Fix README documentation. Remove unnecessary LIBSVM check in CSV inpu…

[ericangelokim]

…t validation.

Description of changes:

• Fix README
• Remove unnecessary LIBSVM validation in CSV validation.
  • Originally, the validation was added in case customer data was formatted with a unorthodox delim such as ' 2:', causing the data to look like LIBSVM: 1 2:1 2:3 2:1 -> 1,2,3,1. However, retrospectively this should not have been added; if the customer wants to use CSV format with such delim, that should be allowed.

Test

Ran all unit tests successfully using python3 -m tox.

By submitting this pull request, I confirm that you can use, modify, copy, and redistribute this contribution, under the terms of your choice.

[aws-patlin]

Did you have a chance to see the customer data and confirm this was the root cause?

[ericangelokim]

Sent customer the image built with this change, and asked her to re-run training job to verify the wrong logs do not show up.

[ericangelokim]

I found a previous training job, whose logs displayed the LIBSVM error log:

ERROR 198.79280739592556,-0.814965842967067,0.11862443460345609,1.441393907312773,-0.3574215393650993,0.260460926398274,-1.5611072676604718,0.19408432846292067,1.8245437744650783,-0.02742789379747009,2.0488629834975347,-1.5521172040593016,1.7112930805229332,1.2401094370037329,0.485322995515301,0.9192109852259204
does not follow LIBSVM label format <label>(:<weight>).

I reran with a custom built image and this no longer shows in logs.

[ericangelokim]

Validated with customer that this fixed the error log.