Night-Quiet commited on
Commit
18f889e
·
1 Parent(s): 1d4d96d

results push

Browse files
Files changed (43) hide show
  1. .gitattributes +5 -0
  2. results/check/audio_sample.joblib +3 -0
  3. results/check/audio_sample_list.joblib +3 -0
  4. results/check/encoder_out.joblib +3 -0
  5. results/check/model_input.joblib +3 -0
  6. results/check/origin_audio.joblib +3 -0
  7. results/check/origin_model_input.joblib +3 -0
  8. results/check/origin_model_output.joblib +3 -0
  9. results/check/origin_model_output1.joblib +3 -0
  10. results/check/sorted_data.joblib +3 -0
  11. results/check/speech.joblib +3 -0
  12. results/check/speech_j.joblib +3 -0
  13. results/draw/mat1.joblib +3 -0
  14. results/draw/mat2.joblib +3 -0
  15. results/draw/mat3.joblib +3 -0
  16. results/draw/mel_spectrogram_Y-2Bev4vU8Uw.png +0 -0
  17. results/draw/mel_spectrogram_Y-2Bev4vU8Uw_0.png +0 -0
  18. results/draw/mel_spectrogram_Y-2Bev4vU8Uw_1.png +0 -0
  19. results/draw/mel_spectrogram_Y-2Bev4vU8Uw_2.png +0 -0
  20. results/draw/mel_spectrogram_Y-2Bev4vU8Uw_output.png +0 -0
  21. results/track_record/bbox_cost.joblib +3 -0
  22. results/track_record/class_cost.joblib +3 -0
  23. results/track_record/collator_print_first.joblib +3 -0
  24. results/track_record/collator_print_first.json +262 -0
  25. results/track_record/enc_text_hidden_state.joblib +3 -0
  26. results/track_record/giou_cost.joblib +3 -0
  27. results/track_record/grad_count.json +1 -0
  28. results/track_record/grad_count_score.json +555 -0
  29. results/track_record/hidden_states.joblib +3 -0
  30. results/track_record/kaldi_fbank1.joblib +3 -0
  31. results/track_record/kaldi_fbank2.joblib +3 -0
  32. results/track_record/labels.joblib +3 -0
  33. results/track_record/load_print_first.joblib +3 -0
  34. results/track_record/load_print_first.json +3 -0
  35. results/track_record/map_print_first.joblib +3 -0
  36. results/track_record/map_print_first.json +132 -0
  37. results/track_record/outputs_loss.joblib +3 -0
  38. results/track_record/source_flatten.joblib +3 -0
  39. results/track_record/text_features.joblib +3 -0
  40. results/track_record/text_features_before.joblib +3 -0
  41. results/track_record/text_projection.joblib +3 -0
  42. results/track_record/waveform1.joblib +3 -0
  43. results/track_record/waveform2.joblib +3 -0
.gitattributes CHANGED
@@ -33,3 +33,8 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
33
  *.zip filter=lfs diff=lfs merge=lfs -text
34
  *.zst filter=lfs diff=lfs merge=lfs -text
35
  *tfevents* filter=lfs diff=lfs merge=lfs -text
 
 
 
 
 
 
33
  *.zip filter=lfs diff=lfs merge=lfs -text
34
  *.zst filter=lfs diff=lfs merge=lfs -text
35
  *tfevents* filter=lfs diff=lfs merge=lfs -text
36
+ results/check/speech.joblib filter=lfs diff=lfs merge=lfs -text
37
+ results/check/speech_j.joblib filter=lfs diff=lfs merge=lfs -text
38
+ results/track_record/collator_print_first.joblib filter=lfs diff=lfs merge=lfs -text
39
+ results/track_record/outputs_loss.joblib filter=lfs diff=lfs merge=lfs -text
40
+ results/track_record/text_projection.joblib filter=lfs diff=lfs merge=lfs -text
results/check/audio_sample.joblib ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:06cc5614ad14441b75d3878023567cea9a5e2e1925681a30c1faa98e6fd3fc8c
3
+ size 4510522
results/check/audio_sample_list.joblib ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:cfc5e5f448227c5fd005e459c23346acc4cc567ac9a12015c4006f0e086827b5
3
+ size 4510525
results/check/encoder_out.joblib ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9bbd017b31a7eab01c9a62301129c84e4f72221ea0470282d4f6b358a5164eb0
3
+ size 1661344
results/check/model_input.joblib ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3c4f41468a3bb873f2ce834d726af77f6f4abc3453a38690f6ce992faf58ac46
3
+ size 6129058
results/check/origin_audio.joblib ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:11dfa284c568674d5b8eda288e5b43ab1907ebc7979c450e16fccc886098e835
3
+ size 210968
results/check/origin_model_input.joblib ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2a9d1990e595cb14be10baa269fefa235b61a209f68e8751b8878543efadf1b0
3
+ size 7714973
results/check/origin_model_output.joblib ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:08efefa4a5b3d56a60342b12a7ce04a922be9d5850b2c2edc71d306d300faa1b
3
+ size 7152032
results/check/origin_model_output1.joblib ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:809629313222531078d71ae028d567cc4a2454e0955cbf55a0cab9d680e955da
3
+ size 7152032
results/check/sorted_data.joblib ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:84518242a14aa9844f49d8981cbabc76cadcccb901f0a91efd5521176d132fcc
3
+ size 1042
results/check/speech.joblib ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:63e2f64975b655c2fe2c8990c2093cc465879c544f586acbe84a623ff00053ec
3
+ size 11272413
results/check/speech_j.joblib ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:af413ec7a6d5af1490cea57477b589198e5a67cd0c685f038b31f097f8d06821
3
+ size 54125065
results/draw/mat1.joblib ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c72a22643f96f28089c1d0d2dd5c530d45b3180885079f88d790c943806ebf3d
3
+ size 639125
results/draw/mat2.joblib ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fcb8f8af7605a2074d7b17fda7c4f74e6a1176021cfbfe0dbd7b46d01c119781
3
+ size 374485
results/draw/mat3.joblib ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:86131bcb96a5b17d10e3e0931d7347339cee8b476602c1fdb487f27ee3843dad
3
+ size 374485
results/draw/mel_spectrogram_Y-2Bev4vU8Uw.png ADDED
results/draw/mel_spectrogram_Y-2Bev4vU8Uw_0.png ADDED
results/draw/mel_spectrogram_Y-2Bev4vU8Uw_1.png ADDED
results/draw/mel_spectrogram_Y-2Bev4vU8Uw_2.png ADDED
results/draw/mel_spectrogram_Y-2Bev4vU8Uw_output.png ADDED
results/track_record/bbox_cost.joblib ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:24dd9c77b4783197b190a80718fc11bd5d7d314fcea2df5e70e82184e3b17488
3
+ size 343612
results/track_record/class_cost.joblib ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:99feb0122c64dab4623220f6ee8c037a01c34ec99190fb81f1b15e74916fad1b
3
+ size 686809
results/track_record/collator_print_first.joblib ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0d23a98229dc9d5078b3f52972cc48f290537aa6c04eabeaa9716498da6a59c8
3
+ size 11326013
results/track_record/collator_print_first.json ADDED
@@ -0,0 +1,262 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "origin_text": "[CLS] ship. gunshot, gunfire. clapping. ratchet, pawl. yell. wind chime. whip. generic impact sounds. [SEP] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD]",
3
+ "new_text": "[CLS] ship. gunshot, gunfire. clapping. ratchet, pawl. yell. wind chime. whip. generic impact sounds. [SEP] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD]",
4
+ "origin_labels": [
5
+ [
6
+ 23,
7
+ [
8
+ 7.974500060081482,
9
+ 0.07350000087171793
10
+ ]
11
+ ],
12
+ [
13
+ 24,
14
+ [
15
+ 7.974500060081482,
16
+ 0.07350000087171793
17
+ ]
18
+ ],
19
+ [
20
+ 25,
21
+ [
22
+ 7.974500060081482,
23
+ 0.07350000087171793
24
+ ]
25
+ ],
26
+ [
27
+ 23,
28
+ [
29
+ 8.258999586105347,
30
+ 0.06499999668449163
31
+ ]
32
+ ],
33
+ [
34
+ 24,
35
+ [
36
+ 8.258999586105347,
37
+ 0.06499999668449163
38
+ ]
39
+ ],
40
+ [
41
+ 25,
42
+ [
43
+ 8.258999586105347,
44
+ 0.06499999668449163
45
+ ]
46
+ ],
47
+ [
48
+ 23,
49
+ [
50
+ 8.762499690055847,
51
+ 0.048500001430511475
52
+ ]
53
+ ],
54
+ [
55
+ 24,
56
+ [
57
+ 8.762499690055847,
58
+ 0.048500001430511475
59
+ ]
60
+ ],
61
+ [
62
+ 25,
63
+ [
64
+ 8.762499690055847,
65
+ 0.048500001430511475
66
+ ]
67
+ ],
68
+ [
69
+ 23,
70
+ [
71
+ 9.322999715805054,
72
+ 0.0570000009611249
73
+ ]
74
+ ],
75
+ [
76
+ 24,
77
+ [
78
+ 9.322999715805054,
79
+ 0.0570000009611249
80
+ ]
81
+ ],
82
+ [
83
+ 25,
84
+ [
85
+ 9.322999715805054,
86
+ 0.0570000009611249
87
+ ]
88
+ ],
89
+ [
90
+ 23,
91
+ [
92
+ 9.635499715805054,
93
+ 0.04449999891221523
94
+ ]
95
+ ],
96
+ [
97
+ 24,
98
+ [
99
+ 9.635499715805054,
100
+ 0.04449999891221523
101
+ ]
102
+ ],
103
+ [
104
+ 25,
105
+ [
106
+ 9.635499715805054,
107
+ 0.04449999891221523
108
+ ]
109
+ ],
110
+ [
111
+ 23,
112
+ [
113
+ 9.921500086784363,
114
+ 0.07850000634789467
115
+ ]
116
+ ],
117
+ [
118
+ 24,
119
+ [
120
+ 9.921500086784363,
121
+ 0.07850000634789467
122
+ ]
123
+ ],
124
+ [
125
+ 25,
126
+ [
127
+ 9.921500086784363,
128
+ 0.07850000634789467
129
+ ]
130
+ ]
131
+ ],
132
+ "new_labels": [
133
+ [
134
+ 23,
135
+ [
136
+ 7.9745001792907715,
137
+ 0.07349999994039536
138
+ ]
139
+ ],
140
+ [
141
+ 24,
142
+ [
143
+ 7.9745001792907715,
144
+ 0.07349999994039536
145
+ ]
146
+ ],
147
+ [
148
+ 25,
149
+ [
150
+ 7.9745001792907715,
151
+ 0.07349999994039536
152
+ ]
153
+ ],
154
+ [
155
+ 23,
156
+ [
157
+ 8.258999824523926,
158
+ 0.06499999761581421
159
+ ]
160
+ ],
161
+ [
162
+ 24,
163
+ [
164
+ 8.258999824523926,
165
+ 0.06499999761581421
166
+ ]
167
+ ],
168
+ [
169
+ 25,
170
+ [
171
+ 8.258999824523926,
172
+ 0.06499999761581421
173
+ ]
174
+ ],
175
+ [
176
+ 23,
177
+ [
178
+ 8.762499809265137,
179
+ 0.048500001430511475
180
+ ]
181
+ ],
182
+ [
183
+ 24,
184
+ [
185
+ 8.762499809265137,
186
+ 0.048500001430511475
187
+ ]
188
+ ],
189
+ [
190
+ 25,
191
+ [
192
+ 8.762499809265137,
193
+ 0.048500001430511475
194
+ ]
195
+ ],
196
+ [
197
+ 23,
198
+ [
199
+ 9.322999954223633,
200
+ 0.05700000002980232
201
+ ]
202
+ ],
203
+ [
204
+ 24,
205
+ [
206
+ 9.322999954223633,
207
+ 0.05700000002980232
208
+ ]
209
+ ],
210
+ [
211
+ 25,
212
+ [
213
+ 9.322999954223633,
214
+ 0.05700000002980232
215
+ ]
216
+ ],
217
+ [
218
+ 23,
219
+ [
220
+ 9.635499954223633,
221
+ 0.04450000077486038
222
+ ]
223
+ ],
224
+ [
225
+ 24,
226
+ [
227
+ 9.635499954223633,
228
+ 0.04450000077486038
229
+ ]
230
+ ],
231
+ [
232
+ 25,
233
+ [
234
+ 9.635499954223633,
235
+ 0.04450000077486038
236
+ ]
237
+ ],
238
+ [
239
+ 23,
240
+ [
241
+ 9.921500205993652,
242
+ 0.07850000262260437
243
+ ]
244
+ ],
245
+ [
246
+ 24,
247
+ [
248
+ 9.921500205993652,
249
+ 0.07850000262260437
250
+ ]
251
+ ],
252
+ [
253
+ 25,
254
+ [
255
+ 9.921500205993652,
256
+ 0.07850000262260437
257
+ ]
258
+ ]
259
+ ],
260
+ "origin_labels2text": "generic impact sounds generic impact sounds generic impact sounds generic impact sounds generic impact sounds generic impact sounds",
261
