Upload 6 files
Browse files- AutoModel.pth +3 -0
- model_config.json +29 -0
- tokenizer_config.json +32 -0
- vocab.json +0 -0
- vocab.txt +0 -0
- 配置权重 +1309 -0
AutoModel.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:27227c5532d3027b044ad00c12c4aed1334e910459edf80b6b0f2bc83e673198
|
3 |
+
size 3237240570
|
model_config.json
ADDED
@@ -0,0 +1,29 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"model_name": "AutoModel",
|
3 |
+
"model_type": "multimodal-transformer",
|
4 |
+
"hidden_size": 768,
|
5 |
+
"num_attention_heads": 12,
|
6 |
+
"num_hidden_layers": 12,
|
7 |
+
"intermediate_size": 2048,
|
8 |
+
"hidden_dropout_prob": 0.1,
|
9 |
+
"attention_probs_dropout_prob": 0.1,
|
10 |
+
"image_size": 224,
|
11 |
+
"image_channels": 3,
|
12 |
+
"patch_size": 16,
|
13 |
+
"max_position_embeddings": 512,
|
14 |
+
"vocab_size": 30522,
|
15 |
+
"type_vocab_size": 2,
|
16 |
+
"audio_sample_rate": 16000,
|
17 |
+
"audio_frame_size": 1024,
|
18 |
+
"audio_hop_size": 512,
|
19 |
+
"enable_vqa": true,
|
20 |
+
"enable_caption": true,
|
21 |
+
"enable_retrieval": true,
|
22 |
+
"enable_asr": true,
|
23 |
+
"enable_realtime_asr": true,
|
24 |
+
"batch_size": 32,
|
25 |
+
"learning_rate": 0.0001,
|
26 |
+
"weight_decay": 0.01,
|
27 |
+
"warmup_steps": 10000,
|
28 |
+
"max_steps": 100000
|
29 |
+
}
|
tokenizer_config.json
ADDED
@@ -0,0 +1,32 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"tokenizer_name": "AutoTokenizer",
|
3 |
+
"pretrained_model_name": "AutoModel",
|
4 |
+
"vocab": {
|
5 |
+
"vocab_size": 30522,
|
6 |
+
"model_max_length": 512,
|
7 |
+
"padding_side": "right",
|
8 |
+
"truncation_side": "right",
|
9 |
+
"special_tokens": {
|
10 |
+
"pad_token": "[PAD]",
|
11 |
+
"unk_token": "[UNK]",
|
12 |
+
"cls_token": "[CLS]",
|
13 |
+
"sep_token": "[SEP]",
|
14 |
+
"mask_token": "[MASK]"
|
15 |
+
},
|
16 |
+
"tokenizer_type": "WordPiece",
|
17 |
+
"lowercase": true,
|
18 |
+
"pad_token_id": 0,
|
19 |
+
"unk_token_id": 100,
|
20 |
+
"cls_token_id": 101,
|
21 |
+
"sep_token_id": 102,
|
22 |
+
"mask_token_id": 103
|
23 |
+
},
|
24 |
+
"normalization": {
|
25 |
+
"lowercase": true,
|
26 |
+
"strip_accents": true
|
27 |
+
},
|
28 |
+
"preprocessing": {
|
29 |
+
"do_lower_case": true,
|
30 |
+
"handle_chinese_chars": true
|
31 |
+
}
|
32 |
+
}
|
vocab.json
ADDED
The diff for this file is too large to render.
See raw diff
|
|
vocab.txt
ADDED
The diff for this file is too large to render.
See raw diff
|
|
配置权重
ADDED
@@ -0,0 +1,1309 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
配置文件已生成: C:\Users\baby7\Desktop\fastAPI\model_config.json
|
2 |
+
{
|
3 |
+
"model_info": {
|
4 |
+
"total_layers": 176,
|
5 |
+
"layers": [
|
6 |
+
{
|
7 |
+
"name": "image_encoder.encoder_layer.0.weight",
|
8 |
+
"shape": [
|
9 |
+
64,
|
10 |
+
3,
|
11 |
+
3,
|
12 |
+
3
|
13 |
+
],
|
14 |
+
"dtype": "torch.float32"
|
15 |
+
},
|
16 |
+
{
|
17 |
+
"name": "image_encoder.encoder_layer.0.bias",
|
18 |
+
"shape": [
|
19 |
+
64
|
20 |
+
],
|
21 |
+
"dtype": "torch.float32"
|
22 |
+
},
|
23 |
+
{
|
24 |
+
"name": "image_encoder.encoder_layer.4.weight",
|
25 |
+
"shape": [
|
26 |
+
768,
|
27 |
+
788544
|
28 |
+
],
|
29 |
+
"dtype": "torch.float32"
|
30 |
+
},
|
31 |
+
{
|
32 |
+
"name": "image_encoder.encoder_layer.4.bias",
|
33 |
+
"shape": [
|
34 |
+
768
|
35 |
+
],
|
36 |
+
"dtype": "torch.float32"
|
37 |
+
},
|
38 |
+
{
|
39 |
+
"name": "text_encoder.transformer_layer.self_attn.in_proj_weight",
|
40 |
+
"shape": [
|
41 |
+
2304,
|
42 |
+
768
|
43 |
+
],
|
44 |
+
"dtype": "torch.float32"
|
45 |
+
},
|
46 |
+
{
|
47 |
+
"name": "text_encoder.transformer_layer.self_attn.in_proj_bias",
|
48 |
+
"shape": [
|
49 |
+
2304
|
50 |
+
],
|
51 |
+
"dtype": "torch.float32"
|
52 |
+
},
|
53 |
+
{
|
54 |
+
"name": "text_encoder.transformer_layer.self_attn.out_proj.weight",
|
55 |
+
"shape": [
|
56 |
+
768,
|
57 |
+
768
|
58 |
+
],
|
59 |
+
"dtype": "torch.float32"
|
60 |
+
},
|
61 |
+
{
|
62 |
+
"name": "text_encoder.transformer_layer.self_attn.out_proj.bias",
|
63 |
+
"shape": [
|
64 |
+
768
|
65 |
+
],
|
66 |
+
"dtype": "torch.float32"
|
67 |
+
},
|
68 |
+
{
|
69 |
+
"name": "text_encoder.transformer_layer.linear1.weight",
|
70 |
+
"shape": [
|
71 |
+
2048,
|
72 |
+
768
|
73 |
+
],
|
74 |
+
"dtype": "torch.float32"
|
75 |
+
},
|
76 |
+
{
|
77 |
+
"name": "text_encoder.transformer_layer.linear1.bias",
|
78 |
+
"shape": [
|
79 |
+
2048
|
80 |
+
],
|
81 |
+
"dtype": "torch.float32"
|
82 |
+
},
|
83 |
+
{
|
84 |
+
"name": "text_encoder.transformer_layer.linear2.weight",
|
85 |
+
"shape": [
|
86 |
+
768,
|
87 |
+
2048
|
88 |
+
],
|
