-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathlanguage_bind_test.py
105 lines (93 loc) · 4.04 KB
/
language_bind_test.py
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
import os
os.environ["CUDA_VISIBLE_DEVICES"] = "7"
import torch
from eagle.model.multimodal_encoder.audio_models.languagebind_audio import LanguageBindAudio, CLIPVisionModel
from eagle.model.multimodal_encoder.audio_models.processing_audio import LanguageBindAudioProcessor
from eagle.model.multimodal_encoder.audio_models.tokenization_audio import LanguageBindAudioTokenizer
# from eagle.model.multimodal_encoder.languagebind import LanguageBindAudio, LanguageBindAudioTokenizer, LanguageBindAudioProcessor
# from eagle.model.multimodal_encoder.video_models.languagebind_video import LanguageBindVideo, CLIPVisionModel, CLIPVisionTransformer
# from eagle.model.multimodal_encoder.video_models.processing_video import LanguageBindVideoProcessor
# from eagle.model.multimodal_encoder.video_models.tokenization_video import LanguageBindVideoTokenizer
# from eagle.model.multimodal_encoder.languagebind_ori import LanguageBindVideo, LanguageBindVideoProcessor, LanguageBindVideoTokenizer
# pretrained_ckpt = './model/LanguageBind_Video_FT'
# vision_tower = CLIPVisionModel.from_pretrained(pretrained_ckpt).cuda()
# model = LanguageBindVideo.from_pretrained(pretrained_ckpt).cuda()
# tokenizer = LanguageBindVideoTokenizer.from_pretrained(pretrained_ckpt)
# video_process = LanguageBindVideoProcessor(model.config, tokenizer)
# model.eval()
# data = video_process(["dataset/Video/train/videochatgpt_tune/videochatgpt_tune/v___c8enCfzqw.mp4",
# "dataset/Video/train/videochatgpt_tune/videochatgpt_tune/v___mIAEE03bE.mp4"], return_tensors='pt')
# print(data['pixel_values'].shape)
# data['pixel_values'] = data['pixel_values'].cuda()
# image_features = model.get_image_features(data['pixel_values'])
# print(image_features.shape)
# torch.Size([2, 3, 8, 224, 224])
# pretrained_ckpt = './model/LanguageBind_Audio_FT'
# vision_tower = CLIPVisionModel.from_pretrained(pretrained_ckpt).cuda()
# model = LanguageBindAudio.from_pretrained(pretrained_ckpt)
# tokenizer = LanguageBindAudioTokenizer.from_pretrained(pretrained_ckpt)
# video_process = LanguageBindAudioProcessor(model.config, tokenizer)
# model.eval()
# data = video_process(["dataset/Audio/AudioSetCaps/example/_7Xe9vD3Hpg_4_10.mp3"], return_tensors='pt')
# print(data['pixel_values'].shape)
# CLIPVisionConfig {
# "_name_or_path": "./model/LanguageBind_Audio_FT",
# "add_time_attn": false,
# "attention_dropout": 0.0,
# "audio_mean": -4.2677393,
# "audio_sample_rate": 16000,
# "audio_std": 4.5689974,
# "force_patch_dropout": 0.0,
# "hidden_act": "gelu",
# "hidden_size": 1024,
# "image_size": 224,
# "initializer_factor": 1.0,
# "initializer_range": 0.02,
# "intermediate_size": 4096,
# "layer_norm_eps": 1e-05,
# "lora_alpha": 16,
# "lora_dropout": 0.0,
# "lora_r": 0,
# "model_type": "clip_vision_model",
# "num_attention_heads": 16,
# "num_channels": 3,
# "num_frames": 1,
# "num_hidden_layers": 24,
# "num_mel_bins": 112,
# "patch_size": 14,
# "projection_dim": 512,
# "target_length": 1036,
# "transformers_version": "4.45.2",
# "video_decode_backend": "decord"
# }
# CLIPVisionConfig {
# "add_time_attn": false,
# "attention_dropout": 0.0,
# "audio_mean": -4.2677393,
# "audio_sample_rate": 16000,
# "audio_std": 4.5689974,
# "force_patch_dropout": 0.0,
# "hidden_act": "gelu",
# "hidden_size": 1024,
# "image_size": 224,
# "initializer_factor": 1.0,
# "initializer_range": 0.02,
# "intermediate_size": 4096,
# "layer_norm_eps": 1e-05,
# "lora_alpha": 16,
# "lora_dropout": 0.0,
# "lora_r": 0,
# "model_type": "clip_vision_model",
# "num_attention_heads": 16,
# "num_channels": 3,
# "num_frames": 1,
# "num_hidden_layers": 24,
# "num_mel_bins": 112,
# "patch_size": 14,
# "projection_dim": 512,
# "target_length": 1036,
# "transformers_version": "4.45.2",
# "video_decode_backend": "decord"
# }
# Error(s) in loading state_dict for CLIPVisionModel:
# size mismatch for vision_model.embeddings.position_embedding.weight: copying a param with shape torch.Size([593, 1024]) from checkpoint, the shape in current model is torch.Size([257, 1024]).