Text Generation
Transformers
Safetensors
English
llava_phi
custom_code
File size: 9,296 Bytes
b491de4
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
from typing import List, Optional, Tuple, Union

import torch
import torch.nn as nn
import math 
import pdb
from typing import Dict, Any
from PIL import Image

from transformers import AutoConfig, AutoModelForCausalLM, PretrainedConfig, PreTrainedModel
                         

from transformers.modeling_outputs import CausalLMOutputWithPast

from .llava_arch import LlavaMetaModel, LlavaMetaForCausalLM

from transformers.cache_utils import Cache, DynamicCache

from transformers.generation.utils import GenerationConfig

import sys
from .modeling_phi import PhiForCausalLM, PhiModel, PhiConfig
from .generation_utils import build_allava_input




################ Phi ###############################

class LlavaPhiConfig(PhiConfig):
    model_type = "llava_phi"

class LlavaPhiModel(LlavaMetaModel, PhiModel):
    config_class = LlavaPhiConfig

    def __init__(self, config: PhiConfig):
        super(LlavaPhiModel, self).__init__(config)



class LlavaPhiForCausalLM(PhiForCausalLM, LlavaMetaForCausalLM):
    config_class = LlavaPhiConfig

    def __init__(self, config, init_vision_encoder_from_ckpt=True): 
        # note that the default value is set to True for this inference version. In training `init_vision_encoder_from_ckpt` is default to be True. 
        config._attn_implementation = "flash_attention_2"

        super(PhiForCausalLM, self).__init__(config)
        # self.model is used in LlavaMetaForCausalLM.get_model(); self.transformer is used in PhiForCausalLM.forward()
        self.model = LlavaPhiModel(config)
        if hasattr(self.model, '_use_flash_attention_2'):
            assert self.model._use_flash_attention_2, 'flash attn is not enabled. check it out!'
        self.vocab_size = config.vocab_size
        self.lm_head = nn.Linear(config.hidden_size, config.vocab_size, bias=False)
        
        if init_vision_encoder_from_ckpt:
            vision_tower = self.get_vision_tower()
            print(f'loading from CLIP first. This should only be used at inference!!!')
            vision_tower.load_model() # 
            
        # Initialize weights and apply final processing
        self.post_init()

    def get_model(self):
        return self.model
    
    def get_tokenizer(self):
        return self.tokenizer
    
    def get_processor(self):
        return self.model.vision_tower.image_processor
    

    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(
            ) = self.prepare_inputs_labels_for_multimodal_new(
                input_ids,
                position_ids,
                attention_mask,
                past_key_values,
                labels,
                images
            )

        # pdb.set_trace()
        return super().forward(
            input_ids=input_ids,
            attention_mask=attention_mask,
            position_ids=position_ids,
            past_key_values=past_key_values,
            inputs_embeds=inputs_embeds,
            labels=labels,
            use_cache=use_cache,
            output_attentions=output_attentions,
            output_hidden_states=output_hidden_states,
            return_dict=return_dict
        )

    def prepare_inputs_for_generation(self, input_ids, past_key_values=None, attention_mask=None, inputs_embeds=None, **kwargs):
        '''
        This function is called for each token at inference
        '''
        # pdb.set_trace()
        images = kwargs.pop("images", None)

        ####################################################
        # lines from modeling_phi.py
        ####################################################

        if past_key_values is not None:
            if isinstance(past_key_values, Cache):
                cache_length = past_key_values.get_seq_length()
                past_length = past_key_values.seen_tokens
                max_cache_length = past_key_values.get_max_length()
            else:
                cache_length = past_length = past_key_values[0][0].shape[2]
                max_cache_length = None

            # Keep only the unprocessed tokens:
            # 1 - If the length of the attention_mask exceeds the length of input_ids, then we are in a setting where
            # some of the inputs are exclusively passed as part of the cache (e.g. when passing input_embeds as
            # input)
            if attention_mask is not None and attention_mask.shape[1] > input_ids.shape[1]:
                input_ids = input_ids[:, -(attention_mask.shape[1] - past_length) :]
            # 2 - If the past_length is smaller than input_ids', then input_ids holds all input tokens. We can discard
            # input_ids based on the past_length.
            elif past_length < input_ids.shape[1]:
                input_ids = input_ids[:, past_length:]
            # 3 - Otherwise (past_length >= input_ids.shape[1]), let's assume input_ids only has unprocessed tokens.
            elif past_length >= input_ids.shape[1]:
                input_ids = input_ids[:, [-1]] # only keep the last one!

            # If we are about to go beyond the maximum cache length, we need to crop the input attention mask.
            if (
                max_cache_length is not None
                and attention_mask is not None
                and cache_length + input_ids.shape[1] > max_cache_length
            ):
                attention_mask = attention_mask[:, -max_cache_length:]

        position_ids = kwargs.get("position_ids", None)
        if attention_mask is not None and position_ids is None:
            # create position_ids on the fly for batch generation
            position_ids = attention_mask.long().cumsum(-1) - 1
            position_ids.masked_fill_(attention_mask == 0, 1)
            if past_key_values:
                position_ids = position_ids[:, -input_ids.shape[1] :]

        # if `inputs_embeds` are passed, we only want to use them in the 1st generation step
        if inputs_embeds is not None and past_key_values is None:
            model_inputs = {"inputs_embeds": inputs_embeds}
        else:
            model_inputs = {"input_ids": input_ids}
        
        model_inputs.update(
            {
                "position_ids": position_ids,
                "past_key_values": past_key_values,
                "use_cache": kwargs.get("use_cache"),
                "attention_mask": attention_mask,
            }
        )
        ####################################################
        # end of lines from modeling_phi.py
        ####################################################


        if images is not None:
            model_inputs['images'] = images
        return model_inputs




    def chat(
        self, 
        texts: Optional[str | list[list[str, str]]], 
        images: Optional[str | list[str]] = None, 
        history: Optional[list[str]] = None, 
        stream = False, 
        return_history = False, 
        **kwargs
    ):
        '''
        texts: if `str`, then generate for a single round; if list[dict], 
        images: str (optional), local path to an image.
        '''
        use_cache = kwargs.pop('use_cache', True)


        ############################
        # merge history
        ############################            
        input_ids, image_tensors, history = build_allava_input(
            tokenizer = self.get_tokenizer(), 
            processor = self.get_processor(), 
            texts = texts, 
            images = images, 
            history=history,
            return_history=return_history,
            device = self.device
        )

        ############################
        # generate response
        ############################
        # with torch.autocast(device_type='cuda'):
        if 'cuda' in str(self.device):
            device_type = 'cuda'
        else:
            device_type = 'cpu'

        with torch.autocast(device_type=device_type, dtype=self.dtype):
            output_ids = self.generate(
                inputs=input_ids,
                images=image_tensors,
                use_cache=use_cache,
                **kwargs)

        answer = self.get_tokenizer().decode(output_ids[0, input_ids.shape[1]:], skip_special_tokens=True).strip()

        if return_history:
            history[-1][-1] = answer
            return answer, history
        return answer


AutoConfig.register("llava_phi", LlavaPhiConfig)
AutoModelForCausalLM.register(LlavaPhiConfig, LlavaPhiForCausalLM)