Enum BatchExtError 

pub enum BatchExtError {
    NullReturn,
    Full(u32),
    InvalidToken(i32),
    InvalidSeqId {
        seq_id: llama_seq_id,
        n_seq_max: u32,
    },
    EmbeddingShape {
        n_rows: usize,
        n_embd: usize,
        len: usize,
    },
    NoSuchEntry {
        entry: i32,
        len: usize,
    },
    PositionArity {
        entry: i32,
        expected: usize,
        got: usize,
    },
    EmbeddingWidthMismatch {
        expected: usize,
        got: usize,
    },
    EmbeddingAlreadySet(i32),
    StateAlreadySet(i32),
    EmbeddingRejected,
}

Errors from building a LlamaBatchExt.

llama.cpp reports most of these as a bare false with the detail in its log, so the variants carry whatever the Rust side can establish on its own.

Variants

NullReturn

llama_batch_ext_init returned null.

Full(u32)

The batch already holds n_batch entries.

InvalidToken(i32)

The token id is negative or not below the model’s vocabulary size.

InvalidSeqId

The sequence id is negative or not below the context’s n_seq_max.

Fields

seq_id: llama_seq_id

The rejected sequence id.

n_seq_max: u32

The context’s n_seq_max.

EmbeddingShape

n_rows * n_embd did not match the length of the value buffer.

Fields

n_rows: usize

Rows the caller declared.

n_embd: usize

Row width the caller declared.

len: usize

Values actually supplied.

NoSuchEntry

The entry handle does not belong to this batch — it came from another batch, or from this one before LlamaBatchExt::clear.

Fields

entry: i32

The rejected entry index.

len: usize

Entries currently in the batch.

PositionArity

The wrong number of positions was supplied for the entry. A token entry takes one; an embedding entry takes LlamaBatchExt::n_pos_per_embd.

Fields

entry: i32

The entry being positioned.

expected: usize

Positions the entry requires.

got: usize

Positions supplied.

EmbeddingWidthMismatch

Every embedding in one batch must have the same total width, fixed by the first one accepted.

Fields

expected: usize

Width fixed by the first embedding in the batch.

got: usize

Width supplied.

EmbeddingAlreadySet(i32)

The entry already carries an input embedding; llama.cpp allows one.

StateAlreadySet(i32)

The entry already carries a state embedding; llama.cpp allows one.

EmbeddingRejected

llama.cpp rejected an embedding: the row width matches neither the decoder nor the encoder input width, or (for a state embedding) this context takes none — only an MTP context does. The specific reason is in the llama.cpp log.

When this comes from LlamaBatchExt::add_embedding, llama.cpp has already appended the entry by the time it validates the block, and offers no way to remove one — so the batch is left holding an entry with no input, and LlamaBatchExt::clear is the only way forward. See LlamaBatchExt::has_incomplete_entry.

Trait Implementations

impl Clone for BatchExtError

fn clone(&self) -> Self

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for BatchExtError

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

Formats the value using the given formatter.

impl Display for BatchExtError

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

Formats the value using the given formatter.

impl Eq for BatchExtError

impl Error for BatchExtError

fn source(&self) -> Option<&(dyn Error + 'static)>

Returns the lower-level source of this error, if any.

fn description(&self) -> &str

👎Deprecated since 1.42.0:

use the Display impl or to_string()

fn cause(&self) -> Option<&dyn Error>

👎Deprecated since 1.33.0:

replaced by Error::source, which can support downcasting

fn provide<'a>(&'a self, request: &mut Request<'a>)

🔬This is a nightly-only experimental API. (error_generic_member_access)

Provides type-based access to context intended for error reports.

impl PartialEq for BatchExtError

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

Equality operator ==.

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

Inequality operator !=.

impl StructuralPartialEq for BatchExtError