sanchit-gandhi HF staff commited on
Commit
467af1f
1 Parent(s): 12e08ef

End of training

Browse files
Files changed (5) hide show
  1. README.md +5 -3
  2. all_results.json +16 -0
  3. eval_results.json +10 -0
  4. train_results.json +9 -0
  5. trainer_state.json +2151 -0
README.md CHANGED
@@ -2,6 +2,8 @@
2
  license: cc-by-nc-4.0
3
  library_name: peft
4
  tags:
 
 
5
  - generated_from_trainer
6
  base_model: facebook/musicgen-melody-large
7
  model-index:
@@ -16,11 +18,11 @@ should probably proofread and complete it, then remove this comment. -->
16
  [<img src="https://raw.githubusercontent.com/wandb/assets/main/wandb-github-badge-28.svg" alt="Visualize in Weights & Biases" width="200" height="32"/>](https://wandb.ai/sanchit-gandhi/black-eyed-peas/runs/mbvcb2wa)
17
  # combined-artists-text-only-3-epochs
18
 
19
- This model is a fine-tuned version of [facebook/musicgen-melody-large](https://huggingface.co/facebook/musicgen-melody-large) on an unknown dataset.
20
  It achieves the following results on the evaluation set:
21
  - Loss: 5.1016
22
- - Text Cosine Sim: 0.1181
23
- - Audio Cosine Sim: 0.3217
24
 
25
  ## Model description
26
 
 
2
  license: cc-by-nc-4.0
3
  library_name: peft
4
  tags:
5
+ - text-to-audio
6
+ - sweet-dreambooths/combined-artists-v1
7
  - generated_from_trainer
8
  base_model: facebook/musicgen-melody-large
9
  model-index:
 
18
  [<img src="https://raw.githubusercontent.com/wandb/assets/main/wandb-github-badge-28.svg" alt="Visualize in Weights & Biases" width="200" height="32"/>](https://wandb.ai/sanchit-gandhi/black-eyed-peas/runs/mbvcb2wa)
19
  # combined-artists-text-only-3-epochs
20
 
21
+ This model is a fine-tuned version of [facebook/musicgen-melody-large](https://huggingface.co/facebook/musicgen-melody-large) on the SWEET-DREAMBOOTHS/COMBINED-ARTISTS-V1 - DEFAULT dataset.
22
  It achieves the following results on the evaluation set:
23
  - Loss: 5.1016
24
+ - Text Cosine Sim: 0.0961
25
+ - Audio Cosine Sim: 0.3267
26
 
27
  ## Model description
28
 
all_results.json ADDED
@@ -0,0 +1,16 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "epoch": 3.0,
3
+ "eval_audio_cosine_sim": 0.32668301463127136,
4
+ "eval_loss": 5.101603984832764,
5
+ "eval_runtime": 150.73,
6
+ "eval_samples": 9,
7
+ "eval_samples_per_second": 0.06,
8
+ "eval_steps_per_second": 0.02,
9
+ "eval_text_cosine_sim": 0.09613808989524841,
10
+ "total_flos": 6235958582540448.0,
11
+ "train_loss": 7.227042491989907,
12
+ "train_runtime": 3530.6424,
13
+ "train_samples": 1580,
14
+ "train_samples_per_second": 1.343,
15
+ "train_steps_per_second": 0.084
16
+ }
eval_results.json ADDED
@@ -0,0 +1,10 @@
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "epoch": 3.0,
3
+ "eval_audio_cosine_sim": 0.32668301463127136,
4
+ "eval_loss": 5.101603984832764,
5
+ "eval_runtime": 150.73,
6
+ "eval_samples": 9,
7
+ "eval_samples_per_second": 0.06,
8
+ "eval_steps_per_second": 0.02,
9
+ "eval_text_cosine_sim": 0.09613808989524841
10
+ }
train_results.json ADDED
@@ -0,0 +1,9 @@
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "epoch": 3.0,
3
+ "total_flos": 6235958582540448.0,
4
+ "train_loss": 7.227042491989907,
5
+ "train_runtime": 3530.6424,
6
+ "train_samples": 1580,
7
+ "train_samples_per_second": 1.343,
8
+ "train_steps_per_second": 0.084
9
+ }
trainer_state.json ADDED
@@ -0,0 +1,2151 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_metric": null,
3
+ "best_model_checkpoint": null,
4
+ "epoch": 3.0,
5
+ "eval_steps": 500,
6
+ "global_step": 297,
7
+ "is_hyper_param_search": false,
8
+ "is_local_process_zero": true,
9
+ "is_world_process_zero": true,
10
+ "log_history": [
11
+ {
12
+ "epoch": 0.010101010101010102,
13
+ "grad_norm": 1.4315768480300903,
14
+ "learning_rate": 2.98989898989899e-05,
15
+ "loss": 10.0726,
16
+ "step": 1
17
+ },
18
+ {
19
+ "epoch": 0.020202020202020204,
20
+ "grad_norm": 1.5613890886306763,
21
+ "learning_rate": 2.97979797979798e-05,
22
+ "loss": 10.1289,
23
+ "step": 2
24
+ },
25
+ {
26
+ "epoch": 0.030303030303030304,
27
+ "grad_norm": 1.4086546897888184,
28
+ "learning_rate": 2.96969696969697e-05,
29
+ "loss": 10.0402,
30
+ "step": 3
31
+ },
32
+ {
33
+ "epoch": 0.04040404040404041,
34
+ "grad_norm": 1.508208155632019,
35
+ "learning_rate": 2.9595959595959595e-05,
36
+ "loss": 10.0206,
37
+ "step": 4
38
+ },
39
+ {
40
+ "epoch": 0.050505050505050504,
41
+ "grad_norm": 1.6154247522354126,
42
+ "learning_rate": 2.9494949494949495e-05,
43
+ "loss": 10.0272,
44
+ "step": 5
45
+ },
46
+ {
47
+ "epoch": 0.06060606060606061,
48
+ "grad_norm": 1.8789210319519043,
49
+ "learning_rate": 2.9393939393939394e-05,
50
+ "loss": 10.0711,
51
+ "step": 6
52
+ },
53
+ {
54
+ "epoch": 0.0707070707070707,
55
+ "grad_norm": 1.879401683807373,
56
+ "learning_rate": 2.9292929292929294e-05,
57
+ "loss": 9.9488,
58
+ "step": 7
59
+ },
60
+ {
61
+ "epoch": 0.08080808080808081,
62
+ "grad_norm": 1.9335699081420898,
63
+ "learning_rate": 2.9191919191919193e-05,
64
+ "loss": 9.8084,
65
+ "step": 8
66
+ },
67
+ {
68
+ "epoch": 0.09090909090909091,
69
+ "grad_norm": 1.9985203742980957,
70
+ "learning_rate": 2.9090909090909093e-05,
71
+ "loss": 9.7694,
72
+ "step": 9
73
+ },
74
+ {
75
+ "epoch": 0.10101010101010101,
76
+ "grad_norm": 2.0941152572631836,
77
+ "learning_rate": 2.8989898989898992e-05,
78
+ "loss": 9.7268,
79
+ "step": 10
80
+ },
81
+ {
82
+ "epoch": 0.1111111111111111,
83
+ "grad_norm": 2.344831943511963,
84
+ "learning_rate": 2.8888888888888888e-05,
85
+ "loss": 9.7139,
86
+ "step": 11
87
+ },
88
+ {
89
+ "epoch": 0.12121212121212122,
90
+ "grad_norm": 2.27956485748291,
91
+ "learning_rate": 2.8787878787878788e-05,
92
+ "loss": 9.5774,
93
+ "step": 12
94
+ },
95
+ {
96
+ "epoch": 0.13131313131313133,
97
+ "grad_norm": 2.440272331237793,
98
+ "learning_rate": 2.8686868686868687e-05,
99
+ "loss": 9.4884,
100
+ "step": 13
101
+ },
102
+ {
103
+ "epoch": 0.1414141414141414,
104
+ "grad_norm": 2.7465505599975586,
105
+ "learning_rate": 2.8585858585858587e-05,
106
+ "loss": 9.4327,
107
+ "step": 14
108
+ },
109
+ {
110
+ "epoch": 0.15151515151515152,
