Struct WeightTensorRecord 

pub struct WeightTensorRecord {
    pub name: String,
    pub dtype: String,
    pub shape: Vec<u64>,
    pub offset_in_shard: u64,
    pub len: u64,
    pub shard_index: u32,
    pub extra: bool,
}

One tensor’s placement inside a staged model, mirroring the safetensors header exactly: offset_in_shard is the ABSOLUTE file offset (8-byte size prefix + header + the tensor’s data-section start), len is the raw contiguous byte count (data_offsets[1] - data_offsets[0]). The client RDMA-READs exactly [shard_base + offset_in_shard .. + len).

Fields

name: String

HuggingFace tensor name — the WeightStore key, verbatim.

dtype: String

Raw safetensors dtype string ("BF16", "F8_E4M3", "I8", …). The client maps it via WeightDtype::from_safetensors_str — the same closed mapping the disk loaders use.

shape: Vec<u64>

offset_in_shard: u64

Absolute byte offset of the tensor’s first byte within its shard file.

len: u64

Tensor byte length (authoritative — do NOT recompute from shape; packed NVFP4 lengths differ from numel * byte_size).

shard_index: u32

Index into WeightManifest::shard_files.

extra: bool

True for tensors from extra_weights.safetensors (grafted MTP etc.): loaded with NO expert-skip filter, exactly like the disk loaders.

Trait Implementations

impl Clone for WeightTensorRecord

fn clone(&self) -> WeightTensorRecord

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for WeightTensorRecord

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl<'de> Deserialize<'de> for WeightTensorRecord

fn deserialize<__D>(__deserializer: __D) -> Result<Self, __D::Error>

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl PartialEq for WeightTensorRecord

fn eq(&self, other: &WeightTensorRecord) -> bool

Tests for self and other values to be equal, and is used by ==.

fn ne(&self, other: &Rhs) -> bool

Tests for !=. The default implementation is almost always sufficient, and should not be overridden without very good reason.

impl Serialize for WeightTensorRecord

fn serialize<__S>(&self, __serializer: __S) -> Result<__S::Ok, __S::Error>

where __S: Serializer,

Serialize this value into the given Serde serializer.

impl Eq for WeightTensorRecord

impl StructuralPartialEq for WeightTensorRecord