File size: 9,622 Bytes
b84549f
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
import torch
import sys
from torch import nn
from new_impl.cv.elasticdnn.api.model import ElasticDNN_OfflineVQAFMModel, ElasticDNN_OfflineVQAMDModel
from new_impl.cv.elasticdnn.api.algs.md_pretraining_wo_fbs import ElasticDNN_MDPretrainingWoFBSAlg
from new_impl.cv.elasticdnn.model.base import ElasticDNNUtil
from new_impl.cv.elasticdnn.pipeline.offline.fm_to_md.base import FM_to_MD_Util
from blip import FM_to_MD_blip_Util
from new_impl.cv.elasticdnn.pipeline.offline.fm_lora.base import FMLoRA_Util
from blip import FMLoRA_blip_Util
from blip import ElasticblipUtil
from utils.dl.common.model import LayerActivation2, get_module, get_parameter
from utils.common.exp import save_models_dict_for_init, get_res_save_dir
from data import build_scenario
from utils.dl.common.loss import CrossEntropyLossSoft
import torch.nn.functional as F
from utils.common.log import logger


class ElasticDNN_blip_OfflineVQAFMModel(ElasticDNN_OfflineVQAFMModel):
    def generate_md_by_reducing_width(self, reducing_width_ratio, samples: torch.Tensor): # TODO:
        return FM_to_MD_blip_Util().init_md_from_fm_by_reducing_width_with_perf_test(self.models_dict['main'], 
                                                                        reducing_width_ratio, samples)
        # raise NotImplementedError

    def get_feature_hook(self) -> LayerActivation2:
        return LayerActivation2(get_module(self.models_dict['main'], 'text_decoder.cls.predictions.decoder'))
    
    def get_elastic_dnn_util(self) -> ElasticDNNUtil: # TODO:
        return ElasticblipUtil()
    
    def forward_to_get_task_loss(self, x, y, *args, **kwargs):
        self.to_train_mode()
        # print(x['input_ids'].size(), x['pixel_values'].size(), )
        #o = self.infer(x)
        o = self.models_dict['main'](**y)
        # print(o.size(), y.size(), o, y)
        
        #return F.binary_cross_entropy_with_logits(o, y) * y.shape[1]
        return o.loss
    def get_lora_util(self) -> FMLoRA_Util:
        return FMLoRA_blip_Util()
    
    def get_task_head_params(self):
        head = get_module(self.models_dict['main'], 'text_decoder.cls.predictions.decoder')
        params_name = {k for k, v in head.named_parameters()}
        logger.info(f'task head params: {params_name}')
        return list(head.parameters())
        
        
class ElasticDNN_blip_OfflineVQAMDModel(ElasticDNN_OfflineVQAMDModel):
    # def __init__(self, name: str, models_dict_path: str, device: str):
    #     super().__init__(name, models_dict_path, device)
        
    #     self.distill_criterion = CrossEntropyLossSoft()
        
    def get_feature_hook(self) -> LayerActivation2:
        return LayerActivation2(get_module(self.models_dict['main'], 'text_decoder.cls.predictions.decoder'))
    
    def forward_to_get_task_loss(self, x, y, *args, **kwargs):
        self.to_train_mode()
        # print(x['input_ids'].size(), x['pixel_values'].size(), )
        #o = self.infer(x)
        o = self.models_dict['main'](**y)
        # print(o.size(), y.size(), o, y)
        
        #return F.binary_cross_entropy_with_logits(o, y) * y.shape[1]
        return o.loss
    
    def get_distill_loss(self, student_output, teacher_output):
        # print(student_output, teacher_output)
        return F.mse_loss(student_output, teacher_output.detach())
    
    def get_matched_param_of_fm(self, self_param_name, fm: nn.Module): # TODO:
        if any([k in self_param_name for k in ['fbs', 'ab', 'embeddings']]):
            return None
        
        p = get_parameter(self.models_dict['main'], self_param_name)
        if p.dim() == 0:
            return None
        
        elif p.dim() == 1 and 'LayerNorm' in self_param_name.lower() and 'weight' in self_param_name:
            # if self_param_name.startswith('norm'):
            #     return None
            return get_parameter(fm, self_param_name)
        elif p.dim() == 1 and 'norm' in self_param_name and 'weight' in self_param_name:
            return get_parameter(fm, self_param_name)
        
        # 1. xx.query.weight -> xx.query.fc.weight and xx.query.ab.0/1
        if ('query' in self_param_name or 'key' in self_param_name or \
            'value' in self_param_name) and ('weight' in self_param_name):
            ss = self_param_name.split('.')
            # raise NotImplementedError() # TODO:
            fm_qkv_name = '.'.join(ss[0: -1]) + '.fc'
            fm_qkv = get_module(fm, fm_qkv_name)
            
            fm_abs_name = '.'.join(ss[0: -1]) + '.ab'
            fm_abs = get_module(fm, fm_abs_name)
            
            return torch.cat([
                fm_qkv.weight.data, # task-agnositc params
                fm_abs[1].weight @ fm_abs[0].weight
            ], dim=0)
            
        elif ('query' in self_param_name or 'key' in self_param_name or \
            'value' in self_param_name) and ('bias' in self_param_name):
            ss = self_param_name.split('.')
            
