Skip to content

vllm.transformers_utils.configs.telechat2

Telechat configuration compatible with LlamaConfig.

Telechat2Config

Bases: PretrainedConfig

Source code in vllm/transformers_utils/configs/telechat2.py 
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
class Telechat2Config(PretrainedConfig):

    model_type = "telechat"
    keys_to_ignore_at_inference = ["past_key_values"]
    attribute_map = {
        "num_hidden_layers": "n_layer",
        "num_attention_heads": "n_head",
        "intermediate_size": "ffn_hidden_size",
        "rms_norm_eps": "layer_norm_epsilon"
    }

    def __init__(
        self,
        vocab_size=160256,
        hidden_size=4096,
        n_layer=30,
        n_head=32,
        layer_norm_epsilon=1e-5,
        initializer_range=0.02,
        use_cache=True,
        bos_token_id=1,
        eos_token_id=2,
        apply_residual_connection_post_layernorm=False,
        hidden_dropout=0.0,
        attention_dropout=0.0,
        ffn_hidden_size=12288,
        training_seqlen=8192,
        logn=True,
        embed_layernorm=False,
        hidden_act="silu",
        **kwargs,
    ):
        self.vocab_size = vocab_size
        n_embed = kwargs.pop("n_embed", None)
        self.hidden_size = hidden_size if n_embed is None else n_embed
        self.n_layer = n_layer
        self.n_head = n_head
        self.layer_norm_epsilon = layer_norm_epsilon
        self.initializer_range = initializer_range
        self.use_cache = use_cache
        self.apply_residual_connection_post_layernorm = (
            apply_residual_connection_post_layernorm)
        self.hidden_dropout = hidden_dropout
        self.attention_dropout = attention_dropout
        self.bos_token_id = bos_token_id
        self.eos_token_id = eos_token_id
        self.logn = logn
        self.training_seqlen = training_seqlen
        self.embed_layernorm = embed_layernorm
        self.num_key_value_heads = kwargs.pop("num_key_value_heads", None)
        self.ffn_hidden_size = ffn_hidden_size
        self.hidden_act = hidden_act
        super().__init__(bos_token_id=bos_token_id,
                         eos_token_id=eos_token_id,
                         **kwargs)

apply_residual_connection_post_layernorm instance-attribute 

apply_residual_connection_post_layernorm = (
    apply_residual_connection_post_layernorm
)

attention_dropout instance-attribute 

attention_dropout = attention_dropout

attribute_map class-attribute instance-attribute 

attribute_map = {
    "num_hidden_layers": "n_layer",
    "num_attention_heads": "n_head",
    "intermediate_size": "ffn_hidden_size",
    "rms_norm_eps": "layer_norm_epsilon",
}

bos_token_id instance-attribute 

bos_token_id = bos_token_id

embed_layernorm instance-attribute 

embed_layernorm = embed_layernorm

eos_token_id instance-attribute 

eos_token_id = eos_token_id

ffn_hidden_size instance-attribute 

ffn_hidden_size = ffn_hidden_size

hidden_act instance-attribute 

hidden_act = hidden_act

hidden_dropout instance-attribute 

hidden_dropout = hidden_dropout

hidden_size instance-attribute 

hidden_size = hidden_size if n_embed is None else n_embed

initializer_range instance-attribute 

initializer_range = initializer_range

keys_to_ignore_at_inference class-attribute instance-attribute 

keys_to_ignore_at_inference = ['past_key_values']

layer_norm_epsilon instance-attribute 

layer_norm_epsilon = layer_norm_epsilon

logn instance-attribute 

logn = logn

model_type class-attribute instance-attribute 

model_type = 'telechat'

n_head instance-attribute 

n_head = n_head

n_layer instance-attribute 

n_layer = n_layer

num_key_value_heads instance-attribute 

num_key_value_heads = pop('num_key_value_heads', None)

training_seqlen instance-attribute 

training_seqlen = training_seqlen

use_cache instance-attribute 

use_cache = use_cache

vocab_size instance-attribute 

vocab_size = vocab_size

__init__ 

__init__(
    vocab_size=160256,
    hidden_size=4096,
    n_layer=30,
    n_head=32,
    layer_norm_epsilon=1e-05,
    initializer_range=0.02,
    use_cache=True,
    bos_token_id=1,
    eos_token_id=2,
    apply_residual_connection_post_layernorm=False,
    hidden_dropout=0.0,
    attention_dropout=0.0,
    ffn_hidden_size=12288,
    training_seqlen=8192,
    logn=True,
    embed_layernorm=False,
    hidden_act="silu",
    **kwargs,
)
Source code in vllm/transformers_utils/configs/telechat2.py 
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
def __init__(
    self,
    vocab_size=160256,
    hidden_size=4096,
    n_layer=30,
    n_head=32,
    layer_norm_epsilon=1e-5,
    initializer_range=0.02,
    use_cache=True,
    bos_token_id=1,
    eos_token_id=2,
    apply_residual_connection_post_layernorm=False,
    hidden_dropout=0.0,
    attention_dropout=0.0,
    ffn_hidden_size=12288,
    training_seqlen=8192,
    logn=True,
    embed_layernorm=False,
    hidden_act="silu",
    **kwargs,
):
    self.vocab_size = vocab_size
    n_embed = kwargs.pop("n_embed", None)
    self.hidden_size = hidden_size if n_embed is None else n_embed
    self.n_layer = n_layer
    self.n_head = n_head
    self.layer_norm_epsilon = layer_norm_epsilon
    self.initializer_range = initializer_range
    self.use_cache = use_cache
    self.apply_residual_connection_post_layernorm = (
        apply_residual_connection_post_layernorm)
    self.hidden_dropout = hidden_dropout
    self.attention_dropout = attention_dropout
    self.bos_token_id = bos_token_id
    self.eos_token_id = eos_token_id
    self.logn = logn
    self.training_seqlen = training_seqlen
    self.embed_layernorm = embed_layernorm
    self.num_key_value_heads = kwargs.pop("num_key_value_heads", None)
    self.ffn_hidden_size = ffn_hidden_size
    self.hidden_act = hidden_act
    super().__init__(bos_token_id=bos_token_id,
                     eos_token_id=eos_token_id,
                     **kwargs)