+ "new_labels2text": "generic impact sounds generic impact sounds generic impact sounds generic impact sounds generic impact sounds generic impact sounds"
262
+ }
results/track_record/enc_text_hidden_state.joblib ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a29a9873dff53ab3f5047466d33a23ac100ab0ef5e1eaf4f516feaf17163cd8e
3
+ size 946594
results/track_record/giou_cost.joblib ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e9e81757b0d9d839642470e84e01c409bf62843a6b9243a551f56f9ef705ba77
3
+ size 343612
results/track_record/grad_count.json ADDED
@@ -0,0 +1 @@
 
 
1
+ "\"{'model.query_position_embeddings.weight': 0.26171875, 'bbox_embed.0.layers.1.weight': 0.2314453125, 'bbox_embed.0.layers.0.weight': 0.2216796875, 'model.text_projection.weight': 0.20703125, 'bbox_embed.0.layers.2.weight': 0.1767578125, 'model.decoder.layers.5.fc1.weight': 0.15625, 'model.decoder.layers.5.fc2.weight': 0.15234375, 'model.encoder.layers.5.text_enhancer_layer.fc1.weight': 0.1123046875, 'model.encoder.layers.5.text_enhancer_layer.fc2.weight': 0.10400390625, 'model.decoder.layers.4.fc1.weight': 0.099609375, 'model.decoder.layers.4.fc2.weight': 0.095703125, 'model.encoder.layers.4.text_enhancer_layer.fc2.weight': 0.078125, 'model.decoder.layers.3.fc1.weight': 0.076171875, 'model.decoder.layers.3.fc2.weight': 0.07470703125, 'model.encoder.layers.4.text_enhancer_layer.fc1.weight': 0.07275390625, 'model.encoder.layers.3.text_enhancer_layer.fc1.weight': 0.06396484375, 'model.decoder.layers.5.encoder_attn.out_proj.weight': 0.06396484375, 'model.decoder.layers.2.fc1.weight': 0.0634765625, 'model.decoder.layers.2.fc2.weight': 0.06298828125, 'model.decoder.layers.5.encoder_attn_text.out_proj.weight': 0.06201171875, 'model.encoder.layers.3.text_enhancer_layer.fc2.weight': 0.061767578125, 'model.decoder.layers.1.fc2.weight': 0.05810546875, 'model.decoder.layers.5.encoder_attn.value.weight': 0.0576171875, 'model.decoder.layers.0.fc2.weight': 0.056884765625, 'model.decoder.layers.1.fc1.weight': 0.056884765625, 'model.encoder.layers.5.text_enhancer_layer.self_attn.out_proj.weight': 0.05517578125, 'model.decoder.layers.0.fc1.weight': 0.054443359375, 'model.decoder.layers.5.encoder_attn_text.value.weight': 0.05419921875, 'model.encoder.layers.5.text_enhancer_layer.self_attn.value.weight': 0.053466796875, 'model.decoder.layers.5.self_attn.out_proj.weight': 0.05224609375, 'model.encoder.layers.2.text_enhancer_layer.fc1.weight': 0.0517578125, 'model.decoder.layers.5.self_attn.value.weight': 0.0498046875, 'model.encoder.layers.2.text_enhancer_layer.fc2.weight': 0.048828125, 'model.encoder.layers.4.text_enhancer_layer.self_attn.out_proj.weight': 0.04541015625, 'model.encoder.layers.4.text_enhancer_layer.self_attn.value.weight': 0.044189453125, 'model.decoder.layers.4.encoder_attn.value.weight': 0.0439453125, 'model.encoder.layers.1.text_enhancer_layer.fc1.weight': 0.043701171875, 'model.encoder.layers.1.text_enhancer_layer.fc2.weight': 0.041748046875, 'model.decoder.layers.4.encoder_attn.out_proj.weight': 0.04052734375, 'model.encoder.layers.0.text_enhancer_layer.fc2.weight': 0.03857421875, 'model.decoder.layers.4.encoder_attn_text.value.weight': 0.03857421875, 'model.encoder.layers.3.text_enhancer_layer.self_attn.out_proj.weight': 0.03759765625, 'model.decoder.layers.3.encoder_attn.out_proj.weight': 0.036865234375, 'model.encoder.layers.0.text_enhancer_layer.fc1.weight': 0.03662109375, 'model.encoder.layers.3.text_enhancer_layer.self_attn.value.weight': 0.035888671875, 'model.decoder.layers.4.encoder_attn_text.out_proj.weight': 0.035400390625, 'model.decoder.layers.3.encoder_attn.value.weight': 0.033935546875, 'model.decoder.layers.3.encoder_attn_text.value.weight': 0.033447265625, 'model.decoder.layers.3.encoder_attn_text.out_proj.weight': 0.0322265625, 'model.decoder.layers.2.encoder_attn.out_proj.weight': 0.03125, 'model.encoder.layers.2.text_enhancer_layer.self_attn.value.weight': 0.031005859375, 'model.decoder.layers.4.self_attn.value.weight': 0.03076171875, 'model.input_proj_vision.0.0.weight': 0.0303955078125, 'model.encoder.layers.2.text_enhancer_layer.self_attn.out_proj.weight': 0.02978515625, 'model.decoder.layers.2.encoder_attn.value.weight': 0.0296630859375, 'model.decoder.layers.4.self_attn.out_proj.weight': 0.02783203125, 'model.encoder.layers.1.text_enhancer_layer.self_attn.out_proj.weight': 0.027587890625, 'model.decoder.layers.0.encoder_attn.out_proj.weight': 0.02734375, 'model.decoder.layers.0.encoder_attn_text.value.weight': 0.027099609375, 'model.decoder.layers.1.encoder_attn.value.weight': 0.026611328125, 'model.decoder.layers.0.encoder_attn.value.weight': 0.0264892578125, 'model.decoder.layers.2.encoder_attn_text.value.weight': 0.0263671875, 'model.decoder.layers.1.encoder_attn.out_proj.weight': 0.0262451171875, 'model.decoder.layers.2.encoder_attn_text.out_proj.weight': 0.0255126953125, 'model.decoder.layers.1.encoder_attn_text.value.weight': 0.025390625, 'model.decoder.layers.0.encoder_attn_text.out_proj.weight': 0.025146484375, 'model.decoder.layers.1.encoder_attn_text.out_proj.weight': 0.0250244140625, 'model.encoder.layers.1.text_enhancer_layer.self_attn.value.weight': 0.02392578125, 'model.encoder.layers.0.text_enhancer_layer.self_attn.value.weight': 0.02294921875, 'model.backbone.scale_feature.0.scale_conv.weight': 0.021728515625, 'model.encoder.layers.0.text_enhancer_layer.self_attn.out_proj.weight': 0.0216064453125, 'model.decoder.layers.3.self_attn.value.weight': 0.021484375, 'model.decoder.layers.3.self_attn.out_proj.weight': 0.0196533203125, 'model.decoder.layers.5.final_layer_norm.bias': 0.017578125, 'model.decoder.layer_norm.bias': 0.0174560546875, 'model.decoder.layer_norm.weight': 0.017333984375, 'model.encoder.layers.5.text_enhancer_layer.layer_norm_after.bias': 0.014892578125, 'model.encoder.layers.5.deformable_layer.fc2.weight': 0.01458740234375, 'model.decoder.layers.2.self_attn.out_proj.weight': 0.01458740234375, 'model.decoder.layers.5.final_layer_norm.weight': 0.01416015625, 'model.encoder.layers.5.deformable_layer.fc1.weight': 0.0137939453125, 'model.decoder.layers.2.self_attn.value.weight': 0.01373291015625, 'model.decoder.layers.5.encoder_attn_layer_norm.bias': 0.01275634765625, 'model.decoder.layers.5.encoder_attn_text_layer_norm.bias': 0.01190185546875, 'model.encoder.layers.4.deformable_layer.fc2.weight': 0.0118408203125, 'model.encoder.layers.5.text_enhancer_layer.layer_norm_after.weight': 0.01165771484375, 'model.encoder.layers.5.text_enhancer_layer.layer_norm_before.bias': 0.0115966796875, 'model.decoder.layers.5.encoder_attn_layer_norm.weight': 0.01123046875, 'model.decoder.layers.5.self_attn_layer_norm.bias': 0.01104736328125, 'model.encoder.layers.4.text_enhancer_layer.layer_norm_after.bias': 0.01092529296875, 'model.encoder.layers.5.fusion_layer.layer_norm_text.bias': 0.01092529296875, 'model.encoder.layers.4.deformable_layer.fc1.weight': 0.01068115234375, 'model.decoder.layers.5.encoder_attn_text_layer_norm.weight': 0.01068115234375, 'model.encoder.layers.3.deformable_layer.fc2.weight': 0.010498046875, 'model.decoder.layers.4.final_layer_norm.bias': 0.01025390625, 'model.decoder.layers.5.self_attn_layer_norm.weight': 0.0098876953125, 'model.encoder.layers.5.deformable_layer.self_attn.out_proj.weight': 0.0096435546875, 'model.decoder.layers.4.final_layer_norm.weight': 0.0093994140625, 'model.encoder.layers.4.text_enhancer_layer.layer_norm_before.bias': 0.00933837890625, 'model.encoder.layers.5.text_enhancer_layer.layer_norm_before.weight': 0.00927734375, 'model.encoder.layers.3.deformable_layer.fc1.weight': 0.00909423828125, 'model.encoder.layers.2.deformable_layer.fc2.weight': 0.009033203125, 'model.decoder.layers.1.self_attn.value.weight': 0.00897216796875, 'model.encoder.layers.4.text_enhancer_layer.layer_norm_after.weight': 0.00872802734375, 'model.encoder.layers.5.fusion_layer.layer_norm_text.weight': 0.00872802734375, 'model.encoder.layers.5.deformable_layer.self_attn.value.weight': 0.00872802734375, 'model.decoder.layers.1.self_attn.out_proj.weight': 0.00872802734375, 'model.encoder.layers.3.text_enhancer_layer.layer_norm_after.bias': 0.00860595703125, 'model.encoder.layers.4.fusion_layer.layer_norm_text.bias': 0.00860595703125, 'model.encoder.layers.4.deformable_layer.self_attn.out_proj.weight': 0.00836181640625, 'model.decoder.layers.4.encoder_attn_layer_norm.bias': 0.0081787109375, 'model.encoder.layers.5.fusion_layer.text_param': 0.00811767578125, 'model.decoder.layers.4.encoder_attn_text_layer_norm.bias': 0.00787353515625, 'model.encoder.layers.2.deformable_layer.fc1.weight': 0.0078125, 'model.encoder.layers.4.text_enhancer_layer.layer_norm_before.weight': 0.0078125, 'model.decoder.layers.4.encoder_attn_layer_norm.weight': 0.0078125, 'model.decoder.layers.4.self_attn_layer_norm.bias': 0.007537841796875, 'model.decoder.layers.4.encoder_attn_text_layer_norm.weight': 0.007537841796875, 'model.encoder.layers.4.deformable_layer.self_attn.value.weight': 0.00750732421875, 'model.encoder.layers.3.text_enhancer_layer.layer_norm_before.bias': 0.007476806640625, 'model.decoder.layers.3.final_layer_norm.bias': 0.00738525390625, 'model.decoder.layers.4.self_attn_layer_norm.weight': 0.007293701171875, 'model.encoder.layers.3.text_enhancer_layer.layer_norm_after.weight': 0.00726318359375, 'model.encoder.layers.4.fusion_layer.layer_norm_text.weight': 0.00726318359375, 'model.encoder.layers.1.deformable_layer.fc2.weight': 0.00714111328125, 'model.decoder.layers.3.final_layer_norm.weight': 0.00714111328125, 'model.encoder.layers.2.text_enhancer_layer.layer_norm_after.bias': 0.00701904296875, 'model.encoder.layers.3.fusion_layer.layer_norm_text.bias': 0.00701904296875, 'model.encoder.layers.3.fusion_layer.text_param': 0.00677490234375, 'model.decoder.layers.0.self_attn.value.weight': 0.006683349609375, 'model.decoder.layers.3.encoder_attn_layer_norm.bias': 0.006683349609375, 'model.decoder.layers.3.encoder_attn_layer_norm.weight': 0.006622314453125, 'model.decoder.layers.3.encoder_attn_text_layer_norm.weight': 0.006500244140625, 'model.encoder.layers.1.deformable_layer.fc1.weight': 0.0064697265625, 'model.decoder.layers.3.encoder_attn_text_layer_norm.bias': 0.0064697265625, 'model.decoder.layers.3.self_attn_layer_norm.weight': 0.006439208984375, 'model.encoder.layers.3.deformable_layer.self_attn.out_proj.weight': 0.00640869140625, 'model.decoder.layers.0.self_attn.out_proj.weight': 