Different Training Results

When I was using dls 2.5, I was able to train my model to achieve a training accuracy of 95% but after installing dls 3.0, i used the same model with the same hyperparameters and my training accuracy never goes 71%. I even tried tuning the model and hyperparameters but still 71% training accuracy. Please what do I do ?
Thanks.

Hi @Jamiepoo17,

Can you please share the project configuration to review this problem?
You can download it from DLS by clicking on download button in the model tab .
image

Thanks
Santosh

Thanks @santosh here is is the configuration
data:
dataset:
name: Breast Cancer New
samples: 277524
type: private
datasetLoadOption: full
kfold: 1
mapping:
Image:
options:
Augmentation: false
Height: ‘100’
Normalization: false
Resize: true
Scaling: 1
Width: ‘100’
height_shift_range: 0
horizontal_flip: false
pretrained: None
rotation_range: 0
shear_range: 0
vertical_flip: false
width_shift_range: 0
port: InputPort0
shape: ‘’
type: Image
Rating:
options: {}
port: OutputPort0
shape: ‘’
type: Categorical
numPorts: 1
samples:
split: 4
test: 27752
training: 222019
validation: 27752
shuffle: true
model:
connections:

  • source: Input_1
    target: VGG16_1
  • source: Dense_1
    target: Dense_2
  • source: Dense_2
    target: Output_1
  • source: VGG16_1
    target: Flatten_1
  • source: Flatten_1
    target: Dense_1
    layers:
  • args: {}
    class: Input
    name: Input_1
    x: 36
    y: 91
  • args: {}
    class: Output
    name: Output_1
    x: 861
    y: 242
  • args:
    activation: relu
    units: ‘64’
    class: Dense
    name: Dense_1
    x: 375
    y: 309
  • args:
    activation: softmax
    units: ‘2’
    class: Dense
    name: Dense_2
    x: 380
    y: 468
  • args: {}
    class: Flatten
    name: Flatten_1
    x: 374
    y: 199
  • args:
    include_top: false
    trainable: 0
    class: VGG16
    name: VGG16_1
    x: 371
    y: 94
    params:
    advance_params: true
    batch_size: 64
    is_custom_loss: false
    loss_func: categorical_crossentropy
    num_epoch: 30
    optimizer:
    name: Adam
    project: JK