111
+ "grad_norm": 2.5827443599700928,
112
+ "learning_rate": 2.8484848484848486e-05,
113
+ "loss": 9.2545,
114
+ "step": 15
115
+ },
116
+ {
117
+ "epoch": 0.16161616161616163,
118
+ "grad_norm": 2.7065556049346924,
119
+ "learning_rate": 2.8383838383838386e-05,
120
+ "loss": 9.2304,
121
+ "step": 16
122
+ },
123
+ {
124
+ "epoch": 0.1717171717171717,
125
+ "grad_norm": 2.944704294204712,
126
+ "learning_rate": 2.8282828282828285e-05,
127
+ "loss": 9.1568,
128
+ "step": 17
129
+ },
130
+ {
131
+ "epoch": 0.18181818181818182,
132
+ "grad_norm": 2.667189598083496,
133
+ "learning_rate": 2.8181818181818185e-05,
134
+ "loss": 9.0014,
135
+ "step": 18
136
+ },
137
+ {
138
+ "epoch": 0.1919191919191919,
139
+ "grad_norm": 2.928550958633423,
140
+ "learning_rate": 2.808080808080808e-05,
141
+ "loss": 8.9312,
142
+ "step": 19
143
+ },
144
+ {
145
+ "epoch": 0.20202020202020202,
146
+ "grad_norm": 2.930938482284546,
147
+ "learning_rate": 2.797979797979798e-05,
148
+ "loss": 8.8656,
149
+ "step": 20
150
+ },
151
+ {
152
+ "epoch": 0.21212121212121213,
153
+ "grad_norm": 2.750476121902466,
154
+ "learning_rate": 2.787878787878788e-05,
155
+ "loss": 8.7607,
156
+ "step": 21
157
+ },
158
+ {
159
+ "epoch": 0.2222222222222222,
160
+ "grad_norm": 2.8917646408081055,
161
+ "learning_rate": 2.777777777777778e-05,
162
+ "loss": 8.699,
163
+ "step": 22
164
+ },
165
+ {
166
+ "epoch": 0.23232323232323232,
167
+ "grad_norm": 2.8639166355133057,
168
+ "learning_rate": 2.767676767676768e-05,
169
+ "loss": 8.5703,
170
+ "step": 23
171
+ },
172
+ {
173
+ "epoch": 0.24242424242424243,
174
+ "grad_norm": 2.9358019828796387,
175
+ "learning_rate": 2.7575757575757578e-05,
176
+ "loss": 8.519,
177
+ "step": 24
178
+ },
179
+ {
180
+ "epoch": 0.25252525252525254,
181
+ "grad_norm": 3.063692092895508,
182
+ "learning_rate": 2.7474747474747478e-05,
183
+ "loss": 8.3564,
184
+ "step": 25
185
+ },
186
+ {
187
+ "epoch": 0.26262626262626265,
188
+ "grad_norm": 2.702965259552002,
189
+ "learning_rate": 2.7373737373737374e-05,
190
+ "loss": 8.3403,
191
+ "step": 26
192
+ },
193
+ {
194
+ "epoch": 0.2727272727272727,
195
+ "grad_norm": 2.8974392414093018,
196
+ "learning_rate": 2.7272727272727273e-05,
197
+ "loss": 8.217,
198
+ "step": 27
199
+ },
200
+ {
201
+ "epoch": 0.2828282828282828,
202
+ "grad_norm": 3.0662784576416016,
203
+ "learning_rate": 2.7171717171717173e-05,
204
+ "loss": 8.1575,
205
+ "step": 28
206
+ },
207
+ {
208
+ "epoch": 0.29292929292929293,
209
+ "grad_norm": 2.8070247173309326,
210
+ "learning_rate": 2.7070707070707072e-05,
211
+ "loss": 8.0814,
212
+ "step": 29
213
+ },
214
+ {
215
+ "epoch": 0.30303030303030304,
216
+ "grad_norm": 2.507457971572876,
217
+ "learning_rate": 2.696969696969697e-05,
218
+ "loss": 8.0387,
219
+ "step": 30
220
+ },
221
+ {
222
+ "epoch": 0.31313131313131315,
223
+ "grad_norm": 2.7476532459259033,
224
+ "learning_rate": 2.686868686868687e-05,
225
+ "loss": 7.96,
226
+ "step": 31
227
+ },
228
+ {
229
+ "epoch": 0.32323232323232326,
230
+ "grad_norm": 2.3512086868286133,
231
+ "learning_rate": 2.676767676767677e-05,
232
+ "loss": 7.8822,
233
+ "step": 32
234
+ },
235
+ {
236
+ "epoch": 0.3333333333333333,
237
+ "grad_norm": 2.1365387439727783,
238
+ "learning_rate": 2.6666666666666667e-05,
239
+ "loss": 7.8282,
240
+ "step": 33
241
+ },
242
+ {
243
+ "epoch": 0.3434343434343434,
244
+ "grad_norm": 2.2421772480010986,
245
+ "learning_rate": 2.6565656565656566e-05,
246
+ "loss": 7.7613,
247
+ "step": 34
248
+ },
249
+ {
250
+ "epoch": 0.35353535353535354,
251
+ "grad_norm": 2.111851930618286,
252
+ "learning_rate": 2.6464646464646466e-05,
253
+ "loss": 7.7516,
254
+ "step": 35
255
+ },
256
+ {
257
+ "epoch": 0.36363636363636365,
258
+ "grad_norm": 2.2756099700927734,
259
+ "learning_rate": 2.6363636363636365e-05,
260
+ "loss": 7.6066,
261
+ "step": 36
262
+ },
263
+ {
264
+ "epoch": 0.37373737373737376,
265
+ "grad_norm": 1.5493675470352173,
266
+ "learning_rate": 2.6262626262626265e-05,
267
+ "loss": 7.6497,
268
+ "step": 37
269
+ },
270
+ {
271
+ "epoch": 0.3838383838383838,
272
+ "grad_norm": 1.7187950611114502,
273
+ "learning_rate": 2.6161616161616164e-05,
274
+ "loss": 7.5753,
275
+ "step": 38
276
+ },
277
+ {
278
+ "epoch": 0.3939393939393939,
279
+ "grad_norm": 1.5378119945526123,
280
+ "learning_rate": 2.6060606060606063e-05,
281
+ "loss": 7.591,
282
+ "step": 39
283
+ },
284
+ {
285
+ "epoch": 0.40404040404040403,
286
+ "grad_norm": 1.3718913793563843,
287
+ "learning_rate": 2.595959595959596e-05,
288
+ "loss": 7.6196,
289
+ "step": 40
290
+ },
291
+ {
292
+ "epoch": 0.41414141414141414,
293
+ "grad_norm": 1.9546277523040771,
294
+ "learning_rate": 2.585858585858586e-05,
295
+ "loss": 7.3466,
296
+ "step": 41
297
+ },
298
+ {
299
+ "epoch": 0.42424242424242425,
300
+ "grad_norm": 1.3100489377975464,
301
+ "learning_rate": 2.575757575757576e-05,
302
+ "loss": 7.588,
303
+ "step": 42
304
+ },
305
+ {
306
+ "epoch": 0.43434343434343436,
307
+ "grad_norm": 1.1814428567886353,
308
+ "learning_rate": 2.5656565656565658e-05,
309
+ "loss": 7.392,
310
+ "step": 43
311
+ },
312
+ {
313
+ "epoch": 0.4444444444444444,
314
+ "grad_norm": 1.3818390369415283,
315
+ "learning_rate": 2.5555555555555557e-05,
316
+ "loss": 7.413,
317
+ "step": 44
318
+ },
319
+ {
320
+ "epoch": 0.45454545454545453,
321
+ "grad_norm": 1.725190281867981,
322
+ "learning_rate": 2.5454545454545457e-05,
323
+ "loss": 7.1846,
324
+ "step": 45
325
+ },
326
+ {
327
+ "epoch": 0.46464646464646464,
328
+ "grad_norm": 1.098010778427124,
329
+ "learning_rate": 2.5353535353535356e-05,
330
+ "loss": 7.384,
331
+ "step": 46
332
+ },
333
+ {
334
+ "epoch": 0.47474747474747475,
335
+ "grad_norm": 1.1586623191833496,
336
+ "learning_rate": 2.5252525252525256e-05,
337
+ "loss": 7.2481,
338
+ "step": 47
339
+ },
340
+ {
341
+ "epoch": 0.48484848484848486,
342
+ "grad_norm": 1.3093281984329224,
343
+ "learning_rate": 2.5151515151515152e-05,
344
+ "loss": 7.1453,
345
+ "step": 48
346
+ },
347
+ {
348
+ "epoch": 0.494949494949495,
349
+ "grad_norm": 1.2143803834915161,
350