0.00640869140625, 'model.decoder.layers.2.final_layer_norm.weight': 0.00640869140625, 'model.encoder.layers.3.deformable_layer.self_attn.value.weight': 0.006378173828125, 'model.encoder.layers.3.text_enhancer_layer.layer_norm_before.weight': 0.006256103515625, 'model.encoder.layers.2.text_enhancer_layer.layer_norm_before.bias': 0.0062255859375, 'model.decoder.layers.3.self_attn_layer_norm.bias': 0.0062255859375, 'model.decoder.layers.2.final_layer_norm.bias': 0.006103515625, 'model.encoder.layers.2.text_enhancer_layer.layer_norm_after.weight': 0.006011962890625, 'model.encoder.layers.1.text_enhancer_layer.layer_norm_after.bias': 0.0059814453125, 'model.encoder.layers.2.fusion_layer.layer_norm_text.bias': 0.0059814453125, 'model.encoder.layers.3.fusion_layer.layer_norm_text.weight': 0.0059814453125, 'model.encoder.layers.0.deformable_layer.fc2.weight': 0.005950927734375, 'model.decoder.layers.2.encoder_attn_layer_norm.weight': 0.005828857421875, 'model.decoder.layers.2.encoder_attn_text_layer_norm.weight': 0.00579833984375, 'model.decoder.layers.1.final_layer_norm.weight': 0.00567626953125, 'model.decoder.layers.2.self_attn_layer_norm.weight': 0.00567626953125, 'model.decoder.layers.2.encoder_attn_layer_norm.bias': 0.00567626953125, 'model.encoder.layers.2.fusion_layer.text_param': 0.005584716796875, 'model.decoder.layers.2.encoder_attn_text_layer_norm.bias': 0.005584716796875, 'model.encoder.layers.2.deformable_layer.self_attn.out_proj.weight': 0.00555419921875, 'model.decoder.layers.1.encoder_attn_text_layer_norm.weight': 0.005523681640625, 'model.decoder.layers.1.encoder_attn_layer_norm.weight': 0.005523681640625, 'model.encoder.layers.2.text_enhancer_layer.layer_norm_before.weight': 0.0054931640625, 'model.decoder.layers.0.final_layer_norm.weight': 0.005462646484375, 'model.decoder.layers.1.self_attn_layer_norm.weight': 0.005462646484375, 'model.decoder.layers.1.final_layer_norm.bias': 0.005462646484375, 'model.decoder.layers.2.self_attn_layer_norm.bias': 0.00543212890625, 'model.encoder.layers.1.text_enhancer_layer.layer_norm_before.bias': 0.005340576171875, 'model.decoder.layers.0.self_attn_layer_norm.weight': 0.005340576171875, 'model.decoder.layers.0.encoder_attn_text_layer_norm.weight': 0.00531005859375, 'model.decoder.layers.0.encoder_attn_layer_norm.weight': 0.00531005859375, 'model.decoder.layers.1.encoder_attn_layer_norm.bias': 0.005279541015625, 'model.encoder.layers.1.text_enhancer_layer.layer_norm_after.weight': 0.0052490234375, 'model.encoder.layers.2.fusion_layer.layer_norm_text.weight': 0.0052490234375, 'model.decoder.layers.1.encoder_attn_text_layer_norm.bias': 0.005218505859375, 'model.encoder.layers.0.deformable_layer.fc1.weight': 0.005157470703125, 'model.encoder.layers.2.deformable_layer.self_attn.value.weight': 0.005157470703125, 'model.decoder.layers.0.self_attn_layer_norm.bias': 0.005157470703125, 'model.decoder.layers.0.final_layer_norm.bias': 0.005157470703125, 'model.decoder.layers.1.self_attn_layer_norm.bias': 0.005157470703125, 'model.encoder.layers.0.text_enhancer_layer.layer_norm_after.bias': 0.005126953125, 'model.decoder.layers.0.encoder_attn_layer_norm.bias': 0.005126953125, 'model.encoder.layers.1.fusion_layer.layer_norm_text.bias': 0.005096435546875, 'model.decoder.layers.0.encoder_attn_text_layer_norm.bias': 0.005096435546875, 'model.encoder.layers.1.deformable_layer.self_attn.out_proj.weight': 0.004913330078125, 'model.encoder.layers.1.text_enhancer_layer.layer_norm_before.weight': 0.004791259765625, 'model.encoder.layers.0.text_enhancer_layer.layer_norm_after.weight': 0.004638671875, 'model.encoder.layers.1.deformable_layer.self_attn.value.weight': 0.004638671875, 'model.encoder.layers.0.text_enhancer_layer.layer_norm_before.bias': 0.004608154296875, 'model.encoder.layers.1.fusion_layer.layer_norm_text.weight': 0.004608154296875, 'model.encoder.layers.0.fusion_layer.layer_norm_text.bias': 0.004486083984375, 'model.encoder.layers.4.fusion_layer.text_param': 0.004241943359375, 'model.encoder.layers.0.text_enhancer_layer.layer_norm_before.weight': 0.00421142578125, 'model.encoder.layers.0.fusion_layer.layer_norm_text.weight': 0.004150390625, 'model.encoder.layers.0.deformable_layer.self_attn.out_proj.weight': 0.003936767578125, 'model.encoder.layers.0.deformable_layer.self_attn.value.weight': 0.0038909912109375, 'model.encoder.layers.0.fusion_layer.text_param': 0.0034942626953125, 'model.encoder.layers.1.fusion_layer.text_param': 0.0030059814453125, 'model.encoder.layers.5.fusion_layer.vision_param': 0.00274658203125, 'model.encoder.layers.5.deformable_layer.layer_norm_after.weight': 0.0025634765625, 'model.encoder.layers.5.deformable_layer.layer_norm_after.bias': 0.002349853515625, 'model.encoder.layers.5.deformable_layer.layer_norm_before.weight': 0.001922607421875, 'model.encoder.layers.5.deformable_layer.layer_norm_before.bias': 0.001922607421875, 'model.encoder.layers.4.fusion_layer.vision_param': 0.0018463134765625, 'model.encoder.layers.4.deformable_layer.layer_norm_after.weight': 0.0018157958984375, 'model.encoder.layers.5.fusion_layer.layer_norm_vision.weight': 0.0018157958984375, 'model.encoder.layers.4.deformable_layer.layer_norm_after.bias': 0.00179290771484375, 'model.encoder.layers.5.fusion_layer.layer_norm_vision.bias': 0.00179290771484375, 'model.encoder.layers.4.deformable_layer.layer_norm_before.weight': 0.00164794921875, 'model.encoder.layers.4.deformable_layer.layer_norm_before.bias': 0.001617431640625, 'model.encoder.layers.3.deformable_layer.layer_norm_after.weight': 0.0015411376953125, 'model.encoder.layers.4.fusion_layer.layer_norm_vision.weight': 0.0015411376953125, 'model.encoder.layers.3.deformable_layer.layer_norm_after.bias': 0.00151824951171875, 'model.encoder.layers.4.fusion_layer.layer_norm_vision.bias': 0.00151824951171875, 'model.encoder.layers.3.fusion_layer.vision_param': 0.00148773193359375, 'model.encoder.layers.3.deformable_layer.layer_norm_before.weight': 0.0013580322265625, 'model.encoder.layers.3.deformable_layer.layer_norm_before.bias': 0.0013275146484375, 'model.encoder.layers.2.deformable_layer.layer_norm_after.weight': 0.00128936767578125, 'model.encoder.layers.3.fusion_layer.layer_norm_vision.weight': 0.00128936767578125, 'model.encoder.layers.2.deformable_layer.layer_norm_after.bias': 0.00125885009765625, 'model.encoder.layers.3.fusion_layer.layer_norm_vision.bias': 0.00125885009765625, 'model.encoder.layers.2.fusion_layer.vision_param': 0.00118255615234375, 'model.encoder.layers.2.deformable_layer.layer_norm_before.weight': 0.001129150390625, 'model.encoder.layers.2.deformable_layer.layer_norm_before.bias': 0.0011138916015625, 'model.encoder.layers.2.fusion_layer.layer_norm_vision.weight': 0.001068115234375, 'model.encoder.layers.1.deformable_layer.layer_norm_after.weight': 0.00106048583984375, 'model.encoder.layers.1.deformable_layer.layer_norm_after.bias': 0.00103759765625, 'model.encoder.layers.2.fusion_layer.layer_norm_vision.bias': 0.00103759765625, 'model.decoder.layers.1.self_attn.query.weight': 0.00099945068359375, 'model.encoder.layers.1.deformable_layer.layer_norm_before.weight': 0.0009918212890625, 'model.encoder.layers.0.fusion_layer.vision_param': 0.000957489013671875, 'model.encoder.layers.0.deformable_layer.layer_norm_after.weight': 0.00095367431640625, 'model.encoder.layers.1.fusion_layer.layer_norm_vision.weight': 0.00095367431640625, 'model.decoder.layers.5.encoder_attn_text.key.weight': 0.000946044921875, 'model.encoder.layers.1.deformable_layer.layer_norm_before.bias': 0.00093841552734375, 'model.decoder.layers.1.self_attn.key.weight': 0.0009307861328125, 'model.decoder.layers.5.encoder_attn_text.query.weight': 0.00091552734375, 'model.decoder.layers.2.self_attn.query.weight': 0.000885009765625, 'model.encoder.layers.0.deformable_layer.layer_norm_after.bias': 0.000881195068359375, 'model.encoder.layers.1.fusion_layer.layer_norm_vision.bias': 0.000881195068359375, 'model.encoder.layers.0.deformable_layer.layer_norm_before.weight': 0.000873565673828125, 'model.encoder.layers.1.fusion_layer.vision_param': 0.000873565673828125, 'model.decoder.layers.2.self_attn.key.weight': 0.000865936279296875, 'model.encoder.layers.0.fusion_layer.layer_norm_vision.weight': 0.000823974609375, 'model.encoder.layers.0.deformable_layer.layer_norm_before.bias': 0.00079345703125, 'model.decoder.layers.4.encoder_attn_text.key.weight': 0.000759124755859375, 'model.encoder.layers.0.fusion_layer.layer_norm_vision.bias': 0.00075531005859375, 'model.decoder.layers.3.self_attn.query.weight': 0.0007476806640625, 'model.decoder.layers.3.self_attn.key.weight': 0.000743865966796875, 'model.decoder.layers.4.encoder_attn_text.query.weight': 0.000705718994140625, 'model.backbone.scale_feature.0.scale_norm.weight': 0.000701904296875, 'model.decoder.layers.4.self_attn.query.weight': 0.000698089599609375, 'model.decoder.layers.5.self_attn.key.weight': 0.000690460205078125, 'model.decoder.layers.4.self_attn.key.weight': 0.000682830810546875, 'model.decoder.layers.5.self_attn.query.weight': 0.00067901611328125, 'model.decoder.layers.3.encoder_attn_text.key.weight': 0.0006561279296875, 'model.decoder.layers.2.encoder_attn_text.key.weight': 0.00060272216796875, 'model.backbone.scale_feature.0.scale_norm.bias': 0.000591278076171875, 'model.decoder.layers.3.encoder_attn_text.query.weight': 0.000591278076171875, 'model.decoder.layers.2.encoder_attn_text.query.weight': 0.00055694580078125, 'model.decoder.layers.0.encoder_attn_text.query.weight': 0.000553131103515625, 'model.decoder.layers.0.encoder_attn_text.key.weight': 0.00054931640625, 'model.decoder.layers.1.encoder_attn_text.query.weight': 0.000522613525390625, 'model.decoder.layers.1.encoder_attn_text.key.weight': 0.000514984130859375, 'model.encoder.layers.5.text_enhancer_layer.self_attn.query.weight': 0.0005035400390625, 'model.encoder.layers.5.text_enhancer_layer.self_attn.key.weight': 0.00048065185546875, 'model.encoder.layers.0.text_enhancer_layer.self_attn.key.weight': 0.0004100799560546875, 'model.encoder.layers.0.text_enhancer_layer.self_attn.query.weight': 0.000400543212890625, 'model.encoder.layers.3.text_enhancer_layer.self_attn.key.weight': 0.00038909912109375, 'model.encoder.layers.4.text_enhancer_layer.self_attn.key.weight': 0.0003871917724609375, 'model.encoder.layers.2.text_enhancer_layer.self_attn.query.weight': 