Beginning Anomaly Detection Using

UHWXUQ

)ORJBVRIWPD[ [GLP 

Figure B-44.  The forward function in the TCN class

The code for the model should look like Figure 

B-45

.

appendix B   intro to pytorch

---

401

Figure B-45.  The code from Figures 

B-43

 and 

B-44

 in a Jupyter cell. This defines 

the entire model

Now you can define your training and testing functions (Figure 

B-46

, Figure 

B-48

, 

and Figure 

B-49

).

appendix B   intro to pytorch

---

402

GHI

WUDLQ PRGHOGHYLFH[BWUDLQ\BWUDLQFULWHULRQRSWLPL]HU

HSRFKVDYHBGLU 7&1B&UHGLW&DUGB3\7RUFKFNSW 

WRWDOBVWHS OHQ [BWUDLQ

[BWUDLQ WRUFK7HQVRU [BWUDLQ FXGD IORDW

\BWUDLQ WRUFK7HQVRU \BWUDLQ FXGD ORQJ

[BWUDLQWR GHYLFH

\BWUDLQWR GHYLFH

)RUZDUGSDVV

RXWSXWV PRGHO [BWUDLQ

ORVV FULWHULRQ RXWSXWV\BWUDLQVTXHH]H 

%DFNZDUGDQGRSWLPL]H

RSWLPL]HU]HURBJUDG

ORVVEDFNZDUG

RSWLPL]HUVWHS

SULQW

(SRFK^`^`/RVV^I` IRUPDW HSRFK

QXPBHSRFKVORVVLWHP

WRUFKVDYH PRGHOVWDWHBGLFW VDYHBGLU

Figure B-46.  The training function. Since you don’t have data loaders, you pass  

in x_train and y_train directly into the GPU after converting them to tensors.  

The inputs then pass through, and the gradients are calculated.

appendix B   intro to pytorch

---

403

Figure B-47.  The code from Figure 

B-46

 in a Jupyter cell

appendix B   intro to pytorch

---

404