+ "learning_rate": 2.505050505050505e-05,
351
+ "loss": 7.1706,
352
+ "step": 49
353
+ },
354
+ {
355
+ "epoch": 0.5050505050505051,
356
+ "grad_norm": 1.4759756326675415,
357
+ "learning_rate": 2.494949494949495e-05,
358
+ "loss": 7.1918,
359
+ "step": 50
360
+ },
361
+ {
362
+ "epoch": 0.5151515151515151,
363
+ "grad_norm": 1.170322060585022,
364
+ "learning_rate": 2.484848484848485e-05,
365
+ "loss": 7.1158,
366
+ "step": 51
367
+ },
368
+ {
369
+ "epoch": 0.5252525252525253,
370
+ "grad_norm": 1.025293231010437,
371
+ "learning_rate": 2.474747474747475e-05,
372
+ "loss": 7.085,
373
+ "step": 52
374
+ },
375
+ {
376
+ "epoch": 0.5353535353535354,
377
+ "grad_norm": 1.673163890838623,
378
+ "learning_rate": 2.464646464646465e-05,
379
+ "loss": 7.4093,
380
+ "step": 53
381
+ },
382
+ {
383
+ "epoch": 0.5454545454545454,
384
+ "grad_norm": 0.892270565032959,
385
+ "learning_rate": 2.454545454545455e-05,
386
+ "loss": 7.1903,
387
+ "step": 54
388
+ },
389
+ {
390
+ "epoch": 0.5555555555555556,
391
+ "grad_norm": 1.1452521085739136,
392
+ "learning_rate": 2.4444444444444445e-05,
393
+ "loss": 7.2127,
394
+ "step": 55
395
+ },
396
+ {
397
+ "epoch": 0.5656565656565656,
398
+ "grad_norm": 1.4096879959106445,
399
+ "learning_rate": 2.4343434343434344e-05,
400
+ "loss": 7.2507,
401
+ "step": 56
402
+ },
403
+ {
404
+ "epoch": 0.5757575757575758,
405
+ "grad_norm": 1.3830610513687134,
406
+ "learning_rate": 2.4242424242424244e-05,
407
+ "loss": 7.2779,
408
+ "step": 57
409
+ },
410
+ {
411
+ "epoch": 0.5858585858585859,
412
+ "grad_norm": 0.88773512840271,
413
+ "learning_rate": 2.4141414141414143e-05,
414
+ "loss": 7.1313,
415
+ "step": 58
416
+ },
417
+ {
418
+ "epoch": 0.5959595959595959,
419
+ "grad_norm": 1.5023548603057861,
420
+ "learning_rate": 2.4040404040404043e-05,
421
+ "loss": 7.3029,
422
+ "step": 59
423
+ },
424
+ {
425
+ "epoch": 0.6060606060606061,
426
+ "grad_norm": 0.9497267603874207,
427
+ "learning_rate": 2.3939393939393942e-05,
428
+ "loss": 7.1915,
429
+ "step": 60
430
+ },
431
+ {
432
+ "epoch": 0.6161616161616161,
433
+ "grad_norm": 1.035706639289856,
434
+ "learning_rate": 2.3838383838383842e-05,
435
+ "loss": 7.2144,
436
+ "step": 61
437
+ },
438
+ {
439
+ "epoch": 0.6262626262626263,
440
+ "grad_norm": 0.7967200875282288,
441
+ "learning_rate": 2.3737373737373738e-05,
442
+ "loss": 7.1566,
443
+ "step": 62
444
+ },
445
+ {
446
+ "epoch": 0.6363636363636364,
447
+ "grad_norm": 0.7237263917922974,
448
+ "learning_rate": 2.3636363636363637e-05,
449
+ "loss": 7.1931,
450
+ "step": 63
451
+ },
452
+ {
453
+ "epoch": 0.6464646464646465,
454
+ "grad_norm": 1.0147182941436768,
455
+ "learning_rate": 2.3535353535353537e-05,
456
+ "loss": 7.2676,
457
+ "step": 64
458
+ },
459
+ {
460
+ "epoch": 0.6565656565656566,
461
+ "grad_norm": 0.8428776264190674,
462
+ "learning_rate": 2.3434343434343436e-05,
463
+ "loss": 7.0818,
464
+ "step": 65
465
+ },
466
+ {
467
+ "epoch": 0.6666666666666666,
468
+ "grad_norm": 0.9503577947616577,
469
+ "learning_rate": 2.3333333333333336e-05,
470
+ "loss": 7.0711,
471
+ "step": 66
472
+ },
473
+ {
474
+ "epoch": 0.6767676767676768,
475
+ "grad_norm": 0.8867577314376831,
476
+ "learning_rate": 2.3232323232323235e-05,
477
+ "loss": 7.0556,
478
+ "step": 67
479
+ },
480
+ {
481
+ "epoch": 0.6868686868686869,
482
+ "grad_norm": 0.8943458795547485,
483
+ "learning_rate": 2.3131313131313135e-05,
484
+ "loss": 7.2772,
485
+ "step": 68
486
+ },
487
+ {
488
+ "epoch": 0.696969696969697,
489
+ "grad_norm": 0.8544779419898987,
490
+ "learning_rate": 2.303030303030303e-05,
491
+ "loss": 7.0846,
492
+ "step": 69
493
+ },
494
+ {
495
+ "epoch": 0.7070707070707071,
496
+ "grad_norm": 0.8547456860542297,
497
+ "learning_rate": 2.292929292929293e-05,
498
+ "loss": 7.092,
499
+ "step": 70
500
+ },
501
+ {
502
+ "epoch": 0.7171717171717171,
503
+ "grad_norm": 0.924784779548645,
504
+ "learning_rate": 2.282828282828283e-05,
505
+ "loss": 7.0622,
506
+ "step": 71
507
+ },
508
+ {
509
+ "epoch": 0.7272727272727273,
510
+ "grad_norm": 1.1316595077514648,
511
+ "learning_rate": 2.272727272727273e-05,
512
+ "loss": 7.0545,
513
+ "step": 72
514
+ },
515
+ {
516
+ "epoch": 0.7373737373737373,
517
+ "grad_norm": NaN,
518
+ "learning_rate": 2.272727272727273e-05,
519
+ "loss": 0.0,
520
+ "step": 73
521
+ },
522
+ {
523
+ "epoch": 0.7474747474747475,
524
+ "grad_norm": 0.7388734221458435,
525
+ "learning_rate": 2.262626262626263e-05,
526
+ "loss": 7.0892,
527
+ "step": 74
528
+ },
529
+ {
530
+ "epoch": 0.7575757575757576,
531
+ "grad_norm": 1.9678359031677246,
532
+ "learning_rate": 2.2525252525252525e-05,
533
+ "loss": 7.0092,
534
+ "step": 75
535
+ },
536
+ {
537
+ "epoch": 0.7676767676767676,
538
+ "grad_norm": 0.736301064491272,
539
+ "learning_rate": 2.2424242424242424e-05,
540
+ "loss": 7.1827,
541
+ "step": 76
542
+ },
543
+ {
544
+ "epoch": 0.7777777777777778,
545
+ "grad_norm": 0.7673182487487793,
546
+ "learning_rate": 2.2323232323232324e-05,
547
+ "loss": 7.0768,
548
+ "step": 77
549
+ },
550
+ {
551
+ "epoch": 0.7878787878787878,
552
+ "grad_norm": 1.4001637697219849,
553
+ "learning_rate": 2.222222222222222e-05,
554
+ "loss": 6.9754,
555
+ "step": 78
556
+ },
557
+ {
558
+ "epoch": 0.797979797979798,
559
+ "grad_norm": 0.8279094696044922,
560
+ "learning_rate": 2.212121212121212e-05,
561
+ "loss": 7.1222,
562
+ "step": 79
563
+ },
564
+ {
565
+ "epoch": 0.8080808080808081,
566
+ "grad_norm": 0.6914481520652771,
567
+ "learning_rate": 2.202020202020202e-05,
568
+ "loss": 7.0812,
569
+ "step": 80
570
+ },
571
+ {
572
+ "epoch": 0.8181818181818182,
573
+ "grad_norm": 0.7313191294670105,
574
+ "learning_rate": 2.191919191919192e-05,
575
+ "loss": 7.05,
576
+ "step": 81
577
+ },
578
+ {
579
+ "epoch": 0.8282828282828283,
580
+ "grad_norm": 1.4300901889801025,
581
+ "learning_rate": 2.1818181818181818e-05,
582
+ "loss": 7.2278,
583
+ "step": 82
584
+ },
585
+ {
586
+ "epoch": 0.8383838383838383,
587
+ "grad_norm": 1.2713080644607544,
588
+ "learning_rate": 2.1717171717171717e-05,
589
+ "loss": 6.9029,
590
+ "step": 83
591
+ },
592
+ {
593
+ "epoch": 0.8484848484848485,
