public GpuQuantizedMatrix(reader as SafetensorsReader, name as String, device as Device, buffer as Byte[] = null)
Loads through a shared descriptor reader, owned and closed by the caller. The reader is seeked and must not be shared concurrently.
| Parameter | Description |
|---|---|
reader | Validated open shard. |
name | Tensor name. |
device | GPU selector. |
buffer | Optional bounded upload block shared by a checkpoint loader. |
matrix = GpuQuantizedMatrix(reader, "projection.weight", Device.gpu[0])