Thanks. In theory this is the model's fault that's not learning "Zero to One" should be considered as a whole book. One limitation I mentioned in my root commen

Thanks. In theory this is the model's fault that's not learning "Zero to One" should be considered as a whole book. One limitation I mentioned in my root comment. Should be fixable with more training data!