89 |
+
"dtype": "torch.float32"
|
90 |
+
},
|
91 |
+
{
|
92 |
+
"name": "text_encoder.transformer_layer.linear2.bias",
|
93 |
+
"shape": [
|
94 |
+
768
|
95 |
+
],
|
96 |
+
"dtype": "torch.float32"
|
97 |
+
},
|
98 |
+
{
|
99 |
+
"name": "text_encoder.transformer_layer.norm1.weight",
|
100 |
+
"shape": [
|
101 |
+
768
|
102 |
+
],
|
103 |
+
"dtype": "torch.float32"
|
104 |
+
},
|
105 |
+
{
|
106 |
+
"name": "text_encoder.transformer_layer.norm1.bias",
|
107 |
+
"shape": [
|
108 |
+
768
|
109 |
+
],
|
110 |
+
"dtype": "torch.float32"
|
111 |
+
},
|
112 |
+
{
|
113 |
+
"name": "text_encoder.transformer_layer.norm2.weight",
|
114 |
+
"shape": [
|
115 |
+
768
|
116 |
+
],
|
117 |
+
"dtype": "torch.float32"
|
118 |
+
},
|
119 |
+
{
|
120 |
+
"name": "text_encoder.transformer_layer.norm2.bias",
|
121 |
+
"shape": [
|
122 |
+
768
|
123 |
+
],
|
124 |
+
"dtype": "torch.float32"
|
125 |
+
},
|
126 |
+
{
|
127 |
+
"name": "text_encoder.transformer_encoder.layers.0.self_attn.in_proj_weight",
|
128 |
+
"shape": [
|
129 |
+
2304,
|
130 |
+
768
|
131 |
+
],
|
132 |
+
"dtype": "torch.float32"
|
133 |
+
},
|
134 |
+
{
|
135 |
+
"name": "text_encoder.transformer_encoder.layers.0.self_attn.in_proj_bias",
|
136 |
+
"shape": [
|
137 |
+
2304
|
138 |
+
],
|
139 |
+
"dtype": "torch.float32"
|
140 |
+
},
|
141 |
+
{
|
142 |
+
"name": "text_encoder.transformer_encoder.layers.0.self_attn.out_proj.weight",
|
143 |
+
"shape": [
|
144 |
+
768,
|
145 |
+
768
|
146 |
+
],
|
147 |
+
"dtype": "torch.float32"
|
148 |
+
},
|
149 |
+
{
|
150 |
+
"name": "text_encoder.transformer_encoder.layers.0.self_attn.out_proj.bias",
|
151 |
+
"shape": [
|
152 |
+
768
|
153 |
+
],
|
154 |
+
"dtype": "torch.float32"
|
155 |
+
},
|
156 |
+
{
|
157 |
+
"name": "text_encoder.transformer_encoder.layers.0.linear1.weight",
|
158 |
+
"shape": [
|
159 |
+
2048,
|
160 |
+
768
|
161 |
+
],
|
162 |
+
"dtype": "torch.float32"
|
163 |
+
},
|
164 |
+
{
|
165 |
+
"name": "text_encoder.transformer_encoder.layers.0.linear1.bias",
|
166 |
+
"shape": [
|
167 |
+
2048
|
168 |
+
],
|
169 |
+
"dtype": "torch.float32"
|
170 |
+
},
|
171 |
+
{
|
172 |
+
"name": "text_encoder.transformer_encoder.layers.0.linear2.weight",
|
173 |
+
"shape": [
|
174 |
+
768,
|
175 |
+
2048
|
176 |
+
],
|
177 |
+
"dtype": "torch.float32"
|
178 |
+
},
|
179 |
+
{
|
180 |
+
"name": "text_encoder.transformer_encoder.layers.0.linear2.bias",
|
181 |
+
"shape": [
|
182 |
+
768
|
183 |
+
],
|
184 |
+
"dtype": "torch.float32"
|
185 |
+
},
|
186 |
+
{
|
187 |
+
"name": "text_encoder.transformer_encoder.layers.0.norm1.weight",
|
188 |
+
"shape": [
|
189 |
+
768
|
190 |
+
],
|
191 |
+
"dtype": "torch.float32"
|
192 |
+
},
|
193 |
+
{
|
194 |
+
"name": "text_encoder.transformer_encoder.layers.0.norm1.bias",
|
195 |
+
"shape": [
|
196 |
+
768
|
197 |
+
],
|
198 |
+
"dtype": "torch.float32"
|
199 |
+
},
|
200 |
+
{
|
201 |
+
"name": "text_encoder.transformer_encoder.layers.0.norm2.weight",
|
202 |
+
"shape": [
|
203 |
+
768
|
204 |
+
],
|
205 |
+
"dtype": "torch.float32"
|
206 |
+
},
|
207 |
+
{
|
208 |
+
"name": "text_encoder.transformer_encoder.layers.0.norm2.bias",
|
209 |
+
"shape": [
|
210 |
+
768
|
211 |
+
],
|
212 |
+
"dtype": "torch.float32"
|
213 |
+
},
|
214 |
+
{
|
215 |
+
"name": "text_encoder.transformer_encoder.layers.1.self_attn.in_proj_weight",
|
216 |
+
"shape": [
|
217 |
+
2304,
|
218 |
+
768
|
219 |
+
],
|
220 |
+
"dtype": "torch.float32"
|
221 |
+
},
|
222 |
+
{
|
223 |
+
"name": "text_encoder.transformer_encoder.layers.1.self_attn.in_proj_bias",
|
224 |
+
"shape": [
|
225 |
+
2304
|
226 |
+
],
|
227 |
+
"dtype": "torch.float32"
|
228 |
+
},
|
229 |
+
{
|
230 |
+
"name": "text_encoder.transformer_encoder.layers.1.self_attn.out_proj.weight",
|
231 |
+
"shape": [
|
232 |
+
768,
|
233 |
+
768
|
234 |
+
],
|
235 |
+
"dtype": "torch.float32"
|
236 |
+
},
|
237 |
+
{
|
238 |
+
"name": "text_encoder.transformer_encoder.layers.1.self_attn.out_proj.bias",
|
239 |
+
"shape": [
|
240 |
+
768
|
241 |
+
],
|
242 |
+
"dtype": "torch.float32"
|
243 |
+
},
|
244 |
+
{
|
245 |
+
"name": "text_encoder.transformer_encoder.layers.1.linear1.weight",
|
246 |
+
"shape": [
|
247 |
+
2048,
|
248 |
+
768
|
249 |
+
],
|
250 |
+
"dtype": "torch.float32"
|
251 |
+
},
|
252 |
+
{
|
253 |
+
"name": "text_encoder.transformer_encoder.layers.1.linear1.bias",
|
254 |
+
"shape": [
|
255 |
+
2048
|
256 |
+
],
|
257 |
+
"dtype": "torch.float32"
|
258 |
+
},
|
259 |
+
{
|
260 |
+
"name": "text_encoder.transformer_encoder.layers.1.linear2.weight",
|
261 |
+
"shape": [
|
262 |
+
768,
|
263 |
+
2048
|
264 |
+
],
|
265 |
