File size: 10,232 Bytes
d123ff3
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
import warnings
from typing import Optional, Tuple
from transformers.models.llama.modeling_llama import (
    LlamaConfig, 
    LlamaModel,
    LlamaForCausalLM,
    LlamaAttention,
    LlamaFlashAttention2,
    LlamaSdpaAttention,
    LlamaMLP,
    LlamaDecoderLayer,
)
from mybitnet.bitnet import BitLinear, BitLinear158b
import torch
from torch import nn

class BitLlamaConfig(LlamaConfig):
    model_type = "bit_llama"

    def __init__(self, bitnet_type="1.58b", bits=8, **kwargs):
        super().__init__(**kwargs)
        self.bitnet_type = bitnet_type
        if self.bitnet_type not in ["1.58b", "1b"]:
            raise ValueError("bitnet_type must be either '1.58b' or '1b'.")
        self.bits = bits


class BitLlamaMLP(LlamaMLP):
    def __init__(self, config):
        super().__init__(config)
        if config.bitnet_type=="1b":
            self.gate_proj = BitLinear(self.hidden_size, self.intermediate_size, bias=False, rms_norm_eps=config.rms_norm_eps, bits=config.bits, flg_before_linear=False)
            self.up_proj = BitLinear(self.hidden_size, self.intermediate_size, bias=False, rms_norm_eps=config.rms_norm_eps, bits=config.bits, flg_before_linear=True)
            self.down_proj = BitLinear(self.intermediate_size, self.hidden_size, bias=False, rms_norm_eps=config.rms_norm_eps, bits=config.bits, flg_before_linear=True)
        elif config.bitnet_type=="1.58b":
            self.gate_proj = BitLinear158b(self.hidden_size, self.intermediate_size, bias=False, rms_norm_eps=config.rms_norm_eps, bits=config.bits, flg_before_linear=False)
            self.up_proj = BitLinear158b(self.hidden_size, self.intermediate_size, bias=False, rms_norm_eps=config.rms_norm_eps, bits=config.bits, flg_before_linear=True)
            self.down_proj = BitLinear158b(self.intermediate_size, self.hidden_size, bias=False, rms_norm_eps=config.rms_norm_eps, bits=config.bits, flg_before_linear=True)
        else:
            raise ValueError("bitnet_type must be either '1.58b' or '1b'.")
        
class BitLlamaAttention(LlamaAttention):
    def __init__(self, config: BitLlamaConfig, layer_idx: Optional[int] = None):
        super().__init__(config)
        if config.bitnet_type=="1b":
            self.q_proj = BitLinear(self.hidden_size, self.num_heads * self.head_dim, bias=False, rms_norm_eps=config.rms_norm_eps, bits=config.bits, flg_before_linear=True)
            self.k_proj = BitLinear(self.hidden_size, self.num_key_value_heads * self.head_dim, bias=False, rms_norm_eps=config.rms_norm_eps, bits=config.bits, flg_before_linear=True)
            self.v_proj = BitLinear(self.hidden_size, self.num_key_value_heads * self.head_dim, bias=False, rms_norm_eps=config.rms_norm_eps, bits=config.bits, flg_before_linear=True)
            self.o_proj = BitLinear(self.hidden_size, self.hidden_size, bias=False, rms_norm_eps=config.rms_norm_eps, bits=config.bits, flg_before_linear=True)
        elif config.bitnet_type=="1.58b":
            self.q_proj = BitLinear158b(self.hidden_size, self.num_heads * self.head_dim, bias=False, rms_norm_eps=config.rms_norm_eps, bits=config.bits, flg_before_linear=True)
            self.k_proj = BitLinear158b(self.hidden_size, self.num_key_value_heads * self.head_dim, bias=False, rms_norm_eps=config.rms_norm_eps, bits=config.bits, flg_before_linear=True)
            self.v_proj = BitLinear158b(self.hidden_size, self.num_key_value_heads * self.head_dim, bias=False, rms_norm_eps=config.rms_norm_eps, bits=config.bits, flg_before_linear=True)
            self.o_proj = BitLinear158b(self.hidden_size, self.hidden_size, bias=False, rms_norm_eps=config.rms_norm_eps, bits=config.bits, flg_before_linear=True)
        else:
            raise ValueError("bitnet_type must be either '1.58b' or '1b'.")

class BitLlamaFlashAttention2(LlamaFlashAttention2):
    def __init__(self, config: BitLlamaConfig, layer_idx: Optional[int] = None):
        super().__init__(config, layer_idx)
        if config.bitnet_type=="1b":
            self.q_proj = BitLinear(self.hidden_size, self.num_heads * self.head_dim, bias=False, rms_norm_eps=config.rms_norm_eps, bits=config.bits, flg_before_linear=True)
            self.k_proj = BitLinear(self.hidden_size, self.num_key_value_heads * self.head_dim, bias=False, rms_norm_eps=config.rms_norm_eps, bits=config.bits, flg_before_linear=True)
            self.v_proj = BitLinear(self.hidden_size, self.num_key_value_heads * self.head_dim, bias=False, rms_norm_eps=config.rms_norm_eps, bits=config.bits, flg_before_linear=True)
            self.o_proj = BitLinear(self.hidden_size, self.hidden_size, bias=False, rms_norm_eps=config.rms_norm_eps, bits=config.bits, flg_before_linear=True)
        elif config.bitnet_type=="1.58b":
            self.q_proj = BitLinear158b(self.hidden_size, self.num_heads * self.head_dim, bias=False, rms_norm_eps=config.rms_norm_eps, bits=config.bits, flg_before_linear=True)
            self.k_proj = BitLinear158b(self.hidden_size, self.num_key_value_heads * self.head_dim, bias=False, rms_norm_eps=config.rms_norm_eps, bits=config.bits, flg_before_linear=True)
            self.v_proj = BitLinear158b(self.hidden_size, self.num_key_value_heads * self.head_dim, bias=False, rms_norm_eps=config.rms_norm_eps, bits=config.bits, flg_before_linear=True)
            self.o_proj = BitLinear158b(self.hidden_size, self.hidden_size, bias=False, rms_norm_eps=config.rms_norm_eps, bits=config.bits, flg_before_linear=True)
        else:
            raise ValueError("bitnet_type must be either '1.58b' or '1b'.")

