Method forward

src/model.cpp:874–883

}

void Model::forward(InferenceState& s, int token, int pos, InferenceMode mode) {
  if (s.device() != _device) {
    std::cerr << "FATAL: inference state device mismatch" << std::endl;
    assert(false);
    return;
  }
  if (_device == Device::CPU) {
    _forward_cpu(s, token, pos, mode);
  }
}

double Model::active_bytes(size_t pos) const {
  double bytes_per_weight = bits_per_weight(config->weight_quant, config->block_size[0] * config->block_size[1]) / 8.0;