Classfication output issue

We have extensively used Clara Train and Deploy SDK for Segmentation. We are now using Train and Deploy SDK for Chest Xray classification. We are able to run successfully for inference with Chexnet dataset using Train SDK. However the results seems to be very wrong. We get classification output for different files as below
/workspace/eval# cat preds_model.csv
/workspace/data/CXR/00000001_000.png,-2.9993646,-5.467203,-8.324304,-7.3575387,-2.380995,-7.250484,-6.2774434,-6.5050864,-6.7666855,-4.047749,-0.6517485,-3.9392443,-5.0880165,-4.476193,-2.6302557
/workspace/data/CXR/00000001_002.png,-2.975049,-5.432899,-8.326101,-7.3514028,-2.3835251,-7.114484,-6.2664895,-6.5266843,-6.8599052,-4.0954847,-0.69324976,-4.037133,-5.1746097,-4.408802,-2.6353362
/workspace/data/CXR/00000002_000.png,-3.0117333,-5.5201297,-8.417332,-7.448345,-2.3047712,-7.315398,-6.3156013,-6.512086,-6.748489,-3.986781,-0.6200426,-3.9380345,-5.098372,-4.520816,-2.7370596
/workspace/data/CXR/00000091_001.png,-2.938099,-5.4107823,-8.332923,-7.3734603,-2.348722,-7.160991,-6.208785,-6.4735045,-6.806645,-4.0232644,-0.66152483,-4.0502024,-5.1563954,-4.4375114,-2.711018
/workspace/data/CXR/00000093_001.png,-2.9866529,-5.393373,-8.315289,-7.3216243,-2.4690082,-7.151875,-6.3634076,-6.5848913,-6.9066076,-4.1519027,-0.69096357,-4.052709,-5.1578555,-4.461323,-2.5067937

If you see the results for different inputs are similar. Are we missing something ? What is the expected format of output?