594
+ "grad_norm": 0.814030110836029,
595
+ "learning_rate": 2.1616161616161617e-05,
596
+ "loss": 7.1241,
597
+ "step": 84
598
+ },
599
+ {
600
+ "epoch": 0.8585858585858586,
601
+ "grad_norm": 1.1318764686584473,
602
+ "learning_rate": 2.1515151515151513e-05,
603
+ "loss": 7.1621,
604
+ "step": 85
605
+ },
606
+ {
607
+ "epoch": 0.8686868686868687,
608
+ "grad_norm": 1.0205698013305664,
609
+ "learning_rate": 2.1414141414141412e-05,
610
+ "loss": 7.0894,
611
+ "step": 86
612
+ },
613
+ {
614
+ "epoch": 0.8787878787878788,
615
+ "grad_norm": 0.5895187854766846,
616
+ "learning_rate": 2.1313131313131312e-05,
617
+ "loss": 6.9731,
618
+ "step": 87
619
+ },
620
+ {
621
+ "epoch": 0.8888888888888888,
622
+ "grad_norm": 0.5949335098266602,
623
+ "learning_rate": 2.121212121212121e-05,
624
+ "loss": 6.9764,
625
+ "step": 88
626
+ },
627
+ {
628
+ "epoch": 0.898989898989899,
629
+ "grad_norm": 0.909782886505127,
630
+ "learning_rate": 2.111111111111111e-05,
631
+ "loss": 6.8954,
632
+ "step": 89
633
+ },
634
+ {
635
+ "epoch": 0.9090909090909091,
636
+ "grad_norm": 0.6604152321815491,
637
+ "learning_rate": 2.101010101010101e-05,
638
+ "loss": 7.015,
639
+ "step": 90
640
+ },
641
+ {
642
+ "epoch": 0.9191919191919192,
643
+ "grad_norm": 1.1579092741012573,
644
+ "learning_rate": 2.090909090909091e-05,
645
+ "loss": 7.1838,
646
+ "step": 91
647
+ },
648
+ {
649
+ "epoch": 0.9292929292929293,
650
+ "grad_norm": 1.1899914741516113,
651
+ "learning_rate": 2.0808080808080806e-05,
652
+ "loss": 7.196,
653
+ "step": 92
654
+ },
655
+ {
656
+ "epoch": 0.9393939393939394,
657
+ "grad_norm": 1.013912558555603,
658
+ "learning_rate": 2.0707070707070705e-05,
659
+ "loss": 7.1402,
660
+ "step": 93
661
+ },
662
+ {
663
+ "epoch": 0.9494949494949495,
664
+ "grad_norm": 0.907071053981781,
665
+ "learning_rate": 2.0606060606060605e-05,
666
+ "loss": 7.129,
667
+ "step": 94
668
+ },
669
+ {
670
+ "epoch": 0.9595959595959596,
671
+ "grad_norm": 0.7546154260635376,
672
+ "learning_rate": 2.0505050505050504e-05,
673
+ "loss": 7.0195,
674
+ "step": 95
675
+ },
676
+ {
677
+ "epoch": 0.9696969696969697,
678
+ "grad_norm": 0.662983775138855,
679
+ "learning_rate": 2.0404040404040404e-05,
680
+ "loss": 7.0075,
681
+ "step": 96
682
+ },
683
+ {
684
+ "epoch": 0.9797979797979798,
685
+ "grad_norm": 0.673218846321106,
686
+ "learning_rate": 2.0303030303030303e-05,
687
+ "loss": 7.0797,
688
+ "step": 97
689
+ },
690
+ {
691
+ "epoch": 0.98989898989899,
692
+ "grad_norm": 1.617126703262329,
693
+ "learning_rate": 2.0202020202020203e-05,
694
+ "loss": 6.8117,
695
+ "step": 98
696
+ },
697
+ {
698
+ "epoch": 1.0,
699
+ "grad_norm": 1.0646132230758667,
700
+ "learning_rate": 2.01010101010101e-05,
701
+ "loss": 7.174,
702
+ "step": 99
703
+ },
704
+ {
705
+ "epoch": 1.0,
706
+ "eval_audio_cosine_sim": 0.383101224899292,
707
+ "eval_loss": 4.921389102935791,
708
+ "eval_runtime": 153.7637,
709
+ "eval_samples_per_second": 0.059,
710
+ "eval_steps_per_second": 0.02,
711
+ "eval_text_cosine_sim": 0.16359001398086548,
712
+ "step": 99
713
+ },
714
+ {
715
+ "epoch": 1.0101010101010102,
716
+ "grad_norm": 0.7128923535346985,
717
+ "learning_rate": 1.9999999999999998e-05,
718
+ "loss": 7.0139,
719
+ "step": 100
720
+ },
721
+ {
722
+ "epoch": 1.02020202020202,
723
+ "grad_norm": 0.8200396299362183,
724
+ "learning_rate": 1.9898989898989898e-05,
725
+ "loss": 7.2107,
726
+ "step": 101
727
+ },
728
+ {
729
+ "epoch": 1.0303030303030303,
730
+ "grad_norm": 0.6839468479156494,
731
+ "learning_rate": 1.9797979797979797e-05,
732
+ "loss": 7.1309,
733
+ "step": 102
734
+ },
735
+ {
736
+ "epoch": 1.0404040404040404,
737
+ "grad_norm": 1.4797368049621582,
738
+ "learning_rate": 1.9696969696969697e-05,
739
+ "loss": 6.9112,
740
+ "step": 103
741
+ },
742
+ {
743
+ "epoch": 1.0505050505050506,
744
+ "grad_norm": 0.7093661427497864,
745
+ "learning_rate": 1.9595959595959596e-05,
746
+ "loss": 7.1014,
747
+ "step": 104
748
+ },
749
+ {
750
+ "epoch": 1.0606060606060606,
751
+ "grad_norm": 1.2138644456863403,
752
+ "learning_rate": 1.9494949494949496e-05,
753
+ "loss": 6.9265,
754
+ "step": 105
755
+ },
756
+ {
757
+ "epoch": 1.0707070707070707,
758
+ "grad_norm": 0.8460476994514465,
759
+ "learning_rate": 1.9393939393939395e-05,
760
+ "loss": 7.2191,
761
+ "step": 106
762
+ },
763
+ {
764
+ "epoch": 1.0808080808080809,
765
+ "grad_norm": 0.7066294550895691,
766
+ "learning_rate": 1.929292929292929e-05,
767
+ "loss": 7.1066,
768
+ "step": 107
769
+ },
770
+ {
771
+ "epoch": 1.0909090909090908,
772
+ "grad_norm": 1.7321295738220215,
773
+ "learning_rate": 1.919191919191919e-05,
774
+ "loss": 6.8371,
775
+ "step": 108
776
+ },
777
+ {
778
+ "epoch": 1.101010101010101,
779
+ "grad_norm": 0.9022343158721924,
780
+ "learning_rate": 1.909090909090909e-05,
781
+ "loss": 6.9655,
782
+ "step": 109
783
+ },
784
+ {
785
+ "epoch": 1.1111111111111112,
786
+ "grad_norm": 0.6070654988288879,
787
+ "learning_rate": 1.898989898989899e-05,
788
+ "loss": 6.9956,
789
+ "step": 110
790
+ },
791
+ {
792
+ "epoch": 1.121212121212121,
793
+ "grad_norm": 0.6426708698272705,
794
+ "learning_rate": 1.888888888888889e-05,
795
+ "loss": 6.964,
796
+ "step": 111
797
+ },
798
+ {
799
+ "epoch": 1.1313131313131313,
800
+ "grad_norm": 2.1155552864074707,
801
+ "learning_rate": 1.878787878787879e-05,
802
+ "loss": 6.7698,
803
+ "step": 112
804
+ },
805
+ {
806
+ "epoch": 1.1414141414141414,
807
+ "grad_norm": 0.9715363383293152,
808
+ "learning_rate": 1.8686868686868688e-05,
809
+ "loss": 7.1096,
810
+ "step": 113
811
+ },
812
+ {
813
+ "epoch": 1.1515151515151516,
814
+ "grad_norm": 1.2896918058395386,
815
+ "learning_rate": 1.8585858585858584e-05,
816
+ "loss": 6.7545,
817
+ "step": 114
818
+ },
819
+ {
820
+ "epoch": 1.1616161616161615,
821
+ "grad_norm": 0.6836603283882141,
822
+ "learning_rate": 1.8484848484848484e-05,
823
+ "loss": 7.042,
824
+ "step": 115
825
+ },
826
+ {
827
+ "epoch": 1.1717171717171717,
828
+ "grad_norm": 0.6922496557235718,
829
+ "learning_rate": 1.8383838383838383e-05,
830
+ "loss": 6.9593,
831
+ "step": 116
