chip_id=<chip_id>  --setupid=<setupid> --gray_file= <gray_file>  --klarf_file=<klarf_file>
  • <rgb_dir>: path that stores RGB images to be inspected.
  • <chip_id>: processor ID, indicating the type of the processor dataset to be trained.
  • <setupid>: processor layer, indicating the specific processor diagram of a layer. This parameter and <chip_id> specify the training dataset.
  • <gray_file>: grayscale image name.
  • <klarf_file>: name of the klarf file.

The model training process is random, and the accuracy of different processor layers is different. The final evaluation accuracy prevails. The training accuracy is displayed as follows:

Figure 1 Training accuracy

After the training is complete, a .ckpt model file is generated in the output directory specified by --train_output_path.

Evaluation Parameters and Value Ranges

Table 3 describes the names, types, value ranges, default values, and descriptions of evaluation parameters of the reference model.

Table 3 Evaluation parameters (wafer_mindspore/reference/eval.py)

Parameter

Type

Value Range

Default Value

Description

--root_path

String

-

"./xxx/yyy/zzz"

Path for saving the generated evaluation file.

--gray_image_path

String

-

""

Path for saving the grayscale images.

--mask_image_path

String

-

""

Path for saving the mask images.

--overlap

Float

[1,2]

1.2

Overlapping rate of two images.

--chip_id

String

-

""

Processor ID.

--setupid

String

-

""

Processor layer.

--model

String

-

""

Path for saving the CKPT model.

--gray_ref_scale

Integer

[1,1000]

248

Cropping size of the grayscale image.

--rgb_ref_scale

Integer

[1,1000]

366

Cropping size of the RGB reference image.

--device_target

String

"Ascend" or "GPU"

"Ascend"

Evaluation platform.

Evaluation Command Reference

The command for evaluating the reference model is as follows:

python3 eval.py --root_path=<root_path> --gray_image_path <gray_image_path> --mask_image_path <mask_image_path>  --chip_id <chip_id> --setupid=<setupid>  --model <model_path>
  • <root_path>: path for saving the generated evaluation file.
  • <gray_image_path>: path for saving the grayscale images.
  • <mask_image_path>: path for saving the mask images.
  • <chip_id>: processor list, indicating the type of the processor dataset to be trained.
  • <setupid>: processor layer, indicating the specific processor diagram of a layer. This parameter and <chip_id> specify the training dataset.
  • <model_path>: path for storing the CKPT file generated during training.

The CKPT output during training is used to evaluate the model accuracy. The following is an example.

Figure 2 Model accuracy evaluation

Generate RGB images in the directory specified by the --root_path parameter. gray_ref stores the grayscale images, mask_ref stores the mask images, and rgb_ref stores the color images.

Figure 3 Generating RGB images