Enum CandleError 

pub enum CandleError {
    EmptyBuffer {
        artifact: &'static str,
    },
    Configuration(String),
    InvalidConfigurationValue {
        field: &'static str,
        reason: String,
    },
    TokenizerLoading(String),
    UnsupportedModelFamily(String),
    ModelFamilyMismatch {
        selected: ConversationProtocol,
        detected: ConversationProtocol,
    },
    ArtifactMismatch {
        artifact: &'static str,
        reason: String,
    },
    TokenizerVocabularyMismatch {
        expected: usize,
        actual: usize,
    },
    MissingSpecialToken {
        token: &'static str,
    },
    SpecialTokenNotMarked {
        token: &'static str,
    },
    TokenIdOutOfRange {
        token: String,
        id: u32,
        vocab_size: usize,
    },
    TokenizerEncoding(String),
    TokenizerDecoding(String),
    InvalidCheckpoint(String),
    InvalidQuantizedCheckpoint(String),
    UnsupportedQuantization(String),
    UnsupportedPromptContent(&'static str),
    ReservedProtocolMarker {
        field: &'static str,
        marker: &'static str,
    },
    MissingTensor(String),
    TensorShapeMismatch {
        tensor: String,
        expected: Vec<usize>,
        actual: Vec<usize>,
    },
    UnsupportedTensorDtype {
        tensor: String,
        dtype: String,
    },
    MissingStopToken,
    ModelLoading(String),
    Inference(String),
    InvalidGeneration(String),
    PromptTooLong {
        prompt_tokens: usize,
        context_limit: usize,
    },
    NoGenerationCapacity {
        prompt_tokens: usize,
        context_limit: usize,
    },
    NumericConversion {
        field: &'static str,
        value: u64,
    },
    InvalidConcurrencyLimit,
    ConcurrencyControllerClosed,
    Cancelled,
    UnsupportedFeature(String),
    InvalidToolDefinition {
        tool: String,
        reason: String,
    },
    MalformedToolCall(String),
    ToolChoiceViolation(String),
    UnmatchedToolResult {
        result_id: String,
    },
    UnsupportedMake,
    BlockingTaskJoin(String),
    StreamingChannelClosed,
}

Available on crate feature candle only.

Why a local Candle completion failed.

Variants

EmptyBuffer

A required artifact buffer was empty.

Fields

artifact: &'static str

Configuration(String)

The Hugging Face configuration could not be parsed or is internally invalid.

InvalidConfigurationValue

A parsed configuration field is incompatible with Candle’s Llama implementation.

Fields

field: &'static str

Configuration field name.

reason: String

Explanation of the invalid value or relationship.

TokenizerLoading(String)

The tokenizer bytes could not be loaded.

UnsupportedModelFamily(String)

The tokenizer metadata does not identify a supported prompt family.

ModelFamilyMismatch

An explicitly selected model family disagrees with validated artifacts.

Fields

selected: ConversationProtocol

Family requested by the caller.

detected: ConversationProtocol

Family detected from the tokenizer.

ArtifactMismatch

Independently supplied model artifacts disagree with one another.

Fields

artifact: &'static str

Artifact or metadata field that disagreed.

reason: String

Human-readable mismatch details.

TokenizerVocabularyMismatch

The tokenizer vocabulary does not agree with the model configuration.

Fields

expected: usize

Vocabulary size required by the model configuration.

actual: usize

Vocabulary size reported by the tokenizer, including added tokens.

MissingSpecialToken

A selected prompt-format token is absent from the tokenizer.

Fields

token: &'static str

Required special-token string.

SpecialTokenNotMarked

A prompt-format token exists but is not registered as special.

Fields

token: &'static str

Formatting token that must be treated atomically by the tokenizer.

TokenIdOutOfRange

A configured or tokenizer-provided token ID lies outside the model vocabulary.

Fields

token: String

Configuration field or special-token name.

id: u32

Invalid token ID.

vocab_size: usize

Model vocabulary size.

TokenizerEncoding(String)

The tokenizer failed to encode the rendered prompt.

TokenizerDecoding(String)

The tokenizer failed to decode generated token IDs.

InvalidCheckpoint(String)

The safetensors checkpoint was malformed or incompatible with the configuration.

InvalidQuantizedCheckpoint(String)

A GGUF checkpoint was malformed or inconsistent with its configuration.

UnsupportedQuantization(String)

The GGUF checkpoint does not use a supported production quantization.

UnsupportedPromptContent(&'static str)

A message contains content that the selected text-only prompt renderer cannot represent.

ReservedProtocolMarker

Caller-controlled content contains a delimiter reserved by the selected chat template.

Fields

field: &'static str

Kind of prompt content containing the marker.

marker: &'static str

Structural delimiter that was rejected.

MissingTensor(String)

A tensor required by the configured architecture was absent.

TensorShapeMismatch

A checkpoint tensor has an incompatible shape.

Fields

tensor: String

Tensor name in the checkpoint.

expected: Vec<usize>

Shape required by Candle’s implementation.

actual: Vec<usize>

Shape stored in the checkpoint.

UnsupportedTensorDtype

A checkpoint tensor uses a dtype outside the portable CPU scope.

Fields

tensor: String

Tensor name in the checkpoint.

dtype: String

Unsupported safetensors dtype.

MissingStopToken

Neither the configuration nor tokenizer supplied a usable stop token.

ModelLoading(String)

Candle could not load the model tensors.

Inference(String)

Candle inference failed.

InvalidGeneration(String)

A generation setting was invalid.

PromptTooLong

The encoded prompt itself exceeds the model context window.

Fields

prompt_tokens: usize

Number of prompt tokens.

context_limit: usize

Configured model context limit.

NoGenerationCapacity

The prompt fills the context window, leaving no capacity for generation.

Fields

prompt_tokens: usize

Number of prompt tokens.

context_limit: usize

Configured model context limit.

NumericConversion

A numeric generation value cannot be represented on the current platform.

Fields

field: &'static str

Name of the value being converted.

value: u64

Original portable value.

InvalidConcurrencyLimit

The native inference concurrency limit is invalid.

ConcurrencyControllerClosed

Available on non-target_family=wasm only.

The native inference admission controller was closed unexpectedly.

Cancelled

Local inference was cooperatively cancelled.

UnsupportedFeature(String)

The request uses a feature outside the selected local-model protocol.

InvalidToolDefinition

A tool definition cannot be represented safely by the selected protocol.

Fields

tool: String

Tool name, or a placeholder when the name itself is invalid.

reason: String

Validation failure.

MalformedToolCall(String)

A model-generated tool call was malformed or incomplete.

ToolChoiceViolation(String)

The generated response violated the requested tool-choice policy.

UnmatchedToolResult

A tool result could not be correlated with a preceding assistant call.

Fields

result_id: String

Result/call identifier supplied in history.

UnsupportedMake

CompletionModel::make cannot load a byte-backed model.

BlockingTaskJoin(String)

Available on non-target_family=wasm only.

A native blocking inference task could not be joined.

StreamingChannelClosed

Available on non-target_family=wasm only.

The consumer of a native streaming response closed its bounded channel.

Trait Implementations

impl Clone for CandleError

fn clone(&self) -> CandleError

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for CandleError

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

Formats the value using the given formatter.

impl Display for CandleError

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

Formats the value using the given formatter.

impl Error for CandleError

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 From<CandleError> for CompletionError

fn from(error: CandleError) -> CompletionError

Converts to this type from the input type.