0.000385284423828125, 'model.encoder.layers.2.text_enhancer_layer.self_attn.key.weight': 0.0003814697265625, 'model.encoder.layers.4.text_enhancer_layer.self_attn.query.weight': 0.0003643035888671875, 'model.encoder.layers.3.text_enhancer_layer.self_attn.query.weight': 0.0003528594970703125, 'model.encoder.layers.1.text_enhancer_layer.self_attn.key.weight': 0.0003376007080078125, 'model.encoder.layers.1.text_enhancer_layer.self_attn.query.weight': 0.00031280517578125, 'model.decoder.layers.0.self_attn.query.weight': 0.0002193450927734375, 'model.decoder.layers.0.self_attn.key.weight': 0.00021266937255859375, 'model.backbone.scale_feature.0.scale_conv.bias': 0.00016498565673828125, 'model.encoder.layers.5.fusion_layer.attn.out_text_proj.weight': 7.62939453125e-05, 'model.encoder.layers.3.fusion_layer.attn.out_text_proj.weight': 7.581710815429688e-05, 'model.encoder.layers.5.fusion_layer.attn.values_vision_proj.weight': 7.581710815429688e-05, 'model.encoder.layers.3.fusion_layer.attn.values_vision_proj.weight': 7.343292236328125e-05, 'model.encoder.layers.2.fusion_layer.attn.out_text_proj.weight': 6.103515625e-05, 'model.encoder.layers.2.fusion_layer.attn.values_vision_proj.weight': 5.9604644775390625e-05, 'model.encoder.layers.3.fusion_layer.attn.text_proj.weight': 5.3882598876953125e-05, 'model.encoder.layers.3.fusion_layer.attn.vision_proj.weight': 5.221366882324219e-05, 'model.encoder.layers.5.fusion_layer.attn.out_vision_proj.weight': 4.76837158203125e-05, 'model.encoder.layers.5.fusion_layer.attn.values_text_proj.weight': 4.696846008300781e-05, 'model.encoder.layers.4.fusion_layer.attn.out_text_proj.weight': 4.506111145019531e-05, 'model.encoder.layers.4.fusion_layer.attn.values_vision_proj.weight': 4.315376281738281e-05, 'model.encoder.layers.0.fusion_layer.attn.values_vision_proj.weight': 3.3855438232421875e-05, 'model.encoder.layers.0.fusion_layer.attn.out_text_proj.weight': 3.361701965332031e-05, 'model.encoder.layers.1.fusion_layer.attn.values_vision_proj.weight': 3.2901763916015625e-05, 'model.encoder.layers.1.fusion_layer.attn.out_text_proj.weight': 3.1948089599609375e-05, 'model.encoder.layers.4.fusion_layer.attn.out_vision_proj.weight': 2.956390380859375e-05, 'model.encoder.layers.4.fusion_layer.attn.values_text_proj.weight': 2.9325485229492188e-05, 'model.encoder.layers.2.fusion_layer.attn.text_proj.weight': 2.5510787963867188e-05, 'model.encoder.layers.0.fusion_layer.attn.text_proj.weight': 2.5391578674316406e-05, 'model.encoder.layers.2.fusion_layer.attn.vision_proj.weight': 2.491474151611328e-05, 'model.encoder.layers.3.fusion_layer.attn.values_text_proj.weight': 2.491474151611328e-05, 'model.encoder.layers.0.fusion_layer.attn.vision_proj.weight': 2.4199485778808594e-05, 'model.encoder.layers.3.fusion_layer.attn.out_vision_proj.weight': 2.4199485778808594e-05, 'model.encoder.layers.1.fusion_layer.attn.vision_proj.weight': 2.3603439331054688e-05, 'model.encoder.layers.5.fusion_layer.attn.vision_proj.weight': 2.300739288330078e-05, 'model.encoder.layers.1.fusion_layer.attn.text_proj.weight': 2.276897430419922e-05, 'model.encoder.layers.4.fusion_layer.attn.text_proj.weight': 2.110004425048828e-05, 'model.encoder.layers.5.fusion_layer.attn.text_proj.weight': 2.110004425048828e-05, 'model.encoder.layers.4.fusion_layer.attn.vision_proj.weight': 2.09808349609375e-05, 'model.encoder.layers.2.fusion_layer.attn.values_text_proj.weight': 2.002716064453125e-05, 'model.encoder.layers.2.fusion_layer.attn.out_vision_proj.weight': 1.990795135498047e-05, 'model.encoder.layers.0.fusion_layer.attn.values_text_proj.weight': 1.5974044799804688e-05, 'model.encoder.layers.0.fusion_layer.attn.out_vision_proj.weight': 1.5735626220703125e-05, 'model.encoder.layers.1.fusion_layer.attn.out_vision_proj.weight': 1.5616416931152344e-05, 'model.encoder.layers.1.fusion_layer.attn.values_text_proj.weight': 1.5497207641601562e-05, 'model.encoder.layers.1.deformable_layer.self_attn.query.weight': 7.987022399902344e-06, 'model.encoder.layers.1.deformable_layer.self_attn.key.weight': 7.927417755126953e-06, 'model.encoder.layers.0.deformable_layer.self_attn.key.weight': 6.16908073425293e-06, 'model.encoder.layers.4.deformable_layer.self_attn.key.weight': 6.16908073425293e-06, 'model.encoder.layers.4.deformable_layer.self_attn.query.weight': 5.811452865600586e-06, 'model.encoder.layers.0.deformable_layer.self_attn.query.weight': 5.453824996948242e-06, 'model.encoder.layers.5.deformable_layer.self_attn.key.weight': 4.202127456665039e-06, 'model.encoder.layers.2.deformable_layer.self_attn.key.weight': 4.112720489501953e-06, 'model.encoder.layers.2.deformable_layer.self_attn.query.weight': 4.0531158447265625e-06, 'model.decoder.layers.5.encoder_attn.key.weight': 4.023313522338867e-06, 'model.encoder.layers.5.deformable_layer.self_attn.query.weight': 3.9637088775634766e-06, 'model.decoder.layers.5.encoder_attn.query.weight': 3.7997961044311523e-06, 'model.encoder.layers.3.deformable_layer.self_attn.key.weight': 3.7550926208496094e-06, 'model.encoder.layers.3.deformable_layer.self_attn.query.weight': 3.5017728805541992e-06, 'model.decoder.layers.4.encoder_attn.query.weight': 3.427267074584961e-06, 'model.decoder.layers.4.encoder_attn.key.weight': 3.293156623840332e-06, 'model.decoder.layers.2.encoder_attn.key.weight': 2.8908252716064453e-06, 'model.decoder.layers.3.encoder_attn.query.weight': 2.8014183044433594e-06, 'model.decoder.layers.3.encoder_attn.key.weight': 2.7567148208618164e-06, 'model.decoder.layers.2.encoder_attn.query.weight': 2.682209014892578e-06, 'model.decoder.layers.0.encoder_attn.key.weight': 2.250075340270996e-06, 'model.decoder.layers.0.encoder_attn.query.weight': 2.1457672119140625e-06, 'model.decoder.layers.1.encoder_attn.key.weight': 2.130866050720215e-06, 'model.decoder.layers.1.encoder_attn.query.weight': 2.041459083557129e-06, 'model.level_embed': 1.601874828338623e-07, 'model.enc_output.weight': 0.0, 'model.enc_output_norm.weight': 0.0, 'model.enc_output_norm.bias': 0.0, 'model.encoder_output_bbox_embed.layers.0.weight': 0.0, 'model.encoder_output_bbox_embed.layers.1.weight': 0.0, 'model.encoder_output_bbox_embed.layers.2.weight': 0.0}\""
results/track_record/grad_count_score.json ADDED
@@ -0,0 +1,555 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "model.decoder.layers.0.self_attn.out_proj.bias": 3184.0,
3
+ "model.decoder.layers.0.self_attn.value.bias": 900.0,
4
+ "model.encoder.layers.5.text_enhancer_layer.fc2.weight": 592.0,
5
+ "model.decoder.layers.1.fc2.weight": 528.0,
6
+ "model.encoder.layers.5.text_enhancer_layer.fc1.weight": 480.0,
7
+ "model.decoder.layers.2.fc2.weight": 466.0,
8
+ "model.query_position_embeddings.weight": 414.0,
9
+ "model.decoder.layers.0.fc2.weight": 412.0,
10
+ "model.decoder.layers.3.fc2.weight": 394.0,
11
+ "model.encoder.layers.4.text_enhancer_layer.fc2.weight": 390.0,
12
+ "model.decoder.layers.1.encoder_attn_text.value.weight": 368.0,
13
+ "model.encoder.layers.5.text_enhancer_layer.self_attn.out_proj.weight": 356.0,
14
+ "model.decoder.layers.1.encoder_attn_text.out_proj.weight": 346.0,
15
+ "model.decoder.layers.0.encoder_attn_text.value.weight": 314.0,
16
+ "model.decoder.layers.0.encoder_attn_text.out_proj.weight": 310.0,
17
+ "model.encoder.layers.4.text_enhancer_layer.fc1.weight": 298.0,
18
+ "model.decoder.layers.4.fc2.weight": 288.0,
19
+ "model.encoder.layers.5.text_enhancer_layer.self_attn.value.weight": 278.0,
20
+ "model.decoder.layers.0.encoder_attn.output_proj.weight": 276.0,
21
+ "model.decoder.layers.2.encoder_attn_text.out_proj.weight": 274.0,
22
+ "model.decoder.layers.2.encoder_attn_text.value.weight": 272.0,
23
+ "model.decoder.layers.3.fc1.weight": 268.0,
24
+ "model.decoder.layers.2.fc1.weight": 264.0,
25
+ "model.encoder.layers.3.text_enhancer_layer.fc2.weight": 262.0,
26
+ "model.encoder.layers.4.text_enhancer_layer.self_attn.out_proj.weight": 247.0,
27
+ "model.decoder.layers.1.fc1.weight": 240.0,
28
+ "model.decoder.layers.4.fc1.weight": 212.0,
29
+ "model.text_projection.weight": 208.0,
30
+ "model.decoder.layers.3.encoder_attn_text.value.weight": 205.0,
31
+ "model.decoder.layers.3.encoder_attn_text.out_proj.weight": 205.0,
32
+ "model.decoder.layers.0.encoder_attn.value_proj.weight": 204.0,
33
+ "model.decoder.layers.1.encoder_attn.output_proj.weight": 203.0,
34
+ "model.decoder.layers.1.encoder_attn.value_proj.weight": 200.0,
35
+ "model.decoder.layers.2.encoder_attn.output_proj.weight": 198.0,
36
+ "model.encoder.layers.3.text_enhancer_layer.fc1.weight": 192.0,
37
+ "model.encoder.layers.4.text_enhancer_layer.self_attn.value.weight": 188.0,
38
+ "model.decoder.layers.2.encoder_attn.value_proj.weight": 183.0,
39
+ "model.decoder.layers.5.fc2.weight": 176.0,
40
+ "model.encoder.layers.2.text_enhancer_layer.fc2.weight": 160.0,
41
+ "model.decoder.layers.5.fc1.weight": 152.0,
42
+ "model.decoder.layer_norm.bias": 151.0,
43
+ "model.decoder.layers.4.encoder_attn_text.value.weight": 150.0,
44
+ "model.encoder.layers.3.text_enhancer_layer.self_attn.out_proj.weight": 149.0,
45
+ "model.decoder.layers.3.encoder_attn.value_proj.weight": 141.0,
46
+ "model.decoder.layers.3.encoder_attn.output_proj.weight": 137.0,
47
+ "model.decoder.layers.4.encoder_attn_text.out_proj.weight": 136.0,
48
+ "model.decoder.layers.0.fc1.weight": 132.0,
49
+ "model.encoder.layers.2.text_enhancer_layer.fc1.weight": 124.0,
50
+ "model.decoder.layers.2.self_attn.out_proj.weight": 122.5,
51
+ "model.encoder.layers.3.text_enhancer_layer.self_attn.value.weight": 119.0,
52
+ "model.decoder.layers.3.self_attn.value.weight": 115.5,
53
+ "model.decoder.layers.1.self_attn.value.weight": 115.0,
54
+ "model.decoder.layers.2.self_attn.value.weight": 114.5,
55
+ "model.decoder.layers.1.self_attn.out_proj.weight": 113.0,
56
+ "model.encoder.layers.1.text_enhancer_layer.fc2.weight": 112.0,
57
+ "model.decoder.layers.3.self_attn.out_proj.weight": 112.0,
58
+ "model.decoder.layers.4.self_attn.out_proj.weight": 104.0,
59
+ "model.encoder.layers.2.text_enhancer_layer.self_attn.out_proj.weight": 97.0,
60
+ "model.decoder.layers.4.encoder_attn.output_proj.weight": 97.0,
61
+ "model.decoder.layers.4.encoder_attn.value_proj.weight": 95.0,
62
+ "model.decoder.layers.4.self_attn.value.weight": 91.0,
63
+ "model.decoder.layers.1.final_layer_norm.bias": 90.5,
64
+ "model.encoder.layers.5.text_enhancer_layer.layer_norm_after.bias": 88.0,