+
"dtype": "torch.float32"
|
266 |
+
},
|
267 |
+
{
|
268 |
+
"name": "text_encoder.transformer_encoder.layers.1.linear2.bias",
|
269 |
+
"shape": [
|
270 |
+
768
|
271 |
+
],
|
272 |
+
"dtype": "torch.float32"
|
273 |
+
},
|
274 |
+
{
|
275 |
+
"name": "text_encoder.transformer_encoder.layers.1.norm1.weight",
|
276 |
+
"shape": [
|
277 |
+
768
|
278 |
+
],
|
279 |
+
"dtype": "torch.float32"
|
280 |
+
},
|
281 |
+
{
|
282 |
+
"name": "text_encoder.transformer_encoder.layers.1.norm1.bias",
|
283 |
+
"shape": [
|
284 |
+
768
|
285 |
+
],
|
286 |
+
"dtype": "torch.float32"
|
287 |
+
},
|
288 |
+
{
|
289 |
+
"name": "text_encoder.transformer_encoder.layers.1.norm2.weight",
|
290 |
+
"shape": [
|
291 |
+
768
|
292 |
+
],
|
293 |
+
"dtype": "torch.float32"
|
294 |
+
},
|
295 |
+
{
|
296 |
+
"name": "text_encoder.transformer_encoder.layers.1.norm2.bias",
|
297 |
+
"shape": [
|
298 |
+
768
|
299 |
+
],
|
300 |
+
"dtype": "torch.float32"
|
301 |
+
},
|
302 |
+
{
|
303 |
+
"name": "text_encoder.transformer_encoder.layers.2.self_attn.in_proj_weight",
|
304 |
+
"shape": [
|
305 |
+
2304,
|
306 |
+
768
|
307 |
+
],
|
308 |
+
"dtype": "torch.float32"
|
309 |
+
},
|
310 |
+
{
|
311 |
+
"name": "text_encoder.transformer_encoder.layers.2.self_attn.in_proj_bias",
|
312 |
+
"shape": [
|
313 |
+
2304
|
314 |
+
],
|
315 |
+
"dtype": "torch.float32"
|
316 |
+
},
|
317 |
+
{
|
318 |
+
"name": "text_encoder.transformer_encoder.layers.2.self_attn.out_proj.weight",
|
319 |
+
"shape": [
|
320 |
+
768,
|
321 |
+
768
|
322 |
+
],
|
323 |
+
"dtype": "torch.float32"
|
324 |
+
},
|
325 |
+
{
|
326 |
+
"name": "text_encoder.transformer_encoder.layers.2.self_attn.out_proj.bias",
|
327 |
+
"shape": [
|
328 |
+
768
|
329 |
+
],
|
330 |
+
"dtype": "torch.float32"
|
331 |
+
},
|
332 |
+
{
|
333 |
+
"name": "text_encoder.transformer_encoder.layers.2.linear1.weight",
|
334 |
+
"shape": [
|
335 |
+
2048,
|
336 |
+
768
|
337 |
+
],
|
338 |
+
"dtype": "torch.float32"
|
339 |
+
},
|
340 |
+
{
|
341 |
+
"name": "text_encoder.transformer_encoder.layers.2.linear1.bias",
|
342 |
+
"shape": [
|
343 |
+
2048
|
344 |
+
],
|
345 |
+
"dtype": "torch.float32"
|
346 |
+
},
|
347 |
+
{
|
348 |
+
"name": "text_encoder.transformer_encoder.layers.2.linear2.weight",
|
349 |
+
"shape": [
|
350 |
+
768,
|
351 |
+
2048
|
352 |
+
],
|
353 |
+
"dtype": "torch.float32"
|
354 |
+
},
|
355 |
+
{
|
356 |
+
"name": "text_encoder.transformer_encoder.layers.2.linear2.bias",
|
357 |
+
"shape": [
|
358 |
+
768
|
359 |
+
],
|
360 |
+
"dtype": "torch.float32"
|
361 |
+
},
|
362 |
+
{
|
363 |
+
"name": "text_encoder.transformer_encoder.layers.2.norm1.weight",
|
364 |
+
"shape": [
|
365 |
+
768
|
366 |
+
],
|
367 |
+
"dtype": "torch.float32"
|
368 |
+
},
|
369 |
+
{
|
370 |
+
"name": "text_encoder.transformer_encoder.layers.2.norm1.bias",
|
371 |
+
"shape": [
|
372 |
+
768
|
373 |
+
],
|
374 |
+
"dtype": "torch.float32"
|
375 |
+
},
|
376 |
+
{
|
377 |
+
"name": "text_encoder.transformer_encoder.layers.2.norm2.weight",
|
378 |
+
"shape": [
|
379 |
+
768
|
380 |
+
],
|
381 |
+
"dtype": "torch.float32"
|
382 |
+
},
|
383 |
+
{
|
384 |
+
"name": "text_encoder.transformer_encoder.layers.2.norm2.bias",
|
385 |
+
"shape": [
|
386 |
+
768
|
387 |
+
],
|
388 |
+
"dtype": "torch.float32"
|
389 |
+
},
|
390 |
+
{
|
391 |
+
"name": "text_encoder.transformer_encoder.layers.3.self_attn.in_proj_weight",
|
392 |
+
"shape": [
|
393 |
+
2304,
|
394 |
+
768
|
395 |
+
],
|
396 |
+
"dtype": "torch.float32"
|
397 |
+
},
|
398 |
+
{
|
399 |
+
"name": "text_encoder.transformer_encoder.layers.3.self_attn.in_proj_bias",
|
400 |
+
"shape": [
|
401 |
+
2304
|
402 |
+
],
|
403 |
+
"dtype": "torch.float32"
|
404 |
+
},
|
405 |
+
{
|
406 |
+
"name": "text_encoder.transformer_encoder.layers.3.self_attn.out_proj.weight",
|
407 |
+
"shape": [
|
408 |
+
768,
|
409 |
+
768
|
410 |
+
],
|
411 |
+
"dtype": "torch.float32"
|
412 |
+
},
|
413 |
+
{
|
414 |
+
"name": "text_encoder.transformer_encoder.layers.3.self_attn.out_proj.bias",
|
415 |
+
"shape": [
|
416 |
+
768
|
417 |
+
],
|
418 |
+
"dtype": "torch.float32"
|
419 |
+
},
|
420 |
+
{
|
421 |
+
"name": "text_encoder.transformer_encoder.layers.3.linear1.weight",
|
422 |
+
"shape": [
|
423 |
+
2048,
|
424 |
+
768
|
425 |
+
],
|
426 |
+
"dtype": "torch.float32"
|
427 |
+
},
|
428 |
+
{
|
429 |
+
"name": "text_encoder.transformer_encoder.layers.3.linear1.bias",
|
430 |
+
"shape": [
|
431 |
+
2048
|
432 |
+
],
|
433 |
+
"dtype": "torch.float32"
|
434 |
+
},
|
435 |
+
{
|
436 |
+
"name": "text_encoder.transformer_encoder.layers.3.linear2.weight",
|
437 |
