float16/complex32?

[charliez0]

Please make sure that this is a feature request. As per our GitHub Policy, we only address code/doc bugs, performance issues, feature requests and build/installation issues on GitHub. tag:feature_template

System information

• TensorFlow.js version (you are using): 4.22.0
• Are you willing to contribute it (Yes/No): Yes

Describe the feature and the current behavior/state.

Will this change the current api? How?

Who will benefit with this feature?

Any Other info.

[shmishra99]

Hi @charliez0 ,

Could you please elaborate on the feature and how it will benefit the community?

Thank You!!

[charliez0]

I'm using tfjs-node-gpu, which supports fp16 fp8 etc, but it seems that there is no ways to initialize tensors with that spec.

Also, is there any support to use fp64 and complex128?

[shmishra99]

Hi @charliez0 ,

We can define the dtype in @tensorflow/tfjs-node-gpu while initializing tensors. Currently, it supports 'float32', 'int32', 'bool', 'complex64', and 'string' data types.

const tf = require('@tensorflow/tfjs-node-gpu');
const floatTensor = tf.tensor2d([[1.0, 2.0], [3.0, 4.0]], [2, 2], 'float32');
console.log(floatTensor);

Let me know if i missed anything. Thank You!!

[charliez0]

missing float64,complex128,float16,bfloat16,int16,int8 etc