@cbnu167
This is a more detailed procedure using simultaneous equations (it is less rigorous than statistical regression), nonetheless it will do the job for you.
Fitting XM430W350 Torque (T) vs Current (C) to a Linear Function from Performance Graph at (XM430W350T/R).
As the T vs C graph looks “linear”, we’ll try to fit a Linear Function to it, so:
T = a * C + b Eq. (1)  where Unknown Parameters are a and b
The “curvefitting” procedure is listed below:

Find (T,C) coordinates for 3 points that are on the T/C Graph. I chose (0.84, 0.60), (2.45, 1.50) and (1.925, 1.20). We’ll use the 3rd data point (1.925, 1.20) to check our function fitting results.

To get rid of Parameter b, apply the first two Data Points to Eq. (1):
 Using (0.84, 0.60), Eq. (1)] yields 0.84 = a * 0.60 + b Eq. (2)
 Using (2.45, 1.50), Eq. (1)] yields 2.45 = a * 1.50 + b Eq. (3)

Determination of Parameter a:
 [Eq. (2) – Eq. (1)] yields 1.61 = a *0.91 therefore a = 1.769

Determination of Parameter b with Eq. (2):
 Eq. (2) is solved for b = 0.84 – (1.769 * 0.60) = 0.2214, so b = 0.2214

So, the T/C Performance Graph can be analytically represented as Eq. (4):
T = 1.769 * C – 0.2214
 Where C is in Amperes (A) and T is in NewtonMeter (Nm)

Checking the results with 3rd Data Point (1.925, 1.2):
 Using C=1.2 in Eq. (4) yields T = 1.769 * 1.2 – 0.2214 = 1.9 (pretty close to 1.925 as shown on the graph).
Application in C++/Python programming
According to the XM430W350 Control Table, at runtime, you can use DXL SDK to read in Present Current (PC126) at Address 126
(XM430W350T/R).
This PC126 number is supposed to be a signed 16bit integer between ( Goal_Current) and (+Goal_Current), to be found at Address 102. By default, this range is between (1193) and (+1193) – in decimal, and its unit value is 2.69 mA.
In ROBOTIS software/firmware, negative integers use the 2complement binary format for a 16bit integer, however Python/C++ does not know that in advance, so it would just print out the 2complement integer as is (between 0 and 65535) – to the confusion of many users.
So, as a programmer for your Python/C++ code, you need to do some checking before converting PC126 into a C value that you can use with your Eq. 4 from above:
// Assuming 16bit 2complement integers  pseudo code
if (PC126 > 32767) // when NEGATIVE VALUES for PC126
PC126 = (65536 – PC126) * (1); // this is a signed decimal integer
else // when POSITIVE VALUES for PC126
PC126 = PC126;
// Computing C in A, using abs(PC126)
C = abs(PC126) * 2.69 / 1000.0; // in A
T = 1.769 * C – 0.2214; // in Nm