+
"shape": [
|
438 |
+
768,
|
439 |
+
2048
|
440 |
+
],
|
441 |
+
"dtype": "torch.float32"
|
442 |
+
},
|
443 |
+
{
|
444 |
+
"name": "text_encoder.transformer_encoder.layers.3.linear2.bias",
|
445 |
+
"shape": [
|
446 |
+
768
|
447 |
+
],
|
448 |
+
"dtype": "torch.float32"
|
449 |
+
},
|
450 |
+
{
|
451 |
+
"name": "text_encoder.transformer_encoder.layers.3.norm1.weight",
|
452 |
+
"shape": [
|
453 |
+
768
|
454 |
+
],
|
455 |
+
"dtype": "torch.float32"
|
456 |
+
},
|
457 |
+
{
|
458 |
+
"name": "text_encoder.transformer_encoder.layers.3.norm1.bias",
|
459 |
+
"shape": [
|
460 |
+
768
|
461 |
+
],
|
462 |
+
"dtype": "torch.float32"
|
463 |
+
},
|
464 |
+
{
|
465 |
+
"name": "text_encoder.transformer_encoder.layers.3.norm2.weight",
|
466 |
+
"shape": [
|
467 |
+
768
|
468 |
+
],
|
469 |
+
"dtype": "torch.float32"
|
470 |
+
},
|
471 |
+
{
|
472 |
+
"name": "text_encoder.transformer_encoder.layers.3.norm2.bias",
|
473 |
+
"shape": [
|
474 |
+
768
|
475 |
+
],
|
476 |
+
"dtype": "torch.float32"
|
477 |
+
},
|
478 |
+
{
|
479 |
+
"name": "text_encoder.transformer_encoder.layers.4.self_attn.in_proj_weight",
|
480 |
+
"shape": [
|
481 |
+
2304,
|
482 |
+
768
|
483 |
+
],
|
484 |
+
"dtype": "torch.float32"
|
485 |
+
},
|
486 |
+
{
|
487 |
+
"name": "text_encoder.transformer_encoder.layers.4.self_attn.in_proj_bias",
|
488 |
+
"shape": [
|
489 |
+
2304
|
490 |
+
],
|
491 |
+
"dtype": "torch.float32"
|
492 |
+
},
|
493 |
+
{
|
494 |
+
"name": "text_encoder.transformer_encoder.layers.4.self_attn.out_proj.weight",
|
495 |
+
"shape": [
|
496 |
+
768,
|
497 |
+
768
|
498 |
+
],
|
499 |
+
"dtype": "torch.float32"
|
500 |
+
},
|
501 |
+
{
|
502 |
+
"name": "text_encoder.transformer_encoder.layers.4.self_attn.out_proj.bias",
|
503 |
+
"shape": [
|
504 |
+
768
|
505 |
+
],
|
506 |
+
"dtype": "torch.float32"
|
507 |
+
},
|
508 |
+
{
|
509 |
+
"name": "text_encoder.transformer_encoder.layers.4.linear1.weight",
|
510 |
+
"shape": [
|
511 |
+
2048,
|
512 |
+
768
|
513 |
+
],
|
514 |
+
"dtype": "torch.float32"
|
515 |
+
},
|
516 |
+
{
|
517 |
+
"name": "text_encoder.transformer_encoder.layers.4.linear1.bias",
|
518 |
+
"shape": [
|
519 |
+
2048
|
520 |
+
],
|
521 |
+
"dtype": "torch.float32"
|
522 |
+
},
|
523 |
+
{
|
524 |
+
"name": "text_encoder.transformer_encoder.layers.4.linear2.weight",
|
525 |
+
"shape": [
|
526 |
+
768,
|
527 |
+
2048
|
528 |
+
],
|
529 |
+
"dtype": "torch.float32"
|
530 |
+
},
|
531 |
+
{
|
532 |
+
"name": "text_encoder.transformer_encoder.layers.4.linear2.bias",
|
533 |
+
"shape": [
|
534 |
+
768
|
535 |
+
],
|
536 |
+
"dtype": "torch.float32"
|
537 |
+
},
|
538 |
+
{
|
539 |
+
"name": "text_encoder.transformer_encoder.layers.4.norm1.weight",
|
540 |
+
"shape": [
|
541 |
+
768
|
542 |
+
],
|
543 |
+
"dtype": "torch.float32"
|
544 |
+
},
|
545 |
+
{
|
546 |
+
"name": "text_encoder.transformer_encoder.layers.4.norm1.bias",
|
547 |
+
"shape": [
|
548 |
+
768
|
549 |
+
],
|
550 |
+
"dtype": "torch.float32"
|
551 |
+
},
|
552 |
+
{
|
553 |
+
"name": "text_encoder.transformer_encoder.layers.4.norm2.weight",
|
554 |
+
"shape": [
|
555 |
+
768
|
556 |
+
],
|
557 |
+
"dtype": "torch.float32"
|
558 |
+
},
|
559 |
+
{
|
560 |
+
"name": "text_encoder.transformer_encoder.layers.4.norm2.bias",
|
561 |
+
"shape": [
|
562 |
+
768
|
563 |
+
],
|
564 |
+
"dtype": "torch.float32"
|
565 |
+
},
|
566 |
+
{
|
567 |
+
"name": "text_encoder.transformer_encoder.layers.5.self_attn.in_proj_weight",
|
568 |
+
"shape": [
|
569 |
+
2304,
|
570 |
+
768
|
571 |
+
],
|
572 |
+
"dtype": "torch.float32"
|
573 |
+
},
|
574 |
+
{
|
575 |
+
"name": "text_encoder.transformer_encoder.layers.5.self_attn.in_proj_bias",
|
576 |
+
"shape": [
|
577 |
+
2304
|
578 |
+
],
|
579 |
+
"dtype": "torch.float32"
|
580 |
+
},
|
581 |
+
{
|
582 |
+
"name": "text_encoder.transformer_encoder.layers.5.self_attn.out_proj.weight",
|
583 |
+
"shape": [
|
584 |
+
768,
|
585 |
+
768
|
586 |
+
],
|
587 |
+
"dtype": "torch.float32"
|
588 |
+
},
|
589 |
+
{
|
590 |
+
"name": "text_encoder.transformer_encoder.layers.5.self_attn.out_proj.bias",
|
591 |
+
"shape": [
|
592 |
+
768
|
593 |
+
],
|
594 |
+
"dtype": "torch.float32"
|
595 |
+
},
|
596 |
+
{
|
597 |
+
"name": "text_encoder.transformer_encoder.layers.5.linear1.weight",
|
598 |
+
"shape": [
|
599 |
+
2048,
|
600 |
+
768
|
601 |
+
],
|
602 |
+
"dtype": "torch.float32"
|
603 |
+
},
|
604 |
+
{
|
605 |
+
"name": "text_encoder.transformer_encoder.layers.5.linear1.bias",
|
606 |
+
"shape": [
|
607 |
+
2048
|
608 |
+
],
|
609 |
+
"dtype": "torch.float32"
|
610 |
+
},
|
611 |
+
{
|
612 |
+
