Expand description
Hardware-independent checkpoint inspection and preparation requirements.
Adapters assign meaning to checkpoint names. Consumers use typed roles, encodings, and data references; display names never select execution behavior.
Modules§
- format 🔒
- index
- Persistent model references and explicit ownership of prepared/source stores.
- inspect 🔒
- preparation 🔒
- qwen 🔒
- Checkpoint names are interpreted only at the architecture boundary.
Structs§
- Checkpoint
- A checkpoint description and the immutable objects backing its tensor views. Pass the packed directory explicitly to inspect prepared data.
- Checkpoint
Format - Container layout and the conventions used to interpret tensor names and encodings.
- Data
Span - A byte range within one object; offsets are independent of tensor encoding.
- Mapped
Bytes - A byte range that keeps its backing mapping alive after the source is dropped. The backing file must remain unchanged for the lifetime of every clone.
- Mapped
Objects - Read-only file mappings, numbered in the order passed to
Self::open. - Model
Description - Architecture-adapted metadata and storage views, independent of execution hardware.
- Ngram
Table - A table’s hashing semantics are separate from its row encoding and placement.
- Object
Id - Identifies an object within an opened checkpoint. It is not a filename or a content digest; a store can assign content identities while copying it.
- Object
Info - An object’s identity and total size within a byte source.
- Stored
Tensor - Byte strides also describe interleaved records without introducing a table-specific layout or treating packed 4-bit values as addressable bytes.
- Table
Shard - A contiguous logical row range backed by a tensor.
- Tensor
- A logical tensor and the physical storage used to represent its values.
- Tensor
Id - Index of a tensor in its containing inventory or model description.
- Tensor
Type - Logical values are distinct from the physical types used to encode them.
Enums§
- Affine
Offset - Per-group offset used to reconstruct affine-quantized values.
- Architecture
- Model architecture recognized by the loader.
- BitPacking
- Placement of quantized codes within addressable storage words.
- Checkpoint
Conventions - Recognized naming and encoding conventions within a container.
- Compatibility
Issue - A representation constraint that prevents preparation by the current engine.
- Container
Format - On-disk container family, including Cherenkov’s prepared format.
- Dtype
- Scalar types supported by physical tensor views.
- Ggml
Encoding - Recognized GGML element and block encodings; unknown numeric codes are preserved.
- Ngram
Hash - Hashing scheme and parameters, separate from the table’s stored representation.
- Preparation
- Representation requirements for loading with the current Qwen4-exp engine.
- Preparation
Step - A transformation needed to produce the engine’s prepared representation.
- Tensor
Encoding - Storage layout and quantization metadata, with explicit references to all data.
- Tensor
Role - An architecture-assigned tensor role, independent of its checkpoint name.
Traits§
- Byte
Source - Object IDs belong to one source; they are neither paths nor content hashes. A copy/import operation must rebind them to its destination’s identities.