File size: 2,879 Bytes
def21c3
 
 
 
 
 
 
7251f46
 
 
def21c3
 
 
 
 
7251f46
 
 
 
def21c3
 
 
7251f46
def21c3
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
from typing import List

from transformers import PretrainedConfig, AutoTokenizer


def config_to_moe_args(config):
    from megablocks.layers.arguments import Arguments as MoEArgs
    import torch.nn.functional as F

    # import pdb; pdb.set_trace()

    kwargs = {
        "activation_fn": F.silu,
        "mlp_type": "glu" if "glu" in config.activation_type.lower() else "mlp",
        "mlp_impl": "sparse",
        "hidden_size": config.d_model,
        "ffn_hidden_size": config.mlp_hidden_size,
        "moe_num_experts": 64,
        "num_layers": config.n_layers,
        # Handled by FSDP (https://github.com/databricks/megablocks/issues/57#issuecomment-1854594483)
        "moe_weight_parallelism": False,
        "moe_expert_model_parallelism": False,
        "moe_top_k": 8,
        # "moe_loss_weight": config.moe_loss_weight,
        # "device": config.init_device,
        # Handled by FSDP
        "bf16": False,
        "fp16": False,
        "bias": False,
        "return_bias": False,
    }

    return MoEArgs(**kwargs)

class MolmoeConfig(PretrainedConfig):
    model_type = "molmoe"
    keys_to_ignore_at_inference = ["past_key_values"]

    def __init__(
        self,
        vocab_size=50304,
        embedding_size=50304,
        hidden_size=4096,
        intermediate_size=11008,
        num_hidden_layers=32,
        num_attention_heads=32,
        num_key_value_heads=None,
        max_position_embeddings=2048,
        initializer_range=0.02,
        use_cache=True,
        layer_norm_eps: float = 1e-5,
        rope_theta=10000.0,
        clip_qkv=None,
        qkv_bias: bool = False,
        weight_tying: bool = False,
        use_position_ids: bool=True,
        tie_word_embeddings: bool=True,
        moe_num_experts: int = 64,
        moe_top_k: int = 8,
        **kwargs,
    ):
        self.vocab_size = vocab_size
        self.embedding_size = embedding_size
        self.max_position_embeddings = max_position_embeddings
        self.hidden_size = hidden_size
        self.intermediate_size = intermediate_size
        self.num_hidden_layers = num_hidden_layers
        self.num_attention_heads = num_attention_heads
        self.layer_norm_eps = layer_norm_eps
        self.weight_tying = weight_tying
        self.use_position_ids = use_position_ids

        # for backward compatibility
        if num_key_value_heads is None:
            num_key_value_heads = num_attention_heads

        self.num_key_value_heads = num_key_value_heads
        self.initializer_range = initializer_range
        self.use_cache = use_cache
        self.rope_theta = rope_theta
        self.clip_qkv = clip_qkv
        self.qkv_bias = qkv_bias
        self.tie_word_embeddings = tie_word_embeddings

        super().__init__(
            tie_word_embeddings=tie_word_embeddings,
            **kwargs,
        )

MolmoeConfig.register_for_auto_class()