how to load a tensorflow model using tensorflowjs

Question

i was building a simple demo that would load a model that i have converted a JSON and binary using tensorfjs. The model is utilizing the following architecture

model = Sequential([
    Flatten(input_shape=(28, 28)),
    Dense(128, activation='relu'),
    Dense(10, activation='softmax')
])

importing this model in the site gives me this issue

my code:

const model = await tf.loadLayersModel('model.json');

error received : Error occurred: n: An InputLayer should be passed either a batchInputShape or an inputShape. at new n (

This feels like an intercompatibility issue between tensorflow and tensorflow where the json is missing some parameters. Are there any workarounds for this? How do i solve this issue?

what ive tried

• ive tried adding input_shape to the JSON (not sure if i did it right though)
• i tried mentioning Input explicitly in the model architecture

Answer 1

Add an Input layer to define the input shape explicitly:

model = Sequential([
    Input(shape=(28, 28)),
    Flatten(),
    Dense(128, activation='relu'),
    Dense(10, activation='softmax')
])

Reconvert your model and try loading it again—it should work!

Answer 2

Use the Functional API instead of Sequential to define your model:

from tensorflow.keras.models import Model
from tensorflow.keras.layers import Input, Flatten, Dense
inputs = Input(shape=(28, 28))
x = Flatten()(inputs)
x = Dense(128, activation='relu')(x)
outputs = Dense(10, activation='softmax')(x)
model = Model(inputs=inputs, outputs=outputs)

Convert and reload the model. Ensure TensorFlow.js tools are updated and model.json includes batchInputShape.