"name": "text_encoder.transformer_encoder.layers.5.linear2.weight",
|
613 |
+
"shape": [
|
614 |
+
768,
|
615 |
+
2048
|
616 |
+
],
|
617 |
+
"dtype": "torch.float32"
|
618 |
+
},
|
619 |
+
{
|
620 |
+
"name": "text_encoder.transformer_encoder.layers.5.linear2.bias",
|
621 |
+
"shape": [
|
622 |
+
768
|
623 |
+
],
|
624 |
+
"dtype": "torch.float32"
|
625 |
+
},
|
626 |
+
{
|
627 |
+
"name": "text_encoder.transformer_encoder.layers.5.norm1.weight",
|
628 |
+
"shape": [
|
629 |
+
768
|
630 |
+
],
|
631 |
+
"dtype": "torch.float32"
|
632 |
+
},
|
633 |
+
{
|
634 |
+
"name": "text_encoder.transformer_encoder.layers.5.norm1.bias",
|
635 |
+
"shape": [
|
636 |
+
768
|
637 |
+
],
|
638 |
+
"dtype": "torch.float32"
|
639 |
+
},
|
640 |
+
{
|
641 |
+
"name": "text_encoder.transformer_encoder.layers.5.norm2.weight",
|
642 |
+
"shape": [
|
643 |
+
768
|
644 |
+
],
|
645 |
+
"dtype": "torch.float32"
|
646 |
+
},
|
647 |
+
{
|
648 |
+
"name": "text_encoder.transformer_encoder.layers.5.norm2.bias",
|
649 |
+
"shape": [
|
650 |
+
768
|
651 |
+
],
|
652 |
+
"dtype": "torch.float32"
|
653 |
+
},
|
654 |
+
{
|
655 |
+
"name": "text_encoder.transformer_encoder.layers.6.self_attn.in_proj_weight",
|
656 |
+
"shape": [
|
657 |
+
2304,
|
658 |
+
768
|
659 |
+
],
|
660 |
+
"dtype": "torch.float32"
|
661 |
+
},
|
662 |
+
{
|
663 |
+
"name": "text_encoder.transformer_encoder.layers.6.self_attn.in_proj_bias",
|
664 |
+
"shape": [
|
665 |
+
2304
|
666 |
+
],
|
667 |
+
"dtype": "torch.float32"
|
668 |
+
},
|
669 |
+
{
|
670 |
+
"name": "text_encoder.transformer_encoder.layers.6.self_attn.out_proj.weight",
|
671 |
+
"shape": [
|
672 |
+
768,
|
673 |
+
768
|
674 |
+
],
|
675 |
+
"dtype": "torch.float32"
|
676 |
+
},
|
677 |
+
{
|
678 |
+
"name": "text_encoder.transformer_encoder.layers.6.self_attn.out_proj.bias",
|
679 |
+
"shape": [
|
680 |
+
768
|
681 |
+
],
|
682 |
+
"dtype": "torch.float32"
|
683 |
+
},
|
684 |
+
{
|
685 |
+
"name": "text_encoder.transformer_encoder.layers.6.linear1.weight",
|
686 |
+
"shape": [
|
687 |
+
2048,
|
688 |
+
768
|
689 |
+
],
|
690 |
+
"dtype": "torch.float32"
|
691 |
+
},
|
692 |
+
{
|
693 |
+
"name": "text_encoder.transformer_encoder.layers.6.linear1.bias",
|
694 |
+
"shape": [
|
695 |
+
2048
|
696 |
+
],
|
697 |
+
"dtype": "torch.float32"
|
698 |
+
},
|
699 |
+
{
|
700 |
+
"name": "text_encoder.transformer_encoder.layers.6.linear2.weight",
|
701 |
+
"shape": [
|
702 |
+
768,
|
703 |
+
2048
|
704 |
+
],
|
705 |
+
"dtype": "torch.float32"
|
706 |
+
},
|
707 |
+
{
|
708 |
+
"name": "text_encoder.transformer_encoder.layers.6.linear2.bias",
|
709 |
+
"shape": [
|
710 |
+
768
|
711 |
+
],
|
712 |
+
"dtype": "torch.float32"
|
713 |
+
},
|
714 |
+
{
|
715 |
+
"name": "text_encoder.transformer_encoder.layers.6.norm1.weight",
|
716 |
+
"shape": [
|
717 |
+
768
|
718 |
+
],
|
719 |
+
"dtype": "torch.float32"
|
720 |
+
},
|
721 |
+
{
|
722 |
+
"name": "text_encoder.transformer_encoder.layers.6.norm1.bias",
|
723 |
+
"shape": [
|
724 |
+
768
|
725 |
+
],
|
726 |
+
"dtype": "torch.float32"
|
727 |
+
},
|
728 |
+
{
|
729 |
+
"name": "text_encoder.transformer_encoder.layers.6.norm2.weight",
|
730 |
+
"shape": [
|
731 |
+
768
|
732 |
+
],
|
733 |
+
"dtype": "torch.float32"
|
734 |
+
},
|
735 |
+
{
|
736 |
+
"name": "text_encoder.transformer_encoder.layers.6.norm2.bias",
|
737 |
+
"shape": [
|
738 |
+
768
|
739 |
+
],
|
740 |
+
"dtype": "torch.float32"
|
741 |
+
},
|
742 |
+
{
|
743 |
+
"name": "text_encoder.transformer_encoder.layers.7.self_attn.in_proj_weight",
|
744 |
+
"shape": [
|
745 |
+
2304,
|
746 |
+
768
|
747 |
+
],
|
748 |
+
"dtype": "torch.float32"
|
749 |
+
},
|
750 |
+
{
|
751 |
+
"name": "text_encoder.transformer_encoder.layers.7.self_attn.in_proj_bias",
|
752 |
+
"shape": [
|
753 |
+
2304
|
754 |
+
],
|
755 |
+
"dtype": "torch.float32"
|
756 |
+
},
|
757 |
+
{
|
758 |
+
"name": "text_encoder.transformer_encoder.layers.7.self_attn.out_proj.weight",
|
759 |
+
"shape": [
|
760 |
+
768,
|
761 |
+
768
|
762 |
+
],
|
763 |
+
"dtype": "torch.float32"
|
764 |
+
},
|
765 |
+
{
|
766 |
+
"name": "text_encoder.transformer_encoder.layers.7.self_attn.out_proj.bias",
|
767 |
+
"shape": [
|
768 |
+
768
|
769 |
+
],
|
770 |
+
"dtype": "torch.float32"
|
771 |
+
},
|
772 |
+
{
|
773 |
+
"name": "text_encoder.transformer_encoder.layers.7.linear1.weight",
|
774 |
+
"shape": [
|
775 |
+
2048,
|
776 |
+
768
|
777 |
+
],
|
778 |
+
"dtype": "torch.float32"
|
779 |
+
},
|
780 |
+
{
|
781 |
+
"name": "text_encoder.transformer_encoder.layers.7.linear1.bias",
|
782 |
+
"shape": [
|
783 |
+
2048
|
784 |
+
],
|
785 |
+
