Tensor Profile: Payload Frame
The payload frame is where tensor carries the actual numeric bytes. The descriptor and body explain how to interpret the bytes; the payload frame is the data itself.
What this layer answers
- How much raw tensor data is present in the current chunk.
- How that chunk maps back to
offsetandlength. - Whether the receiver should process one chunk, reassemble multiple chunks, or wait for a full tensor.
Typical representation
| Name | Typical shape | Description |
|---|---|---|
| tensor_chunk | Binary block | Can hold a full tensor, a region chunk, or another partial numeric segment |
Engineering notes
- The payload frame does not self-describe shape, layout, or dtype; those still come from schema/body.
- You cannot infer complete, partial, or degraded semantics from the payload frame alone; fixed metadata and descriptor still matter.
- When a transport or implementation splits the tensor into chunks, receivers should reassemble with
offsetandlengthrather than assuming full-tensor delivery every time.