832
+ },
833
+ {
834
+ "epoch": 1.1818181818181819,
835
+ "grad_norm": 1.5327054262161255,
836
+ "learning_rate": 1.8282828282828283e-05,
837
+ "loss": 6.7356,
838
+ "step": 117
839
+ },
840
+ {
841
+ "epoch": 1.1919191919191918,
842
+ "grad_norm": 0.686898410320282,
843
+ "learning_rate": 1.8181818181818182e-05,
844
+ "loss": 6.8844,
845
+ "step": 118
846
+ },
847
+ {
848
+ "epoch": 1.202020202020202,
849
+ "grad_norm": 2.4868946075439453,
850
+ "learning_rate": 1.808080808080808e-05,
851
+ "loss": 7.2632,
852
+ "step": 119
853
+ },
854
+ {
855
+ "epoch": 1.2121212121212122,
856
+ "grad_norm": 0.705142080783844,
857
+ "learning_rate": 1.797979797979798e-05,
858
+ "loss": 6.9367,
859
+ "step": 120
860
+ },
861
+ {
862
+ "epoch": 1.2222222222222223,
863
+ "grad_norm": 0.6701662540435791,
864
+ "learning_rate": 1.7878787878787877e-05,
865
+ "loss": 6.8656,
866
+ "step": 121
867
+ },
868
+ {
869
+ "epoch": 1.2323232323232323,
870
+ "grad_norm": 2.3600728511810303,
871
+ "learning_rate": 1.7777777777777777e-05,
872
+ "loss": 7.2908,
873
+ "step": 122
874
+ },
875
+ {
876
+ "epoch": 1.2424242424242424,
877
+ "grad_norm": 0.729280948638916,
878
+ "learning_rate": 1.7676767676767676e-05,
879
+ "loss": 6.9778,
880
+ "step": 123
881
+ },
882
+ {
883
+ "epoch": 1.2525252525252526,
884
+ "grad_norm": 0.9563952684402466,
885
+ "learning_rate": 1.7575757575757576e-05,
886
+ "loss": 6.8367,
887
+ "step": 124
888
+ },
889
+ {
890
+ "epoch": 1.2626262626262625,
891
+ "grad_norm": 0.679764449596405,
892
+ "learning_rate": 1.7474747474747475e-05,
893
+ "loss": 6.9984,
894
+ "step": 125
895
+ },
896
+ {
897
+ "epoch": 1.2727272727272727,
898
+ "grad_norm": 1.1830710172653198,
899
+ "learning_rate": 1.7373737373737375e-05,
900
+ "loss": 7.0285,
901
+ "step": 126
902
+ },
903
+ {
904
+ "epoch": 1.2828282828282829,
905
+ "grad_norm": 0.7950251698493958,
906
+ "learning_rate": 1.7272727272727274e-05,
907
+ "loss": 7.0085,
908
+ "step": 127
909
+ },
910
+ {
911
+ "epoch": 1.2929292929292928,
912
+ "grad_norm": 0.8154851794242859,
913
+ "learning_rate": 1.717171717171717e-05,
914
+ "loss": 6.8398,
915
+ "step": 128
916
+ },
917
+ {
918
+ "epoch": 1.303030303030303,
919
+ "grad_norm": 1.1023356914520264,
920
+ "learning_rate": 1.707070707070707e-05,
921
+ "loss": 7.1022,
922
+ "step": 129
923
+ },
924
+ {
925
+ "epoch": 1.3131313131313131,
926
+ "grad_norm": 1.013553500175476,
927
+ "learning_rate": 1.696969696969697e-05,
928
+ "loss": 6.8513,
929
+ "step": 130
930
+ },
931
+ {
932
+ "epoch": 1.3232323232323233,
933
+ "grad_norm": 1.4370501041412354,
934
+ "learning_rate": 1.686868686868687e-05,
935
+ "loss": 7.208,
936
+ "step": 131
937
+ },
938
+ {
939
+ "epoch": 1.3333333333333333,
940
+ "grad_norm": 0.7258216738700867,
941
+ "learning_rate": 1.6767676767676768e-05,
942
+ "loss": 6.954,
943
+ "step": 132
944
+ },
945
+ {
946
+ "epoch": 1.3434343434343434,
947
+ "grad_norm": 0.7258408069610596,
948
+ "learning_rate": 1.6666666666666667e-05,
949
+ "loss": 7.0258,
950
+ "step": 133
951
+ },
952
+ {
953
+ "epoch": 1.3535353535353536,
954
+ "grad_norm": 0.8000563383102417,
955
+ "learning_rate": 1.6565656565656567e-05,
956
+ "loss": 7.0238,
957
+ "step": 134
958
+ },
959
+ {
960
+ "epoch": 1.3636363636363638,
961
+ "grad_norm": 0.9891355633735657,
962
+ "learning_rate": 1.6464646464646466e-05,
963
+ "loss": 6.9668,
964
+ "step": 135
965
+ },
966
+ {
967
+ "epoch": 1.3737373737373737,
968
+ "grad_norm": 0.7241990566253662,
969
+ "learning_rate": 1.6363636363636363e-05,
970
+ "loss": 7.0859,
971
+ "step": 136
972
+ },
973
+ {
974
+ "epoch": 1.3838383838383839,
975
+ "grad_norm": 0.5333958268165588,
976
+ "learning_rate": 1.6262626262626262e-05,
977
+ "loss": 7.0339,
978
+ "step": 137
979
+ },
980
+ {
981
+ "epoch": 1.393939393939394,
982
+ "grad_norm": 0.7480002641677856,
983
+ "learning_rate": 1.616161616161616e-05,
984
+ "loss": 6.9745,
985
+ "step": 138
986
+ },
987
+ {
988
+ "epoch": 1.404040404040404,
989
+ "grad_norm": 0.8260339498519897,
990
+ "learning_rate": 1.606060606060606e-05,
991
+ "loss": 7.0461,
992
+ "step": 139
993
+ },
994
+ {
995
+ "epoch": 1.4141414141414141,
996
+ "grad_norm": 0.7007152438163757,
997
+ "learning_rate": 1.595959595959596e-05,
998
+ "loss": 7.046,
999
+ "step": 140
1000
+ },
1001
+ {
1002
+ "epoch": 1.4242424242424243,
1003
+ "grad_norm": 1.532183051109314,
1004
+ "learning_rate": 1.585858585858586e-05,
1005
+ "loss": 6.8063,
1006
+ "step": 141
1007
+ },
1008
+ {
1009
+ "epoch": 1.4343434343434343,
1010
+ "grad_norm": 0.6525304317474365,
1011
+ "learning_rate": 1.575757575757576e-05,
1012
+ "loss": 6.9673,
1013
+ "step": 142
1014
+ },
1015
+ {
1016
+ "epoch": 1.4444444444444444,
1017
+ "grad_norm": 0.875150203704834,
1018
+ "learning_rate": 1.5656565656565655e-05,
1019
+ "loss": 6.9106,
1020
+ "step": 143
1021
+ },
1022
+ {
1023
+ "epoch": 1.4545454545454546,
1024
+ "grad_norm": 0.7586594223976135,
1025
+ "learning_rate": 1.5555555555555555e-05,
1026
+ "loss": 6.9954,
1027
+ "step": 144
1028
+ },
1029
+ {
1030
+ "epoch": 1.4646464646464645,
1031
+ "grad_norm": 1.1310893297195435,
1032
+ "learning_rate": 1.5454545454545454e-05,
1033
+ "loss": 7.1989,
1034
+ "step": 145
1035
+ },
1036
+ {
1037
+ "epoch": 1.4747474747474747,
1038
+ "grad_norm": 1.3009579181671143,
1039
+ "learning_rate": 1.5353535353535354e-05,
1040
+ "loss": 7.1353,
1041
+ "step": 146
1042
+ },
1043
+ {
1044
+ "epoch": 1.4848484848484849,
1045
+ "grad_norm": 0.6209822297096252,
1046
+ "learning_rate": 1.5252525252525253e-05,
1047
+ "loss": 6.9722,
1048
+ "step": 147
1049
+ },
1050
+ {
1051
+ "epoch": 1.494949494949495,
1052
+ "grad_norm": 0.7535139322280884,
1053
+ "learning_rate": 1.5151515151515153e-05,
1054
+ "loss": 7.0278,
1055
+ "step": 148
1056
+ },
1057
+ {
1058
+ "epoch": 1.5050505050505052,
1059
+ "grad_norm": 0.6947258114814758,
1060
+ "learning_rate": 1.5050505050505052e-05,
1061
+ "loss": 7.0663,
1062
+ "step": 149
1063
+ },
1064
+ {
1065
+ "epoch": 1.5151515151515151,
1066
+ "grad_norm": 0.7372407913208008,
1067
+ "learning_rate": 1.494949494949495e-05,
1068
+ "loss": 6.9891,