"dtype": "torch.float32"
|
786 |
+
},
|
787 |
+
{
|
788 |
+
"name": "text_encoder.transformer_encoder.layers.7.linear2.weight",
|
789 |
+
"shape": [
|
790 |
+
768,
|
791 |
+
2048
|
792 |
+
],
|
793 |
+
"dtype": "torch.float32"
|
794 |
+
},
|
795 |
+
{
|
796 |
+
"name": "text_encoder.transformer_encoder.layers.7.linear2.bias",
|
797 |
+
"shape": [
|
798 |
+
768
|
799 |
+
],
|
800 |
+
"dtype": "torch.float32"
|
801 |
+
},
|
802 |
+
{
|
803 |
+
"name": "text_encoder.transformer_encoder.layers.7.norm1.weight",
|
804 |
+
"shape": [
|
805 |
+
768
|
806 |
+
],
|
807 |
+
"dtype": "torch.float32"
|
808 |
+
},
|
809 |
+
{
|
810 |
+
"name": "text_encoder.transformer_encoder.layers.7.norm1.bias",
|
811 |
+
"shape": [
|
812 |
+
768
|
813 |
+
],
|
814 |
+
"dtype": "torch.float32"
|
815 |
+
},
|
816 |
+
{
|
817 |
+
"name": "text_encoder.transformer_encoder.layers.7.norm2.weight",
|
818 |
+
"shape": [
|
819 |
+
768
|
820 |
+
],
|
821 |
+
"dtype": "torch.float32"
|
822 |
+
},
|
823 |
+
{
|
824 |
+
"name": "text_encoder.transformer_encoder.layers.7.norm2.bias",
|
825 |
+
"shape": [
|
826 |
+
768
|
827 |
+
],
|
828 |
+
"dtype": "torch.float32"
|
829 |
+
},
|
830 |
+
{
|
831 |
+
"name": "text_encoder.transformer_encoder.layers.8.self_attn.in_proj_weight",
|
832 |
+
"shape": [
|
833 |
+
2304,
|
834 |
+
768
|
835 |
+
],
|
836 |
+
"dtype": "torch.float32"
|
837 |
+
},
|
838 |
+
{
|
839 |
+
"name": "text_encoder.transformer_encoder.layers.8.self_attn.in_proj_bias",
|
840 |
+
"shape": [
|
841 |
+
2304
|
842 |
+
],
|
843 |
+
"dtype": "torch.float32"
|
844 |
+
},
|
845 |
+
{
|
846 |
+
"name": "text_encoder.transformer_encoder.layers.8.self_attn.out_proj.weight",
|
847 |
+
"shape": [
|
848 |
+
768,
|
849 |
+
768
|
850 |
+
],
|
851 |
+
"dtype": "torch.float32"
|
852 |
+
},
|
853 |
+
{
|
854 |
+
"name": "text_encoder.transformer_encoder.layers.8.self_attn.out_proj.bias",
|
855 |
+
"shape": [
|
856 |
+
768
|
857 |
+
],
|
858 |
+
"dtype": "torch.float32"
|
859 |
+
},
|
860 |
+
{
|
861 |
+
"name": "text_encoder.transformer_encoder.layers.8.linear1.weight",
|
862 |
+
"shape": [
|
863 |
+
2048,
|
864 |
+
768
|
865 |
+
],
|
866 |
+
"dtype": "torch.float32"
|
867 |
+
},
|
868 |
+
{
|
869 |
+
"name": "text_encoder.transformer_encoder.layers.8.linear1.bias",
|
870 |
+
"shape": [
|
871 |
+
2048
|
872 |
+
],
|
873 |
+
"dtype": "torch.float32"
|
874 |
+
},
|
875 |
+
{
|
876 |
+
"name": "text_encoder.transformer_encoder.layers.8.linear2.weight",
|
877 |
+
"shape": [
|
878 |
+
768,
|
879 |
+
2048
|
880 |
+
],
|
881 |
+
"dtype": "torch.float32"
|
882 |
+
},
|
883 |
+
{
|
884 |
+
"name": "text_encoder.transformer_encoder.layers.8.linear2.bias",
|
885 |
+
"shape": [
|
886 |
+
768
|
887 |
+
],
|
888 |
+
"dtype": "torch.float32"
|
889 |
+
},
|
890 |
+
{
|
891 |
+
"name": "text_encoder.transformer_encoder.layers.8.norm1.weight",
|
892 |
+
"shape": [
|
893 |
+
768
|
894 |
+
],
|
895 |
+
"dtype": "torch.float32"
|
896 |
+
},
|
897 |
+
{
|
898 |
+
"name": "text_encoder.transformer_encoder.layers.8.norm1.bias",
|
899 |
+
"shape": [
|
900 |
+
768
|
901 |
+
],
|
902 |
+
"dtype": "torch.float32"
|
903 |
+
},
|
904 |
+
{
|
905 |
+
"name": "text_encoder.transformer_encoder.layers.8.norm2.weight",
|
906 |
+
"shape": [
|
907 |
+
768
|
908 |
+
],
|
909 |
+
"dtype": "torch.float32"
|
910 |
+
},
|
911 |
+
{
|
912 |
+
"name": "text_encoder.transformer_encoder.layers.8.norm2.bias",
|
913 |
+
"shape": [
|
914 |
+
768
|
915 |
+
],
|
916 |
+
"dtype": "torch.float32"
|
917 |
+
},
|
918 |
+
{
|
919 |
+
"name": "text_encoder.transformer_encoder.layers.9.self_attn.in_proj_weight",
|
920 |
+
"shape": [
|
921 |
+
2304,
|
922 |
+
768
|
923 |
+
],
|
924 |
+
"dtype": "torch.float32"
|
925 |
+
},
|
926 |
+
{
|
927 |
+
"name": "text_encoder.transformer_encoder.layers.9.self_attn.in_proj_bias",
|
928 |
+
"shape": [
|
929 |
+
2304
|
930 |
+
],
|
931 |
+
"dtype": "torch.float32"
|
932 |
+
},
|
933 |
+
{
|
934 |
+
"name": "text_encoder.transformer_encoder.layers.9.self_attn.out_proj.weight",
|
935 |
+
"shape": [
|
936 |
+
768,
|
937 |
+
768
|
938 |
+
],
|
939 |
+
"dtype": "torch.float32"
|
940 |
+
},
|
941 |
+
{
|
942 |
+
"name": "text_encoder.transformer_encoder.layers.9.self_attn.out_proj.bias",
|
943 |
+
"shape": [
|
944 |
+
768
|
945 |
+
],
|
946 |
+
"dtype": "torch.float32"
|
947 |
+
},
|
948 |
+
{
|
949 |
+
"name": "text_encoder.transformer_encoder.layers.9.linear1.weight",
|
950 |
+
"shape": [
|
951 |
+
2048,
|
952 |
+
768
|
953 |
+
],
|
954 |
+
"dtype": "torch.float32"
|
955 |
+
},
|
956 |
+
{
|
957 |
+
"name": "text_encoder.transformer_encoder.layers.9.linear1.bias",
|
958 |