65
+ "model.decoder.layers.1.fc2.bias": 85.5,
66
+ "model.decoder.layers.5.encoder_attn_text.value.weight": 83.5,
67
+ "model.encoder.layers.1.text_enhancer_layer.fc1.weight": 82.0,
68
+ "model.decoder.layers.5.encoder_attn_text.out_proj.weight": 80.0,
69
+ "model.decoder.layers.2.final_layer_norm.bias": 76.0,
70
+ "model.decoder.layers.1.encoder_attn_layer_norm.bias": 75.5,
71
+ "model.decoder.layers.1.encoder_attn_text_layer_norm.bias": 75.0,
72
+ "model.decoder.layers.1.encoder_attn.output_proj.bias": 74.5,
73
+ "model.encoder.layers.2.text_enhancer_layer.self_attn.value.weight": 74.0,
74
+ "model.decoder.layers.1.self_attn.out_proj.bias": 74.0,
75
+ "model.decoder.layers.1.self_attn_layer_norm.bias": 74.0,
76
+ "model.decoder.layers.1.encoder_attn_text.out_proj.bias": 74.0,
77
+ "model.encoder.layers.0.text_enhancer_layer.fc2.weight": 73.0,
78
+ "model.decoder.layers.0.final_layer_norm.bias": 72.0,
79
+ "model.encoder.layers.1.text_enhancer_layer.self_attn.out_proj.weight": 71.0,
80
+ "model.decoder.layers.2.fc2.bias": 71.0,
81
+ "model.encoder.layers.5.text_enhancer_layer.fc2.bias": 70.0,
82
+ "model.decoder.layers.0.fc2.bias": 69.5,
83
+ "model.encoder.layers.5.text_enhancer_layer.layer_norm_before.bias": 65.5,
84
+ "model.decoder.layers.5.self_attn.value.weight": 65.5,
85
+ "model.decoder.layers.0.encoder_attn_text.out_proj.bias": 65.0,
86
+ "model.decoder.layers.0.encoder_attn_layer_norm.bias": 65.0,
87
+ "model.decoder.layers.0.self_attn_layer_norm.bias": 64.5,
88
+ "model.decoder.layers.0.encoder_attn_text_layer_norm.bias": 64.5,
89
+ "model.decoder.layers.0.encoder_attn.output_proj.bias": 64.5,
90
+ "model.decoder.layers.5.self_attn.out_proj.weight": 62.5,
91
+ "model.decoder.layers.3.final_layer_norm.bias": 61.0,
92
+ "model.decoder.layers.2.encoder_attn_layer_norm.bias": 60.5,
93
+ "model.decoder.layers.2.encoder_attn_text_layer_norm.bias": 59.5,
94
+ "model.decoder.layers.2.encoder_attn.output_proj.bias": 59.5,
95
+ "model.decoder.layers.2.self_attn_layer_norm.bias": 58.75,
96
+ "model.decoder.layers.2.encoder_attn_text.out_proj.bias": 58.75,
97
+ "model.decoder.layers.2.self_attn.out_proj.bias": 58.5,
98
+ "model.encoder.layers.5.text_enhancer_layer.self_attn.out_proj.bias": 57.5,
99
+ "model.decoder.layers.5.encoder_attn.value_proj.weight": 56.75,
100
+ "model.encoder.layers.4.text_enhancer_layer.layer_norm_after.bias": 56.5,
101
+ "model.encoder.layers.5.fusion_layer.layer_norm_text.bias": 56.5,
102
+ "model.decoder.layers.3.fc2.bias": 55.5,
103
+ "model.decoder.layers.5.encoder_attn.output_proj.weight": 55.0,
104
+ "model.encoder.layers.1.text_enhancer_layer.self_attn.value.weight": 52.5,
105
+ "model.encoder.layers.5.deformable_layer.fc2.weight": 49.5,
106
+ "model.decoder.layer_norm.weight": 49.0,
107
+ "model.encoder.layers.0.text_enhancer_layer.fc1.weight": 47.0,
108
+ "model.encoder.layers.5.text_enhancer_layer.layer_norm_after.weight": 46.25,
109
+ "model.decoder.layers.3.encoder_attn_layer_norm.bias": 46.0,
110
+ "model.decoder.layers.3.encoder_attn_text_layer_norm.bias": 45.25,
111
+ "model.encoder.layers.4.text_enhancer_layer.fc2.bias": 45.0,
112
+ "model.decoder.layers.3.encoder_attn.output_proj.bias": 45.0,
113
+ "model.decoder.layers.3.self_attn_layer_norm.bias": 44.25,
114
+ "model.decoder.layers.3.encoder_attn_text.out_proj.bias": 44.25,
115
+ "model.decoder.layers.3.self_attn.out_proj.bias": 44.0,
116
+ "model.encoder.layers.4.text_enhancer_layer.layer_norm_before.bias": 43.0,
117
+ "model.decoder.layers.4.final_layer_norm.bias": 41.5,
118
+ "model.encoder.layers.0.text_enhancer_layer.self_attn.out_proj.weight": 40.75,
119
+ "model.decoder.layers.0.self_attn.out_proj.weight": 40.0,
120
+ "model.decoder.layers.1.fc1.bias": 40.0,
121
+ "model.encoder.layers.4.text_enhancer_layer.self_attn.out_proj.bias": 38.0,
122
+ "model.encoder.layers.4.deformable_layer.fc2.weight": 38.0,
123
+ "model.encoder.layers.4.fusion_layer.layer_norm_text.bias": 37.5,
124
+ "model.encoder.layers.3.text_enhancer_layer.layer_norm_after.bias": 37.25,
125
+ "model.decoder.layers.4.fc2.bias": 37.25,
126
+ "model.encoder.layers.5.text_enhancer_layer.layer_norm_before.weight": 37.0,
127
+ "model.decoder.layers.2.fc1.bias": 33.5,
128
+ "model.encoder.layers.0.text_enhancer_layer.self_attn.value.weight": 33.25,
129
+ "model.encoder.layers.5.deformable_layer.fc1.weight": 32.75,
130
+ "model.encoder.layers.3.deformable_layer.fc2.weight": 32.25,
131
+ "model.encoder.layers.4.text_enhancer_layer.layer_norm_after.weight": 32.25,
132
+ "model.encoder.layers.5.fusion_layer.layer_norm_text.weight": 32.25,
133
+ "model.decoder.layers.4.encoder_attn_layer_norm.bias": 31.5,
134
+ "model.decoder.layers.0.fc1.bias": 31.0,
135
+ "model.decoder.layers.4.encoder_attn_text_layer_norm.bias": 30.75,
136
+ "model.decoder.layers.4.encoder_attn.output_proj.bias": 30.625,
137
+ "model.encoder.layers.5.text_enhancer_layer.fc1.bias": 30.25,
138
+ "model.decoder.layers.4.encoder_attn_text.out_proj.bias": 30.25,
139
+ "model.decoder.layers.4.self_attn_layer_norm.bias": 30.125,
140
+ "model.decoder.layers.4.self_attn.out_proj.bias": 29.75,
141
+ "model.encoder.layers.3.text_enhancer_layer.fc2.bias": 29.25,
142
+ "model.decoder.bbox_embed.0.layers.2.weight": 28.75,
143
+ "model.decoder.layers.0.self_attn.value.weight": 28.0,
144
+ "model.encoder.layers.3.text_enhancer_layer.layer_norm_before.bias": 27.75,
145
+ "model.encoder.layers.2.deformable_layer.fc2.weight": 27.625,
146
+ "model.decoder.layers.3.fc1.bias": 27.625,
147
+ "model.input_proj_vision.0.0.weight": 26.375,
148
+ "model.encoder.layers.4.text_enhancer_layer.layer_norm_before.weight": 25.875,
149
+ "model.text_projection.bias": 25.25,
150
+ "model.decoder.layers.5.final_layer_norm.bias": 25.25,
151
+ "model.decoder.layers.1.final_layer_norm.weight": 25.125,
152
+ "model.encoder.layers.3.text_enhancer_layer.self_attn.out_proj.bias": 24.875,
153
+ "model.encoder.layers.1.deformable_layer.fc2.weight": 24.75,
154
+ "model.encoder.layers.3.fusion_layer.layer_norm_text.bias": 24.5,
155
+ "model.encoder.layers.2.text_enhancer_layer.layer_norm_after.bias": 24.375,
156
+ "model.encoder.layers.4.deformable_layer.fc1.weight": 24.25,
157
+ "model.decoder.layers.1.self_attn.value.bias": 24.125,
158
+ "model.decoder.layers.2.final_layer_norm.weight": 24.0,
159
+ "model.decoder.layers.1.encoder_attn_text.value.bias": 23.375,
160
+ "model.encoder.layers.3.text_enhancer_layer.layer_norm_after.weight": 22.75,
161
+ "model.encoder.layers.4.fusion_layer.layer_norm_text.weight": 22.75,
162
+ "model.backbone.scale_conv.1.weight": 22.625,
163
+ "model.decoder.layers.5.fc2.bias": 22.125,
164
+ "model.encoder.layers.0.deformable_layer.fc2.weight": 22.0,
165
+ "model.decoder.layers.3.final_layer_norm.weight": 22.0,
166
+ "model.backbone.scale_conv.2.weight": 21.5,
167
+ "model.input_proj_vision.1.0.weight": 21.375,
168
+ "model.backbone.scale_conv.0.weight": 21.125,
169
+ "model.decoder.layers.0.encoder_attn_text.value.bias": 19.875,
170
+ "model.encoder.layers.2.text_enhancer_layer.fc2.bias": 19.5,
171
+ "model.encoder.layers.4.text_enhancer_layer.fc1.bias": 19.0,
172
+ "model.decoder.layers.0.encoder_attn.value_proj.bias": 19.0,
173
+ "model.encoder.layers.5.deformable_layer.self_attn.value_proj.weight": 18.75,
174
+ "model.decoder.layers.1.encoder_attn_layer_norm.weight": 18.75,
175
+ "model.encoder.layers.2.text_enhancer_layer.layer_norm_before.bias": 18.625,
176
+ "model.decoder.layers.1.encoder_attn.value_proj.bias": 18.625,
177
+ "model.decoder.layers.4.fc1.bias": 18.625,
178
+ "model.input_proj_vision.2.0.weight": 18.125,
179
+ "model.decoder.layers.2.encoder_attn_layer_norm.weight": 17.875,
180
+ "model.decoder.layers.5.encoder_attn_layer_norm.bias": 17.875,
181
+ "model.encoder.layers.5.deformable_layer.self_attn.output_proj.weight": 17.625,
182
+ "model.decoder.layers.5.encoder_attn.output_proj.bias": 17.625,
183
+ "model.encoder.layers.5.text_enhancer_layer.self_attn.value.bias": 17.5,
184
+ "model.decoder.layers.1.encoder_attn_text_layer_norm.weight": 17.5,
185
+ "model.decoder.layers.5.encoder_attn_text_layer_norm.bias": 17.5,
186
+ "model.encoder.layers.3.text_enhancer_layer.layer_norm_before.weight": 17.25,
187
+ "model.decoder.layers.2.encoder_attn_text.value.bias": 17.25,
188
+ "model.decoder.layers.2.encoder_attn_text_layer_norm.weight": 17.25,
189
+ "model.decoder.layers.2.self_attn.value.bias": 17.0,
190
+ "model.decoder.layers.2.encoder_attn.value_proj.bias": 17.0,
191
+ "model.decoder.layers.5.self_attn_layer_norm.bias": 17.0,
192
+ "model.decoder.layers.5.encoder_attn_text.out_proj.bias": 17.0,
193
+ "model.input_proj_vision.3.0.weight": 16.875,
194
+ "model.encoder.layers.2.text_enhancer_layer.self_attn.out_proj.bias": 16.75,
195
+ "model.encoder.layers.1.text_enhancer_layer.layer_norm_after.bias": 16.625,
196
+ "model.encoder.layers.2.fusion_layer.layer_norm_text.bias": 16.625,
197
+ "model.decoder.layers.5.self_attn.out_proj.bias": 16.625,
198
+ "model.decoder.layers.2.self_attn_layer_norm.weight": 16.375,
199
+ "model.decoder.layers.4.final_layer_norm.weight": 16.0,
200
+ "model.backbone.scale_conv.3.weight": 15.6875,
201
+ "model.decoder.layers.3.encoder_attn_layer_norm.weight": 15.6875,
202
+ "model.decoder.layers.1.self_attn_layer_norm.weight": 15.625,
203
+ "model.decoder.layers.3.encoder_attn_text_layer_norm.weight": 15.3125,
204
+ "model.encoder.layers.2.text_enhancer_layer.layer_norm_after.weight": 15.0,
205
+ "model.encoder.layers.3.fusion_layer.layer_norm_text.weight": 15.0,
206
+ "model.decoder.layers.0.final_layer_norm.weight": 14.6875,
207
+ "model.encoder.layers.4.deformable_layer.self_attn.value_proj.weight": 14.375,
208
+ "model.decoder.layers.3.self_attn_layer_norm.weight": 14.375,
209
+ "model.encoder.layers.2.deformable_layer.fc1.weight": 14.1875,
210
+ "model.encoder.layers.4.deformable_layer.self_attn.output_proj.weight": 13.625,
211
+ "model.encoder.layers.1.text_enhancer_layer.fc2.bias": 13.375,
212
+ "model.decoder.layers.3.self_attn.value.bias": 13.3125,
213
+ "model.encoder.layers.1.text_enhancer_layer.layer_norm_before.bias": 13.125,
214
+ "model.decoder.layers.3.encoder_attn.value_proj.bias": 13.125,
215
+ "model.decoder.layers.3.encoder_attn_text.value.bias": 13.0,
216
+ "model.encoder.layers.3.text_enhancer_layer.fc1.bias": 12.4375,
