Struct CommandRConfig 

pub struct CommandRConfig {

    pub model_name: String,
    pub vocab_size: usize,
    pub hidden_size: usize,
    pub num_attention_heads: usize,
    pub num_key_value_heads: usize,
    pub num_hidden_layers: usize,
    pub intermediate_size: usize,
    pub max_sequence_length: usize,
    pub rms_norm_eps: f32,
    pub rope_theta: f32,
    pub rope_scaling_factor: f32,
    pub attention_dropout: f32,
    pub hidden_dropout: f32,
    pub use_bias: bool,
    pub tie_word_embeddings: bool,
    pub activation_function: String,
    pub layer_norm_eps: f32,
    pub use_logit_bias: bool,
    pub logit_scale: f32,
    pub use_sliding_window: bool,
    pub sliding_window_size: usize,
    pub use_flash_attention: bool,
    pub pad_token_id: Option<usize>,
    pub bos_token_id: Option<usize>,
    pub eos_token_id: Option<usize>,
    pub model_type: String,
    pub torch_dtype: String,
    pub transformers_version: String,
}

Configuration for Command R models

Fields

model_name: String

Model name

vocab_size: usize

Vocabulary size

hidden_size: usize

Hidden size

num_attention_heads: usize

Number of attention heads

num_key_value_heads: usize

Number of key-value heads (for GQA)

num_hidden_layers: usize

Number of hidden layers

intermediate_size: usize

Intermediate size in FFN

max_sequence_length: usize

Maximum sequence length

rms_norm_eps: f32

RMS normalization epsilon

rope_theta: f32

Rope theta

rope_scaling_factor: f32

Rope scaling factor

attention_dropout: f32

Attention dropout

hidden_dropout: f32

Hidden dropout

use_bias: bool

Use bias in linear layers

tie_word_embeddings: bool

Tie word embeddings

activation_function: String

Activation function

layer_norm_eps: f32

Layer norm epsilon

use_logit_bias: bool

Use logit bias

logit_scale: f32

Logit scale

use_sliding_window: bool

Use sliding window attention

sliding_window_size: usize

Sliding window size

use_flash_attention: bool

Use flash attention

pad_token_id: Option<usize>

Pad token id

bos_token_id: Option<usize>

BOS token id

eos_token_id: Option<usize>

EOS token id

model_type: String

Model type

torch_dtype: String

Torch dtype

transformers_version: String

Transformers version

Implementations

impl CommandRConfig

pub fn tiny() -> Self

Create a tiny configuration for testing purposes Uses very small dimensions to allow fast test execution

pub fn command_r() -> Self

Create Command R base model configuration

pub fn command_r_plus() -> Self

Create Command R+ model configuration

pub fn command_r_08_2024() -> Self

Create Command R 08-2024 model configuration

pub fn command_r_plus_08_2024() -> Self

Create Command R+ 08-2024 model configuration

pub fn head_dim(&self) -> usize

Get the head dimension

pub fn kv_head_dim(&self) -> usize

Get the key-value head dimension

pub fn is_gqa(&self) -> bool

Check if grouped query attention is used

pub fn num_query_groups(&self) -> usize

Get the number of query groups

pub fn validate(&self) -> Result<(), String>

Validate configuration

Trait Implementations

impl Clone for CommandRConfig

fn clone(&self) -> CommandRConfig

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Config for CommandRConfig

fn validate(&self) -> Result<()>

Validates the configuration for correctness.

fn architecture(&self) -> &'static str

Returns the architecture name for this configuration.

impl Debug for CommandRConfig

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

Formats the value using the given formatter.

impl Default for CommandRConfig

fn default() -> Self

Returns the “default value” for a type.

impl<'de> Deserialize<'de> for CommandRConfig

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

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl Serialize for CommandRConfig

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

where __S: Serializer,

Serialize this value into the given Serde serializer.