+
"shape": [
|
959 |
+
2048
|
960 |
+
],
|
961 |
+
"dtype": "torch.float32"
|
962 |
+
},
|
963 |
+
{
|
964 |
+
"name": "text_encoder.transformer_encoder.layers.9.linear2.weight",
|
965 |
+
"shape": [
|
966 |
+
768,
|
967 |
+
2048
|
968 |
+
],
|
969 |
+
"dtype": "torch.float32"
|
970 |
+
},
|
971 |
+
{
|
972 |
+
"name": "text_encoder.transformer_encoder.layers.9.linear2.bias",
|
973 |
+
"shape": [
|
974 |
+
768
|
975 |
+
],
|
976 |
+
"dtype": "torch.float32"
|
977 |
+
},
|
978 |
+
{
|
979 |
+
"name": "text_encoder.transformer_encoder.layers.9.norm1.weight",
|
980 |
+
"shape": [
|
981 |
+
768
|
982 |
+
],
|
983 |
+
"dtype": "torch.float32"
|
984 |
+
},
|
985 |
+
{
|
986 |
+
"name": "text_encoder.transformer_encoder.layers.9.norm1.bias",
|
987 |
+
"shape": [
|
988 |
+
768
|
989 |
+
],
|
990 |
+
"dtype": "torch.float32"
|
991 |
+
},
|
992 |
+
{
|
993 |
+
"name": "text_encoder.transformer_encoder.layers.9.norm2.weight",
|
994 |
+
"shape": [
|
995 |
+
768
|
996 |
+
],
|
997 |
+
"dtype": "torch.float32"
|
998 |
+
},
|
999 |
+
{
|
1000 |
+
"name": "text_encoder.transformer_encoder.layers.9.norm2.bias",
|
1001 |
+
"shape": [
|
1002 |
+
768
|
1003 |
+
],
|
1004 |
+
"dtype": "torch.float32"
|
1005 |
+
},
|
1006 |
+
{
|
1007 |
+
"name": "text_encoder.transformer_encoder.layers.10.self_attn.in_proj_weight",
|
1008 |
+
"shape": [
|
1009 |
+
2304,
|
1010 |
+
768
|
1011 |
+
],
|
1012 |
+
"dtype": "torch.float32"
|
1013 |
+
},
|
1014 |
+
{
|
1015 |
+
"name": "text_encoder.transformer_encoder.layers.10.self_attn.in_proj_bias",
|
1016 |
+
"shape": [
|
1017 |
+
2304
|
1018 |
+
],
|
1019 |
+
"dtype": "torch.float32"
|
1020 |
+
},
|
1021 |
+
{
|
1022 |
+
"name": "text_encoder.transformer_encoder.layers.10.self_attn.out_proj.weight",
|
1023 |
+
"shape": [
|
1024 |
+
768,
|
1025 |
+
768
|
1026 |
+
],
|
1027 |
+
"dtype": "torch.float32"
|
1028 |
+
},
|
1029 |
+
{
|
1030 |
+
"name": "text_encoder.transformer_encoder.layers.10.self_attn.out_proj.bias",
|
1031 |
+
"shape": [
|
1032 |
+
768
|
1033 |
+
],
|
1034 |
+
"dtype": "torch.float32"
|
1035 |
+
},
|
1036 |
+
{
|
1037 |
+
"name": "text_encoder.transformer_encoder.layers.10.linear1.weight",
|
1038 |
+
"shape": [
|
1039 |
+
2048,
|
1040 |
+
768
|
1041 |
+
],
|
1042 |
+
"dtype": "torch.float32"
|
1043 |
+
},
|
1044 |
+
{
|
1045 |
+
"name": "text_encoder.transformer_encoder.layers.10.linear1.bias",
|
1046 |
+
"shape": [
|
1047 |
+
2048
|
1048 |
+
],
|
1049 |
+
"dtype": "torch.float32"
|
1050 |
+
},
|
1051 |
+
{
|
1052 |
+
"name": "text_encoder.transformer_encoder.layers.10.linear2.weight",
|
1053 |
+
"shape": [
|
1054 |
+
768,
|
1055 |
+
2048
|
1056 |
+
],
|
1057 |
+
"dtype": "torch.float32"
|
1058 |
+
},
|
1059 |
+
{
|
1060 |
+
"name": "text_encoder.transformer_encoder.layers.10.linear2.bias",
|
1061 |
+
"shape": [
|
1062 |
+
768
|
1063 |
+
],
|
1064 |
+
"dtype": "torch.float32"
|
1065 |
+
},
|
1066 |
+
{
|
1067 |
+
"name": "text_encoder.transformer_encoder.layers.10.norm1.weight",
|
1068 |
+
"shape": [
|
1069 |
+
768
|
1070 |
+
],
|
1071 |
+
"dtype": "torch.float32"
|
1072 |
+
},
|
1073 |
+
{
|
1074 |
+
"name": "text_encoder.transformer_encoder.layers.10.norm1.bias",
|
1075 |
+
"shape": [
|
1076 |
+
768
|
1077 |
+
],
|
1078 |
+
"dtype": "torch.float32"
|
1079 |
+
},
|
1080 |
+
{
|
1081 |
+
"name": "text_encoder.transformer_encoder.layers.10.norm2.weight",
|
1082 |
+
"shape": [
|
1083 |
+
768
|
1084 |
+
],
|
1085 |
+
"dtype": "torch.float32"
|
1086 |
+
},
|
1087 |
+
{
|
1088 |
+
"name": "text_encoder.transformer_encoder.layers.10.norm2.bias",
|
1089 |
+
"shape": [
|
1090 |
+
768
|
1091 |
+
],
|
1092 |
+
"dtype": "torch.float32"
|
1093 |
+
},
|
1094 |
+
{
|
1095 |
+
"name": "text_encoder.transformer_encoder.layers.11.self_attn.in_proj_weight",
|
1096 |
+
"shape": [
|
1097 |
+
2304,
|
1098 |
+
768
|
1099 |
+
],
|
1100 |
+
"dtype": "torch.float32"
|
1101 |
+
},
|
1102 |
+
{
|
1103 |
+
"name": "text_encoder.transformer_encoder.layers.11.self_attn.in_proj_bias",
|
1104 |
+
"shape": [
|
1105 |
+
2304
|
1106 |
+
],
|
1107 |
+
"dtype": "torch.float32"
|
1108 |
+
},
|
1109 |
+
{
|
1110 |
+
"name": "text_encoder.transformer_encoder.layers.11.self_attn.out_proj.weight",
|
1111 |
+
"shape": [
|
1112 |
+
768,
|
1113 |
+
768
|
1114 |
+
],
|
1115 |
+
"dtype": "torch.float32"
|
1116 |
+
},
|
1117 |
+
{
|
1118 |
+
"name": "text_encoder.transformer_encoder.layers.11.self_attn.out_proj.bias",
|
1119 |
+
"shape": [
|
1120 |
+
768
|
1121 |
+
],
|
1122 |
+
"dtype": "torch.float32"
|
1123 |
+
},
|
1124 |
+
{
|
1125 |
+
"name": "text_encoder.transformer_encoder.layers.11.linear1.weight",
|