217
+ "model.decoder.layers.4.encoder_attn_layer_norm.weight": 12.375,
218
+ "model.encoder.layers.4.text_enhancer_layer.self_attn.value.bias": 12.0,
219
+ "model.decoder.layers.5.fc1.bias": 12.0,
220
+ "model.decoder.layers.4.encoder_attn_text_layer_norm.weight": 11.9375,
221
+ "model.encoder.layers.2.text_enhancer_layer.layer_norm_before.weight": 11.875,
222
+ "model.encoder.layers.1.text_enhancer_layer.self_attn.out_proj.bias": 11.6875,
223
+ "model.decoder.layers.0.encoder_attn_layer_norm.weight": 11.625,
224
+ "model.encoder.layers.0.text_enhancer_layer.layer_norm_after.bias": 11.5625,
225
+ "model.encoder.layers.1.fusion_layer.layer_norm_text.bias": 11.5625,
226
+ "model.decoder.layers.4.self_attn_layer_norm.weight": 11.4375,
227
+ "model.encoder.layers.1.deformable_layer.fc1.weight": 11.125,
228
+ "model.encoder.layers.3.deformable_layer.self_attn.value_proj.weight": 10.9375,
229
+ "model.encoder.layers.1.text_enhancer_layer.layer_norm_after.weight": 10.8125,
230
+ "model.encoder.layers.2.fusion_layer.layer_norm_text.weight": 10.8125,
231
+ "model.encoder.layers.3.deformable_layer.self_attn.output_proj.weight": 10.25,
232
+ "model.decoder.layers.5.final_layer_norm.weight": 10.25,
233
+ "model.decoder.layers.0.encoder_attn_text_layer_norm.weight": 9.9375,
234
+ "model.encoder.layers.0.text_enhancer_layer.fc2.bias": 9.6875,
235
+ "model.encoder.layers.0.deformable_layer.fc1.weight": 9.5625,
236
+ "model.decoder.layers.4.encoder_attn_text.value.bias": 9.5,
237
+ "model.encoder.layers.0.text_enhancer_layer.layer_norm_before.bias": 9.375,
238
+ "model.encoder.layers.2.deformable_layer.self_attn.value_proj.weight": 9.3125,
239
+ "model.encoder.layers.2.deformable_layer.self_attn.output_proj.weight": 8.9375,
240
+ "model.encoder.layers.0.text_enhancer_layer.self_attn.out_proj.bias": 8.8125,
241
+ "model.encoder.layers.5.fusion_layer.text_param": 8.8125,
242
+ "model.decoder.layers.4.encoder_attn.value_proj.bias": 8.8125,
243
+ "model.decoder.layers.4.self_attn.value.bias": 8.6875,
244
+ "model.encoder.layers.0.fusion_layer.layer_norm_text.bias": 8.625,
245
+ "model.encoder.layers.1.text_enhancer_layer.layer_norm_before.weight": 8.5,
246
+ "model.encoder.layers.2.text_enhancer_layer.fc1.bias": 8.375,
247
+ "model.decoder.layers.0.self_attn_layer_norm.weight": 8.375,
248
+ "model.encoder.layers.3.text_enhancer_layer.self_attn.value.bias": 7.78125,
249
+ "model.encoder.layers.0.text_enhancer_layer.layer_norm_after.weight": 7.5,
250
+ "model.encoder.layers.1.fusion_layer.layer_norm_text.weight": 7.5,
251
+ "model.encoder.layers.5.deformable_layer.final_layer_norm.bias": 7.4375,
252
+ "model.decoder.layers.5.encoder_attn_layer_norm.weight": 7.34375,
253
+ "model.encoder.layers.1.deformable_layer.self_attn.value_proj.weight": 7.25,
254
+ "model.decoder.layers.5.encoder_attn_text_layer_norm.weight": 7.09375,
255
+ "model.encoder.layers.1.deformable_layer.self_attn.output_proj.weight": 6.875,
256
+ "model.encoder.layers.4.fusion_layer.text_param": 6.78125,
257
+ "model.decoder.layers.5.self_attn_layer_norm.weight": 6.6875,
258
+ "model.encoder.layers.5.deformable_layer.fc2.bias": 6.5,
259
+ "model.encoder.layers.5.deformable_layer.self_attn_layer_norm.bias": 6.15625,
260
+ "model.encoder.layers.5.deformable_layer.self_attn.output_proj.bias": 6.09375,
261
+ "model.encoder.layers.4.deformable_layer.final_layer_norm.bias": 6.03125,
262
+ "model.encoder.layers.5.fusion_layer.layer_norm_vision.bias": 6.03125,
263
+ "model.encoder.layers.1.text_enhancer_layer.fc1.bias": 5.96875,
264
+ "model.encoder.layers.0.text_enhancer_layer.layer_norm_before.weight": 5.875,
265
+ "model.encoder.layers.0.deformable_layer.self_attn.value_proj.weight": 5.78125,
266
+ "model.encoder.layers.0.deformable_layer.self_attn.output_proj.weight": 5.59375,
267
+ "model.encoder.layers.4.deformable_layer.fc2.bias": 5.5,
268
+ "model.encoder.layers.5.deformable_layer.final_layer_norm.weight": 5.5,
269
+ "model.encoder.layers.0.fusion_layer.layer_norm_text.weight": 5.4375,
270
+ "model.decoder.layers.5.self_attn.value.bias": 5.4375,
271
+ "model.decoder.layers.5.encoder_attn.value_proj.bias": 5.3125,
272
+ "model.encoder.layers.4.deformable_layer.self_attn_layer_norm.bias": 5.28125,
273
+ "model.decoder.layers.5.encoder_attn_text.value.bias": 5.28125,
274
+ "model.encoder.layers.4.deformable_layer.self_attn.output_proj.bias": 5.25,
275
+ "model.encoder.layers.3.deformable_layer.final_layer_norm.bias": 5.1875,
276
+ "model.encoder.layers.4.fusion_layer.layer_norm_vision.bias": 5.1875,
277
+ "model.encoder.layers.2.text_enhancer_layer.self_attn.value.bias": 5.03125,
278
+ "model.encoder.layers.3.deformable_layer.fc2.bias": 4.8125,
279
+ "model.encoder.layers.3.deformable_layer.self_attn_layer_norm.bias": 4.6875,
280
+ "model.encoder.layers.3.deformable_layer.self_attn.output_proj.bias": 4.65625,
281
+ "model.encoder.layers.2.deformable_layer.final_layer_norm.bias": 4.59375,
282
+ "model.encoder.layers.3.fusion_layer.layer_norm_vision.bias": 4.59375,
283
+ "model.encoder.layers.2.deformable_layer.fc2.bias": 4.28125,
284
+ "model.encoder.layers.3.fusion_layer.text_param": 4.25,
285
+ "model.encoder.layers.5.deformable_layer.self_attn_layer_norm.weight": 4.25,
286
+ "model.encoder.layers.2.fusion_layer.layer_norm_vision.bias": 4.09375,
287
+ "model.encoder.layers.2.deformable_layer.self_attn.output_proj.bias": 4.09375,
288
+ "model.encoder.layers.2.deformable_layer.self_attn_layer_norm.bias": 4.09375,
289
+ "model.encoder.layers.4.deformable_layer.final_layer_norm.weight": 4.09375,
290
+ "model.encoder.layers.5.fusion_layer.layer_norm_vision.weight": 4.09375,
291
+ "model.encoder.layers.1.deformable_layer.final_layer_norm.bias": 4.0625,
292
+ "model.encoder.layers.0.text_enhancer_layer.fc1.bias": 3.921875,
293
+ "model.encoder.layers.1.text_enhancer_layer.self_attn.value.bias": 3.921875,
294
+ "model.encoder.layers.1.deformable_layer.fc2.bias": 3.84375,
295
+ "model.encoder.layers.0.deformable_layer.final_layer_norm.bias": 3.78125,
296
+ "model.encoder.layers.1.fusion_layer.layer_norm_vision.bias": 3.78125,
297
+ "model.encoder.layers.1.deformable_layer.self_attn.output_proj.bias": 3.78125,
298
+ "model.encoder.layers.1.deformable_layer.self_attn_layer_norm.bias": 3.78125,
299
+ "model.encoder.layers.0.deformable_layer.fc2.bias": 3.625,
300
+ "model.encoder.layers.0.deformable_layer.self_attn.output_proj.bias": 3.546875,
301
+ "model.encoder.layers.0.deformable_layer.self_attn_layer_norm.bias": 3.546875,
302
+ "model.encoder.layers.0.fusion_layer.layer_norm_vision.bias": 3.515625,
303
+ "model.encoder.layers.5.deformable_layer.fc1.bias": 3.28125,
304
+ "model.encoder.layers.4.deformable_layer.self_attn_layer_norm.weight": 3.21875,
305
+ "model.encoder.layers.4.fusion_layer.layer_norm_vision.weight": 3.109375,
306
+ "model.encoder.layers.3.deformable_layer.final_layer_norm.weight": 3.09375,
307
+ "model.encoder.layers.0.text_enhancer_layer.self_attn.value.bias": 2.84375,
308
+ "model.encoder.layers.4.deformable_layer.fc1.bias": 2.71875,
309
+ "model.encoder.layers.2.fusion_layer.text_param": 2.6875,
310
+ "model.encoder.layers.3.deformable_layer.self_attn_layer_norm.weight": 2.625,
311
+ "model.encoder.layers.2.deformable_layer.final_layer_norm.weight": 2.59375,
312
+ "model.encoder.layers.3.fusion_layer.layer_norm_vision.weight": 2.59375,
313
+ "model.input_proj_vision.0.0.bias": 2.296875,
314
+ "model.encoder.layers.3.deformable_layer.fc1.bias": 2.265625,
315
+ "model.input_proj_vision.1.0.bias": 2.234375,
316
+ "model.encoder.layers.1.fusion_layer.text_param": 2.15625,
317
+ "model.decoder.layers.1.encoder_attn.attention_weights.weight": 2.109375,
318
+ "model.encoder.layers.1.deformable_layer.final_layer_norm.weight": 2.015625,
319
+ "model.encoder.layers.2.fusion_layer.layer_norm_vision.weight": 2.015625,
320
+ "model.encoder.layers.2.deformable_layer.self_attn_layer_norm.weight": 2.015625,
321
+ "model.encoder.layers.2.deformable_layer.fc1.bias": 2.0,
322
+ "model.encoder.layers.5.deformable_layer.self_attn.value_proj.bias": 1.9921875,
323
+ "model.decoder.layers.3.encoder_attn.attention_weights.weight": 1.953125,
324
+ "model.backbone.scale_conv.0.bias": 1.7890625,
325
+ "model.encoder.layers.1.deformable_layer.fc1.bias": 1.765625,
326
+ "model.encoder.layers.4.deformable_layer.self_attn.value_proj.bias": 1.7109375,
327
+ "model.encoder.layers.1.deformable_layer.self_attn_layer_norm.weight": 1.6875,
328
+ "model.encoder.layers.0.deformable_layer.final_layer_norm.weight": 1.6484375,
329
+ "model.encoder.layers.1.fusion_layer.layer_norm_vision.weight": 1.6484375,
330
+ "model.encoder.layers.0.deformable_layer.fc1.bias": 1.6328125,
331
+ "model.input_proj_vision.2.0.bias": 1.6171875,
332
+ "model.decoder.bbox_embed.0.layers.0.weight": 1.53125,
333
+ "model.encoder.layers.0.deformable_layer.self_attn_layer_norm.weight": 1.4375,
334
+ "model.encoder.layers.3.deformable_layer.self_attn.value_proj.bias": 1.4375,
335
+ "model.encoder.layers.0.fusion_layer.layer_norm_vision.weight": 1.4140625,
336
+ "model.encoder.layers.2.deformable_layer.self_attn.value_proj.bias": 1.3984375,
337
+ "model.encoder.layers.0.fusion_layer.text_param": 1.3828125,
338
+ "model.decoder.layers.1.self_attn.key.weight": 1.3125,
339
+ "model.decoder.layers.4.encoder_attn.attention_weights.weight": 1.3046875,
340
+ "model.decoder.layers.2.encoder_attn.attention_weights.weight": 1.21875,
341
+ "model.encoder.layers.1.deformable_layer.self_attn.value_proj.bias": 1.1953125,
342
+ "model.decoder.layers.2.self_attn.key.weight": 1.1796875,
343
+ "model.decoder.layers.1.self_attn.query.weight": 1.15625,
344
+ "model.decoder.layers.2.self_attn.query.weight": 1.125,
345
+ "model.decoder.layers.3.self_attn.query.weight": 1.125,
346
+ "model.encoder.layers.3.deformable_layer.self_attn.attention_weights.weight": 1.078125,
347
+ "model.input_proj_vision.0.1.bias": 1.0625,
348
+ "model.encoder.layers.0.deformable_layer.self_attn.value_proj.bias": 1.0625,
349
+ "model.input_proj_vision.1.1.bias": 1.046875,
350
+ "model.encoder.layers.5.deformable_layer.self_attn.attention_weights.weight": 1.03125,
351
+ "model.decoder.layers.0.encoder_attn.attention_weights.weight": 1.03125,
352
+ "model.backbone.scale_norm.0.bias": 1.0234375,
353
+ "model.decoder.layers.3.self_attn.key.weight": 1.015625,
354
+ "model.backbone.scale_conv.1.bias": 0.98046875,
