mistralrs-core 0.8.1

Fast, flexible LLM inference.
Documentation 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55

// https://github.com/huggingface/transformers/blob/f2c388e3f946862f657acc1e21b272ec946fc66c/src/transformers/models/qwen2_vl/configuration_qwen2_vl.py

use mistralrs_quant::QuantizedConfig;

use crate::layers::Activation;

use crate::serde_default_fn;

serde_default_fn!(Activation, default_vision_hidden_act, Activation::QuickGelu);
serde_default_fn!(usize, default_in_channels, 3);

#[derive(Debug, Clone, serde::Deserialize)]
pub struct VisionConfig {
    pub depth: usize,
    pub embed_dim: usize,
    pub hidden_size: usize,
    #[serde(default = "default_vision_hidden_act")]
    pub hidden_act: Activation,
    pub mlp_ratio: f64,
    pub num_heads: usize,
    #[serde(default = "default_in_channels")]
    pub in_channels: usize,
    pub patch_size: usize,
    pub spatial_merge_size: usize,
    pub temporal_patch_size: usize,
}

#[derive(Debug, Clone, serde::Deserialize)]
pub struct MRopeScaling {
    pub mrope_section: Vec<usize>,
}

#[derive(Debug, Clone, serde::Deserialize)]
pub struct Config {
    pub vocab_size: usize,
    pub hidden_size: usize,
    pub intermediate_size: usize,
    pub num_hidden_layers: usize,
    pub num_attention_heads: usize,
    pub num_key_value_heads: usize,
    pub hidden_act: Activation,
    pub max_position_embeddings: usize,
    pub rms_norm_eps: f64,
    pub tie_word_embeddings: bool,
    pub rope_theta: f64,
    pub use_sliding_window: bool,
    pub sliding_window: Option<usize>,
    pub vision_config: VisionConfig,
    pub rope_scaling: MRopeScaling,
    pub quantization_config: Option<QuantizedConfig>,
    pub image_token_id: u32,
    pub video_token_id: u32,
    // pub vision_start_token_id: usize,
    // pub max_window_layers: usize,
}