1069
+ "step": 150
1070
+ },
1071
+ {
1072
+ "epoch": 1.5252525252525253,
1073
+ "grad_norm": 0.8788199424743652,
1074
+ "learning_rate": 1.484848484848485e-05,
1075
+ "loss": 6.9849,
1076
+ "step": 151
1077
+ },
1078
+ {
1079
+ "epoch": 1.5353535353535355,
1080
+ "grad_norm": 0.9445679187774658,
1081
+ "learning_rate": 1.4747474747474747e-05,
1082
+ "loss": 6.976,
1083
+ "step": 152
1084
+ },
1085
+ {
1086
+ "epoch": 1.5454545454545454,
1087
+ "grad_norm": 1.7923697233200073,
1088
+ "learning_rate": 1.4646464646464647e-05,
1089
+ "loss": 6.7889,
1090
+ "step": 153
1091
+ },
1092
+ {
1093
+ "epoch": 1.5555555555555556,
1094
+ "grad_norm": 0.6855000257492065,
1095
+ "learning_rate": 1.4545454545454546e-05,
1096
+ "loss": 6.9884,
1097
+ "step": 154
1098
+ },
1099
+ {
1100
+ "epoch": 1.5656565656565657,
1101
+ "grad_norm": 0.8418505191802979,
1102
+ "learning_rate": 1.4444444444444444e-05,
1103
+ "loss": 6.9349,
1104
+ "step": 155
1105
+ },
1106
+ {
1107
+ "epoch": 1.5757575757575757,
1108
+ "grad_norm": 0.7767158150672913,
1109
+ "learning_rate": 1.4343434343434344e-05,
1110
+ "loss": 6.9518,
1111
+ "step": 156
1112
+ },
1113
+ {
1114
+ "epoch": 1.5858585858585859,
1115
+ "grad_norm": 0.9679882526397705,
1116
+ "learning_rate": 1.4242424242424243e-05,
1117
+ "loss": 6.8211,
1118
+ "step": 157
1119
+ },
1120
+ {
1121
+ "epoch": 1.595959595959596,
1122
+ "grad_norm": 0.9250569343566895,
1123
+ "learning_rate": 1.4141414141414143e-05,
1124
+ "loss": 6.9044,
1125
+ "step": 158
1126
+ },
1127
+ {
1128
+ "epoch": 1.606060606060606,
1129
+ "grad_norm": 0.8403336405754089,
1130
+ "learning_rate": 1.404040404040404e-05,
1131
+ "loss": 6.8393,
1132
+ "step": 159
1133
+ },
1134
+ {
1135
+ "epoch": 1.6161616161616161,
1136
+ "grad_norm": 0.9166443347930908,
1137
+ "learning_rate": 1.393939393939394e-05,
1138
+ "loss": 6.8174,
1139
+ "step": 160
1140
+ },
1141
+ {
1142
+ "epoch": 1.6262626262626263,
1143
+ "grad_norm": 0.5435317754745483,
1144
+ "learning_rate": 1.383838383838384e-05,
1145
+ "loss": 6.9202,
1146
+ "step": 161
1147
+ },
1148
+ {
1149
+ "epoch": 1.6363636363636362,
1150
+ "grad_norm": 1.0361636877059937,
1151
+ "learning_rate": 1.3737373737373739e-05,
1152
+ "loss": 6.9836,
1153
+ "step": 162
1154
+ },
1155
+ {
1156
+ "epoch": 1.6464646464646466,
1157
+ "grad_norm": 1.709415316581726,
1158
+ "learning_rate": 1.3636363636363637e-05,
1159
+ "loss": 7.1939,
1160
+ "step": 163
1161
+ },
1162
+ {
1163
+ "epoch": 1.6565656565656566,
1164
+ "grad_norm": 1.3511853218078613,
1165
+ "learning_rate": 1.3535353535353536e-05,
1166
+ "loss": 7.0895,
1167
+ "step": 164
1168
+ },
1169
+ {
1170
+ "epoch": 1.6666666666666665,
1171
+ "grad_norm": 0.6736574172973633,
1172
+ "learning_rate": 1.3434343434343436e-05,
1173
+ "loss": 6.8909,
1174
+ "step": 165
1175
+ },
1176
+ {
1177
+ "epoch": 1.676767676767677,
1178
+ "grad_norm": 0.6862733364105225,
1179
+ "learning_rate": 1.3333333333333333e-05,
1180
+ "loss": 6.9872,
1181
+ "step": 166
1182
+ },
1183
+ {
1184
+ "epoch": 1.6868686868686869,
1185
+ "grad_norm": 1.0613605976104736,
1186
+ "learning_rate": 1.3232323232323233e-05,
1187
+ "loss": 6.9787,
1188
+ "step": 167
1189
+ },
1190
+ {
1191
+ "epoch": 1.696969696969697,
1192
+ "grad_norm": 0.9397658705711365,
1193
+ "learning_rate": 1.3131313131313132e-05,
1194
+ "loss": 6.85,
1195
+ "step": 168
1196
+ },
1197
+ {
1198
+ "epoch": 1.7070707070707072,
1199
+ "grad_norm": 0.6116166114807129,
1200
+ "learning_rate": 1.3030303030303032e-05,
1201
+ "loss": 6.9257,
1202
+ "step": 169
1203
+ },
1204
+ {
1205
+ "epoch": 1.7171717171717171,
1206
+ "grad_norm": 1.0234121084213257,
1207
+ "learning_rate": 1.292929292929293e-05,
1208
+ "loss": 7.0316,
1209
+ "step": 170
1210
+ },
1211
+ {
1212
+ "epoch": 1.7272727272727273,
1213
+ "grad_norm": 1.68846595287323,
1214
+ "learning_rate": 1.2828282828282829e-05,
1215
+ "loss": 7.1854,
1216
+ "step": 171
1217
+ },
1218
+ {
1219
+ "epoch": 1.7373737373737375,
1220
+ "grad_norm": 0.686038076877594,
1221
+ "learning_rate": 1.2727272727272728e-05,
1222
+ "loss": 6.9311,
1223
+ "step": 172
1224
+ },
1225
+ {
1226
+ "epoch": 1.7474747474747474,
1227
+ "grad_norm": 1.3312983512878418,
1228
+ "learning_rate": 1.2626262626262628e-05,
1229
+ "loss": 7.1633,
1230
+ "step": 173
1231
+ },
1232
+ {
1233
+ "epoch": 1.7575757575757576,
1234
+ "grad_norm": 0.688218355178833,
1235
+ "learning_rate": 1.2525252525252526e-05,
1236
+ "loss": 7.0336,
1237
+ "step": 174
1238
+ },
1239
+ {
1240
+ "epoch": 1.7676767676767677,
1241
+ "grad_norm": 0.768675684928894,
1242
+ "learning_rate": 1.2424242424242425e-05,
1243
+ "loss": 6.9246,
1244
+ "step": 175
1245
+ },
1246
+ {
1247
+ "epoch": 1.7777777777777777,
1248
+ "grad_norm": 0.6171760559082031,
1249
+ "learning_rate": 1.2323232323232325e-05,
1250
+ "loss": 7.0127,
1251
+ "step": 176
1252
+ },
1253
+ {
1254
+ "epoch": 1.7878787878787878,
1255
+ "grad_norm": 0.79632967710495,
1256
+ "learning_rate": 1.2222222222222222e-05,
1257
+ "loss": 7.0955,
1258
+ "step": 177
1259
+ },
1260
+ {
1261
+ "epoch": 1.797979797979798,
1262
+ "grad_norm": 0.5743001699447632,
1263
+ "learning_rate": 1.2121212121212122e-05,
1264
+ "loss": 7.0305,
1265
+ "step": 178
1266
+ },
1267
+ {
1268
+ "epoch": 1.808080808080808,
1269
+ "grad_norm": 0.648186445236206,
1270
+ "learning_rate": 1.2020202020202021e-05,
1271
+ "loss": 7.0791,
1272
+ "step": 179
1273
+ },
1274
+ {
1275
+ "epoch": 1.8181818181818183,
1276
+ "grad_norm": 0.9388221502304077,
1277
+ "learning_rate": 1.1919191919191921e-05,
1278
+ "loss": 6.8963,
1279
+ "step": 180
1280
+ },
1281
+ {
1282
+ "epoch": 1.8282828282828283,
1283
+ "grad_norm": 0.6248198747634888,
1284
+ "learning_rate": 1.1818181818181819e-05,
1285
+ "loss": 7.1456,
1286
+ "step": 181
1287
+ },
1288
+ {
1289
+ "epoch": 1.8383838383838382,
1290
+ "grad_norm": 0.7853441834449768,
1291
+ "learning_rate": 1.1717171717171718e-05,
1292
+ "loss": 7.1302,
1293
+ "step": 182
1294
+ },
1295
+ {
1296
+ "epoch": 1.8484848484848486,
1297
+ "grad_norm": 0.7538427710533142,
1298
+ "learning_rate": 1.1616161616161618e-05,
1299
