oxidized-transformers 0.1.1

Transformers library (not functional yet)
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
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90

use serde::{Deserialize, Serialize};

use crate::error::BoxedError;
use crate::models::hf::FromHF;
use crate::models::llama::decoder::HFLlamaDecoderConfig;
use crate::models::llama::LlamaDecoder;
use crate::models::transformer::{
    TransformerCausalLM, TransformerCausalLMConfig, TransformerDecoderConfig,
};

/// Llama causal language model (Touvron et al., 2023).
///
/// See:
/// * [LLaMA: Open and Efficient Foundation Language Models](https://arxiv.org/abs/2302.13971)
/// * [Llama 2: Open Foundation and Fine-Tuned Chat Models](https://arxiv.org/abs/2307.09288)
pub struct LlamaCausalLM;

/// Llama causal language model configuration.
#[derive(Clone, Debug, Deserialize, Serialize)]
pub struct HfLlamaCausalLMConfig {
    #[serde(flatten)]
    decoder: HFLlamaDecoderConfig,
}

impl TryFrom<HfLlamaCausalLMConfig> for TransformerCausalLMConfig {
    type Error = BoxedError;

    fn try_from(config: HfLlamaCausalLMConfig) -> Result<Self, Self::Error> {
        Ok(Self::default()
            .hidden_size(config.decoder.hidden_size)
            // Input and output vocab sizes are the same.
            .n_pieces(config.decoder.vocab_size)
            .decoder(Box::new(TransformerDecoderConfig::try_from(
                config.decoder,
            )?)))
    }
}

impl FromHF for LlamaCausalLM {
    type Config = TransformerCausalLMConfig;

    type HFConfig = HfLlamaCausalLMConfig;

    type Model = TransformerCausalLM;

    fn rename_parameters() -> impl Fn(&str) -> String {
        LlamaDecoder::rename_parameters()
    }
}

#[cfg(test)]
mod tests {
    use ndarray::array;
    use snafu::{report, Whatever};

    use crate::models::llama::causal_lm::LlamaCausalLM;
    use crate::models::util::tests::check_causal_lm;

    #[test]
    #[report]
    fn llama_causal_lm_emits_correct_output() -> Result<(), Whatever> {
        check_causal_lm!(
            LlamaCausalLM,
            "explosion-testing/llama2-kv-sharing",
            None,
            array![
                [0.0000, 0.0000, 3.9272, 0.0000, 0.0000, -0.2746, 0.0000, 0.0000],
                [-1.6657, 0.0000, 0.0000, 0.1828, 0.0000, 0.7174, 0.4477, -0.4943],
                [0.0000, -2.5876, -1.4347, 0.0000, 0.0000, -0.2561, -0.6859, 0.0000]
            ],
            epsilon = 1e-4,
        )
    }

    #[test]
    #[report]
    fn llama_causal_lm_emits_correct_output_float16() -> Result<(), Whatever> {
        check_causal_lm!(
            LlamaCausalLM,
            "explosion-testing/llama2-kv-sharing-float16",
            None,
            array![
                [0.0000, 0.0000, 3.9272, 0.0000, 0.0000, -0.2746, 0.0000, 0.0000],
                [-1.6657, 0.0000, 0.0000, 0.1828, 0.0000, 0.7174, 0.4477, -0.4943],
                [0.0000, -2.5876, -1.4347, 0.0000, 0.0000, -0.2561, -0.6859, 0.0000]
            ],
            epsilon = 1e-1
        )
    }
}