1126 |
+
"shape": [
|
1127 |
+
2048,
|
1128 |
+
768
|
1129 |
+
],
|
1130 |
+
"dtype": "torch.float32"
|
1131 |
+
},
|
1132 |
+
{
|
1133 |
+
"name": "text_encoder.transformer_encoder.layers.11.linear1.bias",
|
1134 |
+
"shape": [
|
1135 |
+
2048
|
1136 |
+
],
|
1137 |
+
"dtype": "torch.float32"
|
1138 |
+
},
|
1139 |
+
{
|
1140 |
+
"name": "text_encoder.transformer_encoder.layers.11.linear2.weight",
|
1141 |
+
"shape": [
|
1142 |
+
768,
|
1143 |
+
2048
|
1144 |
+
],
|
1145 |
+
"dtype": "torch.float32"
|
1146 |
+
},
|
1147 |
+
{
|
1148 |
+
"name": "text_encoder.transformer_encoder.layers.11.linear2.bias",
|
1149 |
+
"shape": [
|
1150 |
+
768
|
1151 |
+
],
|
1152 |
+
"dtype": "torch.float32"
|
1153 |
+
},
|
1154 |
+
{
|
1155 |
+
"name": "text_encoder.transformer_encoder.layers.11.norm1.weight",
|
1156 |
+
"shape": [
|
1157 |
+
768
|
1158 |
+
],
|
1159 |
+
"dtype": "torch.float32"
|
1160 |
+
},
|
1161 |
+
{
|
1162 |
+
"name": "text_encoder.transformer_encoder.layers.11.norm1.bias",
|
1163 |
+
"shape": [
|
1164 |
+
768
|
1165 |
+
],
|
1166 |
+
"dtype": "torch.float32"
|
1167 |
+
},
|
1168 |
+
{
|
1169 |
+
"name": "text_encoder.transformer_encoder.layers.11.norm2.weight",
|
1170 |
+
"shape": [
|
1171 |
+
768
|
1172 |
+
],
|
1173 |
+
"dtype": "torch.float32"
|
1174 |
+
},
|
1175 |
+
{
|
1176 |
+
"name": "text_encoder.transformer_encoder.layers.11.norm2.bias",
|
1177 |
+
"shape": [
|
1178 |
+
768
|
1179 |
+
],
|
1180 |
+
"dtype": "torch.float32"
|
1181 |
+
},
|
1182 |
+
{
|
1183 |
+
"name": "audio_encoder.encoder_layer.0.weight",
|
1184 |
+
"shape": [
|
1185 |
+
768,
|
1186 |
+
16000
|
1187 |
+
],
|
1188 |
+
"dtype": "torch.float32"
|
1189 |
+
},
|
1190 |
+
{
|
1191 |
+
"name": "audio_encoder.encoder_layer.0.bias",
|
1192 |
+
"shape": [
|
1193 |
+
768
|
1194 |
+
],
|
1195 |
+
"dtype": "torch.float32"
|
1196 |
+
},
|
1197 |
+
{
|
1198 |
+
"name": "audio_encoder.encoder_layer.2.weight",
|
1199 |
+
"shape": [
|
1200 |
+
768,
|
1201 |
+
768
|
1202 |
+
],
|
1203 |
+
"dtype": "torch.float32"
|
1204 |
+
},
|
1205 |
+
{
|
1206 |
+
"name": "audio_encoder.encoder_layer.2.bias",
|
1207 |
+
"shape": [
|
1208 |
+
768
|
1209 |
+
],
|
1210 |
+
"dtype": "torch.float32"
|
1211 |
+
},
|
1212 |
+
{
|
1213 |
+
"name": "fusion_layer.fusion_layer.weight",
|
1214 |
+
"shape": [
|
1215 |
+
768,
|
1216 |
+
2304
|
1217 |
+
],
|
1218 |
+
"dtype": "torch.float32"
|
1219 |
+
},
|
1220 |
+
{
|
1221 |
+
"name": "fusion_layer.fusion_layer.bias",
|
1222 |
+
"shape": [
|
1223 |
+
768
|
1224 |
+
],
|
1225 |
+
"dtype": "torch.float32"
|
1226 |
+
},
|
1227 |
+
{
|
1228 |
+
"name": "vqa_layer.vqa_layer.weight",
|
1229 |
+
"shape": [
|
1230 |
+
30522,
|
1231 |
+
768
|
1232 |
+
],
|
1233 |
+
"dtype": "torch.float32"
|
1234 |
+
},
|
1235 |
+
{
|
1236 |
+
"name": "vqa_layer.vqa_layer.bias",
|
1237 |
+
"shape": [
|
1238 |
+
30522
|
1239 |
+
],
|
1240 |
+
"dtype": "torch.float32"
|
1241 |
+
},
|
1242 |
+
{
|
1243 |
+
"name": "caption_layer.caption_layer.weight",
|
1244 |
+
"shape": [
|
1245 |
+
30522,
|
1246 |
+
768
|
1247 |
+
],
|
1248 |
+
"dtype": "torch.float32"
|
1249 |
+
},
|
1250 |
+
{
|
1251 |
+
"name": "caption_layer.caption_layer.bias",
|
1252 |
+
"shape": [
|
1253 |
+
30522
|
1254 |
+
],
|
1255 |
+
"dtype": "torch.float32"
|
1256 |
+
},
|
1257 |
+
{
|
1258 |
+
"name": "retrieval_layer.retrieval_layer.weight",
|
1259 |
+
"shape": [
|
1260 |
+
30522,
|
1261 |
+
768
|
1262 |
+
],
|
1263 |
+
"dtype": "torch.float32"
|
1264 |
+
},
|
1265 |
+
{
|
1266 |
+
"name": "retrieval_layer.retrieval_layer.bias",
|
1267 |
+
"shape": [
|
1268 |
+
30522
|
1269 |
+
],
|
1270 |
+
"dtype": "torch.float32"
|
1271 |
+
},
|
1272 |
+
{
|
1273 |
+
"name": "asr_layer.asr_layer.weight",
|
1274 |
+
"shape": [
|
1275 |
+
30522,
|
1276 |
+
768
|
1277 |
+
],
|
1278 |
+
"dtype": "torch.float32"
|
1279 |
+
},
|
1280 |
+
{
|
1281 |
+
"name": "asr_layer.asr_layer.bias",
|
1282 |
+
"shape": [
|
1283 |
+
30522
|
1284 |
+
],
|
1285 |
+
"dtype": "torch.float32"
|
1286 |
+
},
|
1287 |
+
{
|
1288 |
+
"name": "realtime_asr_layer.realtime_asr_layer.weight",
|
1289 |
+
"shape": [
|
1290 |
+
30522,
|
1291 |
+
768
|
1292 |
+
],
|
1293 |
+
"dtype": "torch.float32"
|
1294 |
+
},
|
1295 |
+
{
|
1296 |
+
"name": "realtime_asr_layer.realtime_asr_layer.bias",
|
1297 |
+
"shape": [
|
1298 |
+
30522
|
1299 |
+
],
|
1300 |
+
"dtype": "torch.float32"
|
1301 |
+
}
|
1302 |
+
]
|
1303 |
+
},
|
1304 |
+
"file_info": {
|
1305 |
+
"path": "C:\\Users\\baby7\\Desktop\\fastAPI\\AutoModel.pth",
|
1306 |
+
"size": 3237240570,
|
1307 |
+
"last_modified": 1735983514.6732724
|
1308 |
+
}
|
1309 |
+
}
|