+ "loss": 7.0855,
1300
+ "step": 183
1301
+ },
1302
+ {
1303
+ "epoch": 1.8585858585858586,
1304
+ "grad_norm": 1.3595892190933228,
1305
+ "learning_rate": 1.1515151515151515e-05,
1306
+ "loss": 6.8817,
1307
+ "step": 184
1308
+ },
1309
+ {
1310
+ "epoch": 1.8686868686868687,
1311
+ "grad_norm": 1.2711068391799927,
1312
+ "learning_rate": 1.1414141414141415e-05,
1313
+ "loss": 6.9406,
1314
+ "step": 185
1315
+ },
1316
+ {
1317
+ "epoch": 1.878787878787879,
1318
+ "grad_norm": 0.7485206127166748,
1319
+ "learning_rate": 1.1313131313131314e-05,
1320
+ "loss": 7.0214,
1321
+ "step": 186
1322
+ },
1323
+ {
1324
+ "epoch": 1.8888888888888888,
1325
+ "grad_norm": 1.068521499633789,
1326
+ "learning_rate": 1.1212121212121212e-05,
1327
+ "loss": 6.9544,
1328
+ "step": 187
1329
+ },
1330
+ {
1331
+ "epoch": 1.898989898989899,
1332
+ "grad_norm": 0.6432831287384033,
1333
+ "learning_rate": 1.111111111111111e-05,
1334
+ "loss": 7.0268,
1335
+ "step": 188
1336
+ },
1337
+ {
1338
+ "epoch": 1.9090909090909092,
1339
+ "grad_norm": 1.0608913898468018,
1340
+ "learning_rate": 1.101010101010101e-05,
1341
+ "loss": 6.9549,
1342
+ "step": 189
1343
+ },
1344
+ {
1345
+ "epoch": 1.9191919191919191,
1346
+ "grad_norm": 0.7320489883422852,
1347
+ "learning_rate": 1.0909090909090909e-05,
1348
+ "loss": 6.9866,
1349
+ "step": 190
1350
+ },
1351
+ {
1352
+ "epoch": 1.9292929292929293,
1353
+ "grad_norm": 1.776832103729248,
1354
+ "learning_rate": 1.0808080808080808e-05,
1355
+ "loss": 6.8107,
1356
+ "step": 191
1357
+ },
1358
+ {
1359
+ "epoch": 1.9393939393939394,
1360
+ "grad_norm": 0.8337501883506775,
1361
+ "learning_rate": 1.0707070707070706e-05,
1362
+ "loss": 6.9117,
1363
+ "step": 192
1364
+ },
1365
+ {
1366
+ "epoch": 1.9494949494949494,
1367
+ "grad_norm": 0.9856388568878174,
1368
+ "learning_rate": 1.0606060606060606e-05,
1369
+ "loss": 6.8311,
1370
+ "step": 193
1371
+ },
1372
+ {
1373
+ "epoch": 1.9595959595959596,
1374
+ "grad_norm": 1.3855446577072144,
1375
+ "learning_rate": 1.0505050505050505e-05,
1376
+ "loss": 7.1336,
1377
+ "step": 194
1378
+ },
1379
+ {
1380
+ "epoch": 1.9696969696969697,
1381
+ "grad_norm": 1.5992196798324585,
1382
+ "learning_rate": 1.0404040404040403e-05,
1383
+ "loss": 6.6699,
1384
+ "step": 195
1385
+ },
1386
+ {
1387
+ "epoch": 1.9797979797979797,
1388
+ "grad_norm": 2.3153018951416016,
1389
+ "learning_rate": 1.0303030303030302e-05,
1390
+ "loss": 7.2326,
1391
+ "step": 196
1392
+ },
1393
+ {
1394
+ "epoch": 1.98989898989899,
1395
+ "grad_norm": 0.5802290439605713,
1396
+ "learning_rate": 1.0202020202020202e-05,
1397
+ "loss": 6.9314,
1398
+ "step": 197
1399
+ },
1400
+ {
1401
+ "epoch": 2.0,
1402
+ "grad_norm": 0.8023201823234558,
1403
+ "learning_rate": 1.0101010101010101e-05,
1404
+ "loss": 6.7947,
1405
+ "step": 198
1406
+ },
1407
+ {
1408
+ "epoch": 2.0,
1409
+ "eval_audio_cosine_sim": 0.37472105026245117,
1410
+ "eval_loss": 5.077270030975342,
1411
+ "eval_runtime": 152.9953,
1412
+ "eval_samples_per_second": 0.059,
1413
+ "eval_steps_per_second": 0.02,
1414
+ "eval_text_cosine_sim": 0.1508849412202835,
1415
+ "step": 198
1416
+ },
1417
+ {
1418
+ "epoch": 2.01010101010101,
1419
+ "grad_norm": 0.8875887989997864,
1420
+ "learning_rate": 9.999999999999999e-06,
1421
+ "loss": 6.9461,
1422
+ "step": 199
1423
+ },
1424
+ {
1425
+ "epoch": 2.0202020202020203,
1426
+ "grad_norm": 0.9611765742301941,
1427
+ "learning_rate": 9.898989898989899e-06,
1428
+ "loss": 6.9659,
1429
+ "step": 200
1430
+ },
1431
+ {
1432
+ "epoch": 2.0303030303030303,
1433
+ "grad_norm": 1.539200782775879,
1434
+ "learning_rate": 9.797979797979798e-06,
1435
+ "loss": 6.6096,
1436
+ "step": 201
1437
+ },
1438
+ {
1439
+ "epoch": 2.04040404040404,
1440
+ "grad_norm": 0.8153275847434998,
1441
+ "learning_rate": 9.696969696969698e-06,
1442
+ "loss": 6.8189,
1443
+ "step": 202
1444
+ },
1445
+ {
1446
+ "epoch": 2.0505050505050506,
1447
+ "grad_norm": 0.8883192539215088,
1448
+ "learning_rate": 9.595959595959595e-06,
1449
+ "loss": 6.8387,
1450
+ "step": 203
1451
+ },
1452
+ {
1453
+ "epoch": 2.0606060606060606,
1454
+ "grad_norm": 1.5102068185806274,
1455
+ "learning_rate": 9.494949494949495e-06,
1456
+ "loss": 6.5686,
1457
+ "step": 204
1458
+ },
1459
+ {
1460
+ "epoch": 2.0707070707070705,
1461
+ "grad_norm": 0.86061692237854,
1462
+ "learning_rate": 9.393939393939394e-06,
1463
+ "loss": 6.8586,
1464
+ "step": 205
1465
+ },
1466
+ {
1467
+ "epoch": 2.080808080808081,
1468
+ "grad_norm": 1.1872936487197876,
1469
+ "learning_rate": 9.292929292929292e-06,
1470
+ "loss": 6.9541,
1471
+ "step": 206
1472
+ },
1473
+ {
1474
+ "epoch": 2.090909090909091,
1475
+ "grad_norm": 0.6017976999282837,
1476
+ "learning_rate": 9.191919191919192e-06,
1477
+ "loss": 6.7969,
1478
+ "step": 207
1479
+ },
1480
+ {
1481
+ "epoch": 2.101010101010101,
1482
+ "grad_norm": 2.1137189865112305,
1483
+ "learning_rate": 9.090909090909091e-06,
1484
+ "loss": 7.2209,
1485
+ "step": 208
1486
+ },
1487
+ {
1488
+ "epoch": 2.111111111111111,
1489
+ "grad_norm": 0.7061384916305542,
1490
+ "learning_rate": 8.98989898989899e-06,
1491
+ "loss": 6.7687,
1492
+ "step": 209
1493
+ },
1494
+ {
1495
+ "epoch": 2.121212121212121,
1496
+ "grad_norm": 0.7401531934738159,
1497
+ "learning_rate": 8.888888888888888e-06,
1498
+ "loss": 6.8944,
1499
+ "step": 210
1500
+ },
1501
+ {
1502
+ "epoch": 2.1313131313131315,
1503
+ "grad_norm": 1.1122565269470215,
1504
+ "learning_rate": 8.787878787878788e-06,
1505
+ "loss": 6.6628,
1506
+ "step": 211
1507
+ },
1508
+ {
1509
+ "epoch": 2.1414141414141414,
1510
+ "grad_norm": 0.6532828211784363,
1511
+ "learning_rate": 8.686868686868687e-06,
1512
+ "loss": 6.8288,
1513
+ "step": 212
1514
+ },
1515
+ {
1516
+ "epoch": 2.1515151515151514,
1517
+ "grad_norm": 1.3121882677078247,
1518
+ "learning_rate": 8.585858585858585e-06,
1519
+ "loss": 7.0173,
1520
+ "step": 213
1521
+ },
1522
+ {
1523
+ "epoch": 2.1616161616161618,
1524
+ "grad_norm": 1.4856034517288208,
1525
+ "learning_rate": 8.484848484848485e-06,
1526
+ "loss": 7.0473,
1527
+ "step": 214
1528
+ },
1529
+ {
1530
+ "epoch": 2.1717171717171717,
