File size: 5,022 Bytes
e19aac6
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
# Copyright 2024 NVIDIA CORPORATION & AFFILIATES
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#      http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
#
# SPDX-License-Identifier: Apache-2.0

# This file is modified from https://github.com/haotian-liu/LLaVA/

from typing import List, Optional, Tuple, Union

import torch
import torch.nn as nn

from transformers import AutoConfig, AutoModelForCausalLM, \
                         MistralConfig, MistralModel, MistralForCausalLM

from transformers.modeling_outputs import CausalLMOutputWithPast

from ..llava_arch import LlavaMetaModel, LlavaMetaForCausalLM


class LlavaMistralConfig(MistralConfig):
    model_type = "llava_mistral"
    pretraining_tp = 1


class LlavaMistralModel(MistralModel, LlavaMetaModel):
    config_class = LlavaMistralConfig

    def __init__(self, config: MistralConfig):
        super(LlavaMistralModel, self).__init__(config)


class LlavaMistralForCausalLM(MistralForCausalLM, LlavaMetaForCausalLM):
    config_class = LlavaMistralConfig

    def __init__(self, config):
        super(MistralForCausalLM, self).__init__(config)
        self.model = LlavaMistralModel(config)
        self.pretraining_tp = config.pretraining_tp
        self.vocab_size = config.vocab_size
        self.lm_head = nn.Linear(config.hidden_size, config.vocab_size, bias=False)

        # Initialize weights and apply final processing
        self.post_init()

    def get_model(self):
        return self.model
    
    def get_lm_head(self):
        return self.lm_head

    def forward(
        self,
        input_ids: torch.LongTensor = None,
        attention_mask: Optional[torch.Tensor] = None,
        position_ids: Optional[torch.LongTensor] = None,
        past_key_values: Optional[List[torch.FloatTensor]] = None,
        inputs_embeds: Optional[torch.FloatTensor] = None,
        labels: Optional[torch.LongTensor] = None,
        use_cache: Optional[bool] = None,
        output_attentions: Optional[bool] = None,
        output_hidden_states: Optional[bool] = None,
        images: Optional[torch.FloatTensor] = None,
        return_dict: Optional[bool] = None,
    ) -> Union[Tuple, CausalLMOutputWithPast]:
        if inputs_embeds is None:
            (
                input_ids,
                position_ids,
                attention_mask,
                past_key_values,
                inputs_embeds,
                labels
            ) = self.prepare_inputs_labels_for_multimodal(
                input_ids,
                position_ids,
                attention_mask,
                past_key_values,
                labels,
                images
            )
        if self.training:
            (
                _,
                new_position_ids,
                new_attention_mask,
                _,
                new_inputs_embeds,
                new_labels,
                sorted_seqlens_in_batch
            ) = self.repack_multimodal_data(
                input_ids,
                position_ids,
                attention_mask,
                past_key_values,
                inputs_embeds,
                labels
            )
            new_input_ids = None
            past_key_values = None
        else:
            new_attention_mask = attention_mask
            new_position_ids = position_ids
            new_inputs_embeds = inputs_embeds
            new_labels = labels
            sorted_seqlens_in_batch = attention_mask.sum(-1).int()
            new_input_ids = input_ids

        outputs = super().forward(
            input_ids=new_input_ids,
            attention_mask=new_attention_mask,
            position_ids=new_position_ids,
            past_key_values=past_key_values,
            inputs_embeds=new_inputs_embeds,
            labels=new_labels,
            use_cache=use_cache,
            output_attentions=output_attentions,
            output_hidden_states=output_hidden_states,
            return_dict=return_dict,
            seqlens_in_batch=sorted_seqlens_in_batch,
        )
        return outputs

    def prepare_inputs_for_generation(self, input_ids, past_key_values=None, inputs_embeds=None, **kwargs):
        images = kwargs.pop("images", None)
        _inputs = super().prepare_inputs_for_generation(
            input_ids, past_key_values=past_key_values, inputs_embeds=inputs_embeds, **kwargs
        )
        if images is not None:
            _inputs['images'] = images
        return _inputs

AutoConfig.register("llava_mistral", LlavaMistralConfig)
AutoModelForCausalLM.register(LlavaMistralConfig, LlavaMistralForCausalLM)