class BitLlamaSdpaAttention(LlamaSdpaAttention):
    def __init__(self, config: BitLlamaConfig, layer_idx: Optional[int] = None):
        super().__init__(config, layer_idx)
        if config.bitnet_type=="1b":
            self.q_proj = BitLinear(self.hidden_size, self.num_heads * self.head_dim, bias=False, rms_norm_eps=config.rms_norm_eps, bits=config.bits, flg_before_linear=True)
            self.k_proj = BitLinear(self.hidden_size, self.num_key_value_heads * self.head_dim, bias=False, rms_norm_eps=config.rms_norm_eps, bits=config.bits, flg_before_linear=True)
            self.v_proj = BitLinear(self.hidden_size, self.num_key_value_heads * self.head_dim, bias=False, rms_norm_eps=config.rms_norm_eps, bits=config.bits, flg_before_linear=True)
            self.o_proj = BitLinear(self.hidden_size, self.hidden_size, bias=False, rms_norm_eps=config.rms_norm_eps, bits=config.bits, flg_before_linear=True)
        elif config.bitnet_type=="1.58b":
            self.q_proj = BitLinear158b(self.hidden_size, self.num_heads * self.head_dim, bias=False, rms_norm_eps=config.rms_norm_eps, bits=config.bits, flg_before_linear=True)
            self.k_proj = BitLinear158b(self.hidden_size, self.num_key_value_heads * self.head_dim, bias=False, rms_norm_eps=config.rms_norm_eps, bits=config.bits, flg_before_linear=True)
            self.v_proj = BitLinear158b(self.hidden_size, self.num_key_value_heads * self.head_dim, bias=False, rms_norm_eps=config.rms_norm_eps, bits=config.bits, flg_before_linear=True)
            self.o_proj = BitLinear158b(self.hidden_size, self.hidden_size, bias=False, rms_norm_eps=config.rms_norm_eps, bits=config.bits, flg_before_linear=True)
        else:
            raise ValueError("bitnet_type must be either '1.58b' or '1b'.")

BITLLAMA_ATTENTION_CLASSES = {
    "eager": BitLlamaAttention,
    "flash_attention_2": BitLlamaFlashAttention2,
    "sdpa": BitLlamaSdpaAttention,
}

class BitLlamaDecoderLayer(LlamaDecoderLayer):
    def __init__(self, config: BitLlamaConfig, layer_idx: int):
        super().__init__(config, layer_idx)
        self.self_attn = BITLLAMA_ATTENTION_CLASSES[config._attn_implementation](config=config, layer_idx=layer_idx)
        self.mlp = BitLlamaMLP(config)
        del self.input_layernorm
        del self.post_attention_layernorm

    def forward(
        self,
        hidden_states: torch.Tensor,
        attention_mask: Optional[torch.Tensor] = None,
        position_ids: Optional[torch.LongTensor] = None,
        past_key_value: Optional[Tuple[torch.Tensor]] = None,
        output_attentions: Optional[bool] = False,
        use_cache: Optional[bool] = False,
        cache_position: Optional[torch.LongTensor] = None,
        **kwargs,
    ) -> Tuple[torch.FloatTensor, Optional[Tuple[torch.FloatTensor, torch.FloatTensor]]]:
        """
        refers: https://github.com/huggingface/transformers/blob/c5f0288bc7d76f65996586f79f69fba8867a0e67/src/transformers/models/llama/modeling_llama.py#L693
        """
        if "padding_mask" in kwargs:
            warnings.warn(
                "Passing `padding_mask` is deprecated and will be removed in v4.37. Please make sure use `attention_mask` instead.`"
            )

        residual = hidden_states

        # Self Attention
        hidden_states, self_attn_weights, present_key_value = self.self_attn(
            hidden_states=hidden_states,
            attention_mask=attention_mask,
            position_ids=position_ids,
            past_key_value=past_key_value,
            output_attentions=output_attentions,
            use_cache=use_cache,
            cache_position=cache_position,
            **kwargs,
        )
        hidden_states = residual + hidden_states

        # Fully Connected
        residual = hidden_states
        hidden_states = self.mlp(hidden_states)
        hidden_states = residual + hidden_states

        outputs = (hidden_states,)

        if output_attentions:
            outputs += (self_attn_weights,)

        if use_cache:
            outputs += (present_key_value,)

        return outputs

class BitLlamaModel(LlamaModel):
    config_class = BitLlamaConfig

    def __init__(self, config: BitLlamaConfig):
        super().__init__(config)
        self.layers = nn.ModuleList(
            [BitLlamaDecoderLayer(config, layer_idx) for layer_idx in range(config.num_hidden_layers)]
        )

class BitLlamaForCausalLM(LlamaForCausalLM):
    config_class = BitLlamaConfig

    def __init__(self, config: BitLlamaConfig):
        super().__init__(config)
        self.model = BitLlamaModel(config)
        self.lm_head = BitLinear(config.hidden_size, config.vocab_size, bias=False, bits=config.bits, flg_before_linear=True)