Closed carlos-r-l-rodrigues closed 6 years ago
Indeed, thank you for noticing it ! Corrected.
The model should be like this:
const agentConfig = { model: model, agentConfig: { memorySize: 1000, batchSize: 10, temporalWindow: temporalWindow } };
Also, the academy.step should be:
await academy.step([ { teacherName: teacher, agentsInput: inputs } ])
I'm also debugging another issue with tensorflow.js but I haven't figured out why it's happen. Error: Input 0 is incompatible with layer dense_Dense1: expected axis NaN of input shape to have value
Thank you once more, I forgot the second part of the correction...
Can you provide an example to reproduce your bug ?
Please ignore the error, it was caused by another third part code overriding the Object.prototype.clone. At some point it was changing the type of the objects used by ts.js and failing in the assert.
Thanks for this awesome framework 👍
It seems that the model must goes inside the Agent config:
model.compile({loss: 'categoricalCrossentropy', optimizer: 'sgd'}) const agentConfig = { model: model, memorySize: 5000, batchSize: 128, temporalWindow: temporalWindow };
await academy.step([ {teacherName: teacher, inputs: inputs} ]);
The parameters in the interface AcademyStepInput are "teacherName" and "agentsInput".