355
+ "model.input_proj_vision.3.0.bias": 0.9765625,
356
+ "model.backbone.scale_norm.1.bias": 0.92578125,
357
+ "model.encoder.layers.4.deformable_layer.self_attn.attention_weights.weight": 0.91015625,
358
+ "model.encoder.layers.4.fusion_layer.vision_param": 0.90234375,
359
+ "model.encoder.layers.5.fusion_layer.vision_param": 0.87109375,
360
+ "model.input_proj_vision.2.1.bias": 0.84375,
361
+ "model.encoder.layers.2.deformable_layer.self_attn.attention_weights.weight": 0.7890625,
362
+ "model.decoder.layers.5.encoder_attn.attention_weights.weight": 0.7421875,
363
+ "model.encoder.layers.3.fusion_layer.vision_param": 0.73046875,
364
+ "model.backbone.scale_norm.2.bias": 0.703125,
365
+ "model.decoder.layers.4.self_attn.query.weight": 0.6796875,
366
+ "model.input_proj_vision.3.1.bias": 0.66015625,
367
+ "model.encoder.layers.2.fusion_layer.vision_param": 0.640625,
368
+ "model.input_proj_vision.3.1.weight": 0.62109375,
369
+ "model.encoder.layers.1.deformable_layer.self_attn.attention_weights.weight": 0.62109375,
370
+ "model.input_proj_vision.2.1.weight": 0.60546875,
371
+ "model.decoder.layers.4.self_attn.key.weight": 0.60546875,
372
+ "model.decoder.reference_points_head.layers.1.weight": 0.58203125,
373
+ "model.input_proj_vision.0.1.weight": 0.54296875,
374
+ "model.encoder.layers.5.text_enhancer_layer.self_attn.query.weight": 0.53125,
375
+ "model.backbone.scale_norm.0.weight": 0.5078125,
376
+ "model.input_proj_vision.1.1.weight": 0.5078125,
377
+ "model.decoder.bbox_embed.0.layers.1.weight": 0.50390625,
378
+ "model.encoder.layers.5.text_enhancer_layer.self_attn.key.weight": 0.49609375,
379
+ "model.encoder.layers.0.deformable_layer.self_attn.attention_weights.weight": 0.494140625,
380
+ "model.backbone.scale_conv.2.bias": 0.4765625,
381
+ "model.encoder.layers.0.fusion_layer.vision_param": 0.474609375,
382
+ "model.encoder.layers.1.fusion_layer.vision_param": 0.44921875,
383
+ "model.backbone.scale_norm.3.bias": 0.423828125,
384
+ "model.backbone.scale_norm.1.weight": 0.384765625,
385
+ "model.decoder.layers.1.encoder_attn.attention_weights.bias": 0.3828125,
386
+ "model.decoder.layers.5.self_attn.key.weight": 0.37890625,
387
+ "model.encoder.layers.0.text_enhancer_layer.self_attn.key.weight": 0.376953125,
388
+ "model.encoder.layers.1.text_enhancer_layer.self_attn.query.weight": 0.376953125,
389
+ "model.encoder.layers.1.text_enhancer_layer.self_attn.key.weight": 0.375,
390
+ "model.decoder.layers.5.self_attn.query.weight": 0.375,
391
+ "model.encoder.layers.0.text_enhancer_layer.self_attn.query.weight": 0.369140625,
392
+ "model.backbone.scale_norm.2.weight": 0.337890625,
393
+ "model.decoder.layers.0.encoder_attn.attention_weights.bias": 0.32421875,
394
+ "model.backbone.scale_norm.3.weight": 0.310546875,
395
+ "model.encoder.layers.4.text_enhancer_layer.self_attn.query.weight": 0.30859375,
396
+ "model.decoder.layers.0.encoder_attn.sampling_offsets.weight": 0.302734375,
397
+ "model.encoder.layers.4.text_enhancer_layer.self_attn.key.weight": 0.2890625,
398
+ "model.encoder.layers.3.text_enhancer_layer.self_attn.query.weight": 0.236328125,
399
+ "model.decoder.layers.1.self_attn.query.bias": 0.2353515625,
400
+ "model.encoder.layers.3.text_enhancer_layer.self_attn.key.weight": 0.224609375,
401
+ "model.decoder.layers.3.encoder_attn.attention_weights.bias": 0.2080078125,
402
+ "model.encoder.layers.5.deformable_layer.self_attn.sampling_offsets.weight": 0.1953125,
403
+ "model.decoder.layers.2.encoder_attn.sampling_offsets.weight": 0.1884765625,
404
+ "model.encoder.layers.5.fusion_layer.attn.out_text_proj.weight": 0.1865234375,
405
+ "model.encoder.layers.4.deformable_layer.self_attn.sampling_offsets.weight": 0.1787109375,
406
+ "model.decoder.layers.1.encoder_attn.sampling_offsets.weight": 0.1767578125,
407
+ "model.encoder.layers.2.text_enhancer_layer.self_attn.key.weight": 0.1640625,
408
+ "model.decoder.layers.2.self_attn.query.bias": 0.1640625,
409
+ "model.decoder.layers.1.encoder_attn_text.query.weight": 0.1611328125,
410
+ "model.decoder.layers.3.encoder_attn.sampling_offsets.weight": 0.1611328125,
411
+ "model.decoder.layers.2.encoder_attn.attention_weights.bias": 0.16015625,
412
+ "model.backbone.scale_conv.3.bias": 0.1572265625,
413
+ "model.encoder.layers.2.text_enhancer_layer.self_attn.query.weight": 0.154296875,
414
+ "model.encoder.layers.0.deformable_layer.self_attn.sampling_offsets.weight": 0.15234375,
415
+ "model.encoder.layers.2.deformable_layer.self_attn.sampling_offsets.weight": 0.150390625,
416
+ "model.encoder.layers.3.deformable_layer.self_attn.sampling_offsets.weight": 0.150390625,
417
+ "model.encoder.layers.1.deformable_layer.self_attn.sampling_offsets.weight": 0.1455078125,
418
+ "model.decoder.layers.1.encoder_attn_text.key.weight": 0.1376953125,
419
+ "model.encoder.layers.5.fusion_layer.attn.values_vision_proj.weight": 0.13671875,
420
+ "model.encoder.layers.4.fusion_layer.attn.out_text_proj.weight": 0.1318359375,
421
+ "model.decoder.layers.2.encoder_attn_text.query.weight": 0.130859375,
422
+ "model.decoder.reference_points_head.layers.0.weight": 0.130859375,
423
+ "model.decoder.layers.2.encoder_attn_text.key.weight": 0.1298828125,
424
+ "model.decoder.layers.3.self_attn.query.bias": 0.1279296875,
425
+ "model.decoder.layers.3.encoder_attn_text.query.weight": 0.1171875,
426
+ "model.decoder.layers.4.encoder_attn.attention_weights.bias": 0.1171875,
427
+ "model.decoder.layers.3.encoder_attn_text.key.weight": 0.1162109375,
428
+ "model.decoder.layers.4.encoder_attn_text.key.weight": 0.10791015625,
429
+ "model.decoder.layers.4.encoder_attn_text.query.weight": 0.095703125,
430
+ "model.encoder.layers.4.fusion_layer.attn.values_vision_proj.weight": 0.09033203125,
431
+ "model.encoder.layers.3.fusion_layer.attn.out_text_proj.weight": 0.08935546875,
432
+ "model.encoder.layers.3.deformable_layer.self_attn.attention_weights.bias": 0.07568359375,
433
+ "model.decoder.layers.4.encoder_attn.sampling_offsets.weight": 0.06884765625,
434
+ "model.encoder.layers.5.deformable_layer.self_attn.attention_weights.bias": 0.0673828125,
435
+ "model.decoder.layers.0.encoder_attn_text.query.weight": 0.06591796875,
436
+ "model.encoder.layers.3.fusion_layer.attn.values_vision_proj.weight": 0.064453125,
437
+ "model.decoder.layers.4.self_attn.query.bias": 0.06396484375,
438
+ "model.encoder.layers.4.deformable_layer.self_attn.attention_weights.bias": 0.061767578125,
439
+ "model.encoder.layers.2.fusion_layer.attn.out_text_proj.weight": 0.060546875,
440
+ "model.decoder.layers.0.encoder_attn_text.key.weight": 0.0595703125,
441
+ "model.decoder.layers.5.encoder_attn.attention_weights.bias": 0.05859375,
442
+ "model.encoder.layers.2.deformable_layer.self_attn.attention_weights.bias": 0.056396484375,
443
+ "model.encoder.layers.1.fusion_layer.attn.out_text_proj.weight": 0.04833984375,
444
+ "model.decoder.layers.5.encoder_attn_text.query.weight": 0.04736328125,
445
+ "model.encoder.layers.1.deformable_layer.self_attn.attention_weights.bias": 0.044677734375,
446
+ "model.encoder.layers.2.fusion_layer.attn.values_vision_proj.weight": 0.04443359375,
447
+ "model.decoder.layers.5.encoder_attn_text.key.weight": 0.04248046875,
448
+ "model.decoder.layers.5.encoder_attn.sampling_offsets.weight": 0.037353515625,
449
+ "model.encoder.layers.0.deformable_layer.self_attn.attention_weights.bias": 0.035888671875,
450
+ "model.encoder.layers.0.fusion_layer.attn.out_text_proj.weight": 0.03564453125,
451
+ "model.decoder.layers.0.encoder_attn_text.query.bias": 0.032470703125,
452
+ "model.encoder.layers.1.fusion_layer.attn.values_vision_proj.weight": 0.0322265625,
453
+ "model.decoder.layers.1.encoder_attn_text.query.bias": 0.0322265625,
454
+ "model.decoder.layers.5.self_attn.query.bias": 0.03076171875,
455
+ "model.encoder.layers.5.text_enhancer_layer.self_attn.query.bias": 0.029296875,
456
+ "model.level_embed": 0.02587890625,
457
+ "model.encoder.layers.0.fusion_layer.attn.values_vision_proj.weight": 0.0250244140625,
458
+ "model.encoder.layers.5.fusion_layer.attn.out_text_proj.bias": 0.0234375,
459
+ "model.encoder.layers.0.text_enhancer_layer.self_attn.query.bias": 0.0230712890625,
460
+ "model.encoder.layers.1.text_enhancer_layer.self_attn.query.bias": 0.0225830078125,
461
+ "model.decoder.layers.2.encoder_attn_text.query.bias": 0.0186767578125,
462
+ "model.encoder.layers.4.fusion_layer.attn.out_text_proj.bias": 0.0179443359375,
463
+ "model.encoder.layers.4.text_enhancer_layer.self_attn.query.bias": 0.0172119140625,
464
+ "model.encoder.layers.5.fusion_layer.attn.values_text_proj.weight": 0.0172119140625,
465
+ "model.encoder.layers.2.fusion_layer.attn.values_text_proj.weight": 0.01611328125,
466
+ "model.encoder.layers.4.fusion_layer.attn.values_text_proj.weight": 0.0157470703125,
467
+ "model.encoder.layers.5.fusion_layer.attn.out_vision_proj.weight": 0.015625,
468
+ "model.encoder.layers.5.fusion_layer.attn.values_vision_proj.bias": 0.0152587890625,
469
+ "model.encoder.layers.2.fusion_layer.attn.out_vision_proj.weight": 0.01495361328125,
470
+ "model.encoder.layers.4.fusion_layer.attn.out_vision_proj.weight": 0.0147705078125,
471
+ "model.encoder.layers.3.fusion_layer.attn.out_text_proj.bias": 0.01446533203125,
472
+ "model.encoder.layers.3.fusion_layer.attn.out_vision_proj.weight": 0.01416015625,
473
+ "model.encoder.layers.3.fusion_layer.attn.values_text_proj.weight": 0.01409912109375,
474
+ "model.decoder.layers.3.encoder_attn_text.query.bias": 0.0130615234375,
475
+ "model.encoder.layers.3.text_enhancer_layer.self_attn.query.bias": 0.012939453125,
476
+ "model.encoder.layers.4.fusion_layer.attn.values_vision_proj.bias": 0.01141357421875,
477
+ "model.encoder.layers.2.fusion_layer.attn.out_text_proj.bias": 0.0106201171875,
478
+ "model.encoder.layers.1.fusion_layer.attn.values_text_proj.weight": 0.00970458984375,
479
+ "model.encoder.layers.1.fusion_layer.attn.out_vision_proj.weight": 0.00927734375,
480
+ "model.encoder.layers.3.fusion_layer.attn.values_vision_proj.bias": 0.00921630859375,
481
+ "model.encoder.layers.1.fusion_layer.attn.out_text_proj.bias": 0.00897216796875,
482
+ "model.encoder.layers.0.fusion_layer.attn.values_text_proj.weight": 0.0089111328125,
483
+ "model.decoder.layers.4.encoder_attn_text.query.bias": 0.00885009765625,
484
+ "model.encoder.layers.2.text_enhancer_layer.self_attn.query.bias": 0.00836181640625,