            fm_qkv_name = '.'.join(ss[0: -1]) + '.fc.bias'
            return get_parameter(fm, fm_qkv_name)
            
        elif 'intermediate.dense' in self_param_name:
            fm_param_name = self_param_name.replace('.linear', '')
            return get_parameter(fm, fm_param_name)
        
        elif 'qkv.weight' in self_param_name:
            ss = self_param_name.split('.')
            
            fm_qkv_name = '.'.join(ss[0: -1]) + '.fc'
            fm_qkv = get_module(fm, fm_qkv_name)
            
            fm_abs_name = '.'.join(ss[0: -1]) + '.ab'
            fm_abs = get_module(fm, fm_abs_name)
            
            return torch.cat([
                fm_qkv.weight.data, # task-agnositc params
                fm_abs[1].weight @ fm_abs[0].weight # task-specific params (LoRA)
            ], dim=0)
            
        # elif 'to_qkv.bias' in self_param_name:
        #     ss = self_param_name.split('.')
            
        #     fm_qkv_name = '.'.join(ss[0: -2]) + '.qkv.bias'
        #     return get_parameter(fm, fm_qkv_name)
            
        elif 'mlp.fc1' in self_param_name and 'weight' in self_param_name:
            fm_param_name = self_param_name.replace('.linear', '')
            return get_parameter(fm, fm_param_name)

        elif 'mlp.fc2' in self_param_name and 'weight' in self_param_name:
            fm_param_name = self_param_name
            res = get_parameter(fm, fm_param_name)
            # print('mlp fc2 debug', fm_param_name, res is None)
            return res
        else:
            #return get_parameter(fm, self_param_name)
            return None
        
        
        
if __name__ == '__main__':
    from utils.dl.common.env import set_random_seed
    set_random_seed(1)
    
    # 3. init scenario
    scenario = build_scenario(
        source_datasets_name=['VQA_split1'],
        target_datasets_order=['VQA_split1_c'] * 1, # TODO
        da_mode='close_set',
        data_dirs={
            'VQA_split1': '/data/zql/datasets/vqav2',
            'VQA_split1_c': '/data/zql/datasets/vqav2'
        },
    )
    
    # 1. init model
    # from dnns.deeplabv3.head import modify_forward_head
    # modify_forward_head() # TODO: bring a bug
    # from dnns.vit import vit_b_16
    fm_models_dict_path = 'new_impl/mm/Vis_bert/QuestionAnswering/results/blip_fbs.py/20231020/999999-162038-/data/zql/concept-drift-in-edge-projects/UniversalElasticNet/new_impl/mm/Vis_bert/QuestionAnswering/blip_fbs.py/models/fm_best.pt'
    fm_models = torch.load(fm_models_dict_path)
    fm_models_dict_path = save_models_dict_for_init(fm_models, __file__, 'fm_blip_vqa_lora')
    md_models_dict_path = save_models_dict_for_init(
        torch.load('new_impl/mm/Vis_bert/QuestionAnswering/results/blip_fbs.py/20231020/999999-162038-/data/zql/concept-drift-in-edge-projects/UniversalElasticNet/new_impl/mm/Vis_bert/QuestionAnswering/blip_fbs.py/models/md_best.pt'), 
        __file__, 'md_blip_vqa_raw_pretrained')
    device = 'cuda'
    
    fm_model = ElasticDNN_blip_OfflineVQAFMModel('fm', fm_models_dict_path, device)
    md_model = ElasticDNN_blip_OfflineVQAMDModel('md', md_models_dict_path, device)
    
    # 2. init alg
    models = {
        'fm': fm_model,
        'md': md_model
    }
    from new_impl.cv.elasticdnn.api.algs.md_pretraining_index_v2_train_index_and_md import ElasticDNN_MDPretrainingIndexAlg
    fm_to_md_alg = ElasticDNN_MDPretrainingIndexAlg(models, get_res_save_dir(__file__, sys.argv[0]))

    sample_dataset = list(scenario.get_offline_datasets().values())[0]['train']
    sample = sample_dataset[0][0]
    
    from utils.dl.common.lr_scheduler import get_linear_schedule_with_warmup
    fm_to_md_alg.run(scenario, hyps={
        'launch_tbboard': False,
        
        'samples_size': sample,
        
        'FBS_r': 8,
        'FBS_ignore_layers': [],
        
        'train_batch_size': 16,
        'val_batch_size': 256,
        'num_workers': 16,
        'optimizer': 'AdamW',
        'optimizer_args': {'lr': 1e-4, 'betas': [0.9, 0.999], 'weight_decay': 0.01},
        'scheduler': 'LambdaLR',
        'scheduler_args': {'lr_lambda': get_linear_schedule_with_warmup(10000, 70000)},
        
        'indexes_optimizer_args': {'lr': 3e-3, 'momentum': 0.9, 'weight_decay': 5e-4},
        
        'num_iters': 80000,
        'val_freq': 20,
        
        'max_sparsity': 0.9,
        'min_sparsity': 0.0,
        'l1_reg_loss_weight': 1e-9,
        'index_loss_weight': 1e-4,
        'val_num_sparsities': 4,
        
        'bn_cal_num_iters': 0,
        
        'index_init': 'zero',
        'index_guided_linear_comb_split_size': 512
    })