485
+ "model.encoder.layers.0.fusion_layer.attn.out_vision_proj.weight": 0.00811767578125,
486
+ "model.encoder.layers.0.fusion_layer.attn.out_text_proj.bias": 0.0072021484375,
487
+ "model.encoder.layers.2.fusion_layer.attn.values_vision_proj.bias": 0.0069580078125,
488
+ "model.encoder.layers.5.fusion_layer.attn.vision_proj.weight": 0.005950927734375,
489
+ "model.encoder.layers.1.fusion_layer.attn.values_vision_proj.bias": 0.005706787109375,
490
+ "model.encoder.layers.5.fusion_layer.attn.text_proj.weight": 0.005584716796875,
491
+ "model.encoder.layers.0.fusion_layer.attn.values_vision_proj.bias": 0.004730224609375,
492
+ "model.decoder.layers.0.self_attn.query.bias": 0.004638671875,
493
+ "model.decoder.layers.5.encoder_attn_text.query.bias": 0.0038299560546875,
494
+ "model.encoder.layers.4.fusion_layer.attn.vision_proj.weight": 0.003692626953125,
495
+ "model.encoder.layers.4.fusion_layer.attn.text_proj.weight": 0.003692626953125,
496
+ "model.encoder.layers.3.fusion_layer.attn.text_proj.weight": 0.0024261474609375,
497
+ "model.encoder.layers.3.fusion_layer.attn.vision_proj.weight": 0.002410888671875,
498
+ "model.encoder.layers.0.fusion_layer.attn.text_proj.weight": 0.00213623046875,
499
+ "model.encoder.layers.2.fusion_layer.attn.text_proj.weight": 0.00213623046875,
500
+ "model.encoder.layers.1.fusion_layer.attn.vision_proj.weight": 0.0021209716796875,
501
+ "model.encoder.layers.0.fusion_layer.attn.vision_proj.weight": 0.0020751953125,
502
+ "model.encoder.layers.1.fusion_layer.attn.text_proj.weight": 0.002044677734375,
503
+ "model.encoder.layers.2.fusion_layer.attn.vision_proj.weight": 0.0020294189453125,
504
+ "model.encoder.layers.2.fusion_layer.attn.out_vision_proj.bias": 0.00168609619140625,
505
+ "model.encoder.layers.5.fusion_layer.attn.out_vision_proj.bias": 0.00164794921875,
506
+ "model.encoder.layers.4.fusion_layer.attn.out_vision_proj.bias": 0.00154876708984375,
507
+ "model.encoder.layers.3.fusion_layer.attn.out_vision_proj.bias": 0.00148773193359375,
508
+ "model.encoder.layers.0.fusion_layer.attn.out_vision_proj.bias": 0.001220703125,
509
+ "model.encoder.layers.1.fusion_layer.attn.out_vision_proj.bias": 0.00115966796875,
510
+ "model.encoder.layers.2.fusion_layer.attn.values_text_proj.bias": 0.001129150390625,
511
+ "model.encoder.layers.5.fusion_layer.attn.values_text_proj.bias": 0.00109100341796875,
512
+ "model.encoder.layers.4.fusion_layer.attn.values_text_proj.bias": 0.00101470947265625,
513
+ "model.encoder.layers.3.fusion_layer.attn.values_text_proj.bias": 0.000934600830078125,
514
+ "model.encoder.layers.0.fusion_layer.attn.values_text_proj.bias": 0.00080108642578125,
515
+ "model.encoder.layers.1.fusion_layer.attn.values_text_proj.bias": 0.000751495361328125,
516
+ "model.decoder.layers.0.self_attn.query.weight": 0.000518798828125,
517
+ "model.decoder.layers.0.self_attn.key.weight": 0.00051116943359375,
518
+ "model.encoder.layers.5.fusion_layer.attn.text_proj.bias": 0.0003509521484375,
519
+ "model.encoder.layers.1.text_enhancer_layer.self_attn.key.bias": 0.0002956390380859375,
520
+ "model.encoder.layers.4.text_enhancer_layer.self_attn.key.bias": 0.0002689361572265625,
521
+ "model.encoder.layers.4.fusion_layer.attn.text_proj.bias": 0.0002346038818359375,
522
+ "model.encoder.layers.5.text_enhancer_layer.self_attn.key.bias": 0.000217437744140625,
523
+ "model.encoder.layers.2.text_enhancer_layer.self_attn.key.bias": 0.000202178955078125,
524
+ "model.encoder.layers.3.text_enhancer_layer.self_attn.key.bias": 0.0002002716064453125,
525
+ "model.encoder.layers.3.fusion_layer.attn.text_proj.bias": 0.00015735626220703125,
526
+ "model.encoder.layers.2.fusion_layer.attn.text_proj.bias": 0.00013637542724609375,
527
+ "model.encoder.layers.1.fusion_layer.attn.text_proj.bias": 0.00013065338134765625,
528
+ "model.encoder.layers.0.fusion_layer.attn.text_proj.bias": 0.00011968612670898438,
529
+ "model.encoder.layers.0.text_enhancer_layer.self_attn.key.bias": 7.867813110351562e-05,
530
+ "model.decoder.layers.0.encoder_attn_text.key.bias": 5.0067901611328125e-05,
531
+ "model.decoder.layers.1.self_attn.key.bias": 4.744529724121094e-05,
532
+ "model.decoder.layers.2.encoder_attn_text.key.bias": 4.673004150390625e-05,
533
+ "model.decoder.layers.1.encoder_attn_text.key.bias": 4.0531158447265625e-05,
534
+ "model.decoder.layers.2.self_attn.key.bias": 3.552436828613281e-05,
535
+ "model.decoder.layers.3.encoder_attn_text.key.bias": 3.457069396972656e-05,
536
+ "model.decoder.layers.4.encoder_attn_text.key.bias": 3.218650817871094e-05,
537
+ "model.decoder.layers.4.self_attn.key.bias": 2.9206275939941406e-05,
538
+ "model.decoder.layers.3.self_attn.key.bias": 2.86102294921875e-05,
539
+ "model.decoder.layers.5.self_attn.key.bias": 1.7642974853515625e-05,
540
+ "model.decoder.layers.5.encoder_attn_text.key.bias": 1.5854835510253906e-05,
541
+ "model.encoder.layers.0.fusion_layer.attn.vision_proj.bias": 1.52587890625e-05,
542
+ "model.encoder.layers.2.fusion_layer.attn.vision_proj.bias": 1.0848045349121094e-05,
543
+ "model.encoder.layers.3.fusion_layer.attn.vision_proj.bias": 7.4803829193115234e-06,
544
+ "model.encoder.layers.1.fusion_layer.attn.vision_proj.bias": 7.271766662597656e-06,
545
+ "model.encoder.layers.4.fusion_layer.attn.vision_proj.bias": 2.8461217880249023e-06,
546
+ "model.decoder.layers.0.self_attn.key.bias": 2.8312206268310547e-06,
547
+ "model.encoder.layers.5.fusion_layer.attn.vision_proj.bias": 2.4437904357910156e-06,
548
+ "model.enc_output.weight": 0,
549
+ "model.enc_output.bias": 0,
550
+ "model.enc_output_norm.weight": 0,
551
+ "model.enc_output_norm.bias": 0,
552
+ "model.encoder_output_bbox_embed.layers.0.weight": 0,
553
+ "model.encoder_output_bbox_embed.layers.1.weight": 0,
554
+ "model.encoder_output_bbox_embed.layers.2.weight": 0
555
+ }
results/track_record/hidden_states.joblib ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3d9a3c02c513bdce8c2b46a74201630be782957daf89b6581450beacd6a3bf74
3
+ size 2703785
results/track_record/kaldi_fbank1.joblib ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4eeec9ba52cef30deb46dbc8560acd94ad42dbe758a28131e921e63b3c90b79c
3
+ size 639125
results/track_record/kaldi_fbank2.joblib ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2b1daa3f2d74a2bf822b874426a49ecf4d6b06ea571fa7b667e67ab731e5b89c
3
+ size 639125
results/track_record/labels.joblib ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c2f0902a37d68d20c37b2e44e87b02e3fd5237213cc9af9a36c503ebb98c3cb0
3
+ size 27770
results/track_record/load_print_first.joblib ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:20b667d2d54dbb85655beb657bd6ffcf800cc2cfbe3445bb0d5ae0ed10912883
3
+ size 2560367
results/track_record/load_print_first.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ {
2
+ "file_path": "/root/autodl-tmp/gaudio/data/AudioSetStrongLabel_Demo/train/Y--12UOziMF0.wav"
3
+ }
results/track_record/map_print_first.joblib ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:66808f07152aa390e38bfd6a896edbf8c1b9eef6071b031791e2cebb714dff8b
3
+ size 843696
results/track_record/map_print_first.json ADDED
@@ -0,0 +1,132 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "text": "ship. gunshot, gunfire. clapping. ratchet, pawl. yell. wind chime. whip. generic impact sounds.",
3
+ "labels": [
4
+ [
5
+ 23,
6
+ [
7
+ 7.974500060081482,
8
+ 0.07350000087171793
9
+ ]
10
+ ],
11
+ [
12
+ 24,
13
+ [
14
+ 7.974500060081482,
15
+ 0.07350000087171793
16
+ ]
17
+ ],
18
+ [
19
+ 25,
20
+ [
21
+ 7.974500060081482,
22
+ 0.07350000087171793
23
+ ]
24
+ ],
25
+ [
26
+ 23,
27
+ [
28
+ 8.258999586105347,
29
+ 0.06499999668449163
30
+ ]
31
+ ],
32
+ [
33
+ 24,
34
+ [
35
+ 8.258999586105347,
36
+ 0.06499999668449163
37
+ ]
38
+ ],
39
+ [
40
+ 25,
41
+ [
42
+ 8.258999586105347,
43
+ 0.06499999668449163
44
+ ]
45
+ ],
46
+ [
47
+ 23,
48
+ [
49
+ 8.762499690055847,
50
+ 0.048500001430511475
51
+ ]
52
+ ],
53
+ [
54
+ 24,
55
+ [
56
+ 8.762499690055847,
57
+ 0.048500001430511475
58
+ ]
59
+ ],
60
+ [
61
+ 25,
62
+ [
63
+ 8.762499690055847,
64
+ 0.048500001430511475
65
+ ]
66
+ ],
67
+ [
68
+ 23,
69
+ [
70
+ 9.322999715805054,
71
+ 0.0570000009611249
72
+ ]
73
+ ],
74
+ [
75
+ 24,
76
+ [
77
+ 9.322999715805054,
78
+ 0.0570000009611249
79
+ ]
80
+ ],
81
+ [
82
+ 25,
83
+ [
84
+ 9.322999715805054,
85
+ 0.0570000009611249
86
+ ]
87
+ ],
88
+ [
89
+ 23,
90
+ [
91
+ 9.635499715805054,
92
+ 0.04449999891221523
93
+ ]
94
+ ],
95
+ [
96
+ 24,
97
+ [
98
+ 9.635499715805054,
99
+ 0.04449999891221523
100
+ ]
101
+ ],
102
+ [
103
+ 25,
104
+ [
105
+ 9.635499715805054,
106
+ 0.04449999891221523
107
+ ]
108
+ ],
109
+ [
110
+ 23,
111
+ [
112
+ 9.921500086784363,
113
+ 0.07850000634789467
114
+ ]
115
+ ],
116
+ [
117
+ 24,
118
+ [
119
+ 9.921500086784363,
120
+ 0.07850000634789467
121
+ ]
122
+ ],
123
+ [
124
+ 25,
125
+ [
126
+ 9.921500086784363,
127
+ 0.07850000634789467
128
+ ]
129
+ ]
130
+ ],
131
+ "labels2text": "generic impact sounds generic impact sounds generic impact sounds generic impact sounds generic impact sounds generic impact sounds"
132
+ }
results/track_record/outputs_loss.joblib ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c697961aa4bb4e209039b7cba04f17ff52e055be5fa1957a7ba06e8d37a077a0
3
+ size 32697545
results/track_record/source_flatten.joblib ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:503df09f8c16a772c8fd33f3bce6fadf5799f6bbe659b6c784b7eae80aef8e43
3
+ size 7186853
results/track_record/text_features.joblib ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:faedb00e5dd7eb4417d60d5d0b72b196a9e44fea2ce30a9e919cb7153189cb12
3
+ size 946594
results/track_record/text_features_before.joblib ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7b9bbbf9c49c917ad91b5bc01ff69bf7030d35667d80f13d036b017da7ce6393
3
+ size 2838946
results/track_record/text_projection.joblib ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f6dcfe2f4ffbff15dcc44d73d6e467132656fb3e8d3d6a7e507049eda8eef26b
3
+ size 69882271
results/track_record/waveform1.joblib ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:264819631ff1e1525c928e73a7b48162a4077622eee9e94c7781cfa287807b20
3
+ size 2560410
results/track_record/waveform2.joblib ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:762a570f0547310f05b48de0aa7bd478199dfb2acf76a49d133203410f599b1d
3
+ size 2560410