1531
+ "grad_norm": 1.0639023780822754,
1532
+ "learning_rate": 8.383838383838384e-06,
1533
+ "loss": 6.9259,
1534
+ "step": 215
1535
+ },
1536
+ {
1537
+ "epoch": 2.1818181818181817,
1538
+ "grad_norm": 0.9515113830566406,
1539
+ "learning_rate": 8.282828282828283e-06,
1540
+ "loss": 6.9661,
1541
+ "step": 216
1542
+ },
1543
+ {
1544
+ "epoch": 2.191919191919192,
1545
+ "grad_norm": 0.6862073540687561,
1546
+ "learning_rate": 8.181818181818181e-06,
1547
+ "loss": 6.7838,
1548
+ "step": 217
1549
+ },
1550
+ {
1551
+ "epoch": 2.202020202020202,
1552
+ "grad_norm": 0.6430158615112305,
1553
+ "learning_rate": 8.08080808080808e-06,
1554
+ "loss": 6.9264,
1555
+ "step": 218
1556
+ },
1557
+ {
1558
+ "epoch": 2.212121212121212,
1559
+ "grad_norm": 0.9119675159454346,
1560
+ "learning_rate": 7.97979797979798e-06,
1561
+ "loss": 6.9994,
1562
+ "step": 219
1563
+ },
1564
+ {
1565
+ "epoch": 2.2222222222222223,
1566
+ "grad_norm": 1.3677645921707153,
1567
+ "learning_rate": 7.87878787878788e-06,
1568
+ "loss": 6.6887,
1569
+ "step": 220
1570
+ },
1571
+ {
1572
+ "epoch": 2.2323232323232323,
1573
+ "grad_norm": 1.3011837005615234,
1574
+ "learning_rate": 7.777777777777777e-06,
1575
+ "loss": 6.6735,
1576
+ "step": 221
1577
+ },
1578
+ {
1579
+ "epoch": 2.242424242424242,
1580
+ "grad_norm": 0.8368801474571228,
1581
+ "learning_rate": 7.676767676767677e-06,
1582
+ "loss": 6.9868,
1583
+ "step": 222
1584
+ },
1585
+ {
1586
+ "epoch": 2.2525252525252526,
1587
+ "grad_norm": 1.2124541997909546,
1588
+ "learning_rate": 7.5757575757575764e-06,
1589
+ "loss": 7.0907,
1590
+ "step": 223
1591
+ },
1592
+ {
1593
+ "epoch": 2.2626262626262625,
1594
+ "grad_norm": 0.5335760116577148,
1595
+ "learning_rate": 7.474747474747475e-06,
1596
+ "loss": 6.9152,
1597
+ "step": 224
1598
+ },
1599
+ {
1600
+ "epoch": 2.2727272727272725,
1601
+ "grad_norm": 1.2292662858963013,
1602
+ "learning_rate": 7.373737373737374e-06,
1603
+ "loss": 7.0845,
1604
+ "step": 225
1605
+ },
1606
+ {
1607
+ "epoch": 2.282828282828283,
1608
+ "grad_norm": 1.1617155075073242,
1609
+ "learning_rate": 7.272727272727273e-06,
1610
+ "loss": 6.8596,
1611
+ "step": 226
1612
+ },
1613
+ {
1614
+ "epoch": 2.292929292929293,
1615
+ "grad_norm": 1.7857162952423096,
1616
+ "learning_rate": 7.171717171717172e-06,
1617
+ "loss": 6.7016,
1618
+ "step": 227
1619
+ },
1620
+ {
1621
+ "epoch": 2.303030303030303,
1622
+ "grad_norm": 0.7340604066848755,
1623
+ "learning_rate": 7.070707070707071e-06,
1624
+ "loss": 6.924,
1625
+ "step": 228
1626
+ },
1627
+ {
1628
+ "epoch": 2.313131313131313,
1629
+ "grad_norm": 1.880449652671814,
1630
+ "learning_rate": 6.96969696969697e-06,
1631
+ "loss": 7.1343,
1632
+ "step": 229
1633
+ },
1634
+ {
1635
+ "epoch": 2.323232323232323,
1636
+ "grad_norm": 0.9435675144195557,
1637
+ "learning_rate": 6.868686868686869e-06,
1638
+ "loss": 6.8979,
1639
+ "step": 230
1640
+ },
1641
+ {
1642
+ "epoch": 2.3333333333333335,
1643
+ "grad_norm": 0.877120316028595,
1644
+ "learning_rate": 6.767676767676768e-06,
1645
+ "loss": 6.8256,
1646
+ "step": 231
1647
+ },
1648
+ {
1649
+ "epoch": 2.3434343434343434,
1650
+ "grad_norm": 0.9317502975463867,
1651
+ "learning_rate": 6.666666666666667e-06,
1652
+ "loss": 6.8409,
1653
+ "step": 232
1654
+ },
1655
+ {
1656
+ "epoch": 2.3535353535353534,
1657
+ "grad_norm": 1.0060955286026,
1658
+ "learning_rate": 6.565656565656566e-06,
1659
+ "loss": 7.0845,
1660
+ "step": 233
1661
+ },
1662
+ {
1663
+ "epoch": 2.3636363636363638,
1664
+ "grad_norm": 0.8277117609977722,
1665
+ "learning_rate": 6.464646464646465e-06,
1666
+ "loss": 7.092,
1667
+ "step": 234
1668
+ },
1669
+ {
1670
+ "epoch": 2.3737373737373737,
1671
+ "grad_norm": 0.755837619304657,
1672
+ "learning_rate": 6.363636363636364e-06,
1673
+ "loss": 6.8859,
1674
+ "step": 235
1675
+ },
1676
+ {
1677
+ "epoch": 2.3838383838383836,
1678
+ "grad_norm": 0.6454436182975769,
1679
+ "learning_rate": 6.262626262626263e-06,
1680
+ "loss": 7.0516,
1681
+ "step": 236
1682
+ },
1683
+ {
1684
+ "epoch": 2.393939393939394,
1685
+ "grad_norm": 1.4383716583251953,
1686
+ "learning_rate": 6.161616161616162e-06,
1687
+ "loss": 6.7176,
1688
+ "step": 237
1689
+ },
1690
+ {
1691
+ "epoch": 2.404040404040404,
1692
+ "grad_norm": 1.2009567022323608,
1693
+ "learning_rate": 6.060606060606061e-06,
1694
+ "loss": 6.9433,
1695
+ "step": 238
1696
+ },
1697
+ {
1698
+ "epoch": 2.4141414141414144,
1699
+ "grad_norm": 0.7479004859924316,
1700
+ "learning_rate": 5.9595959595959605e-06,
1701
+ "loss": 6.9019,
1702
+ "step": 239
1703
+ },
1704
+ {
1705
+ "epoch": 2.4242424242424243,
1706
+ "grad_norm": 0.6840262413024902,
1707
+ "learning_rate": 5.858585858585859e-06,
1708
+ "loss": 6.9329,
1709
+ "step": 240
1710
+ },
1711
+ {
1712
+ "epoch": 2.4343434343434343,
1713
+ "grad_norm": 0.6570662260055542,
1714
+ "learning_rate": 5.757575757575758e-06,
1715
+ "loss": 6.9839,
1716
+ "step": 241
1717
+ },
1718
+ {
1719
+ "epoch": 2.4444444444444446,
1720
+ "grad_norm": 1.0435160398483276,
1721
+ "learning_rate": 5.656565656565657e-06,
1722
+ "loss": 6.9886,
1723
+ "step": 242
1724
+ },
1725
+ {
1726
+ "epoch": 2.4545454545454546,
1727
+ "grad_norm": 1.2880451679229736,
1728
+ "learning_rate": 5.555555555555555e-06,
1729
+ "loss": 7.0024,
1730
+ "step": 243
1731
+ },
1732
+ {
1733
+ "epoch": 2.4646464646464645,
1734
+ "grad_norm": 1.2956496477127075,
1735
+ "learning_rate": 5.4545454545454545e-06,
1736
+ "loss": 7.1377,
1737
+ "step": 244
1738
+ },
1739
+ {
1740
+ "epoch": 2.474747474747475,
1741
+ "grad_norm": 1.3433921337127686,
1742
+ "learning_rate": 5.353535353535353e-06,
1743
+ "loss": 7.0797,
1744
+ "step": 245
1745
+ },
1746
+ {
1747
+ "epoch": 2.484848484848485,
1748
+ "grad_norm": 0.6715502738952637,
1749
+ "learning_rate": 5.2525252525252526e-06,
1750
+ "loss": 6.8772,
1751
+ "step": 246
1752
+ },
1753
+ {
1754
+ "epoch": 2.494949494949495,
1755
+ "grad_norm": 0.8352294564247131,
1756
+ "learning_rate": 5.151515151515151e-06,
1757
+ "loss": 7.0575,
1758
+ "step": 247
1759
+ },
1760
+ {
1761
+ "epoch": 2.505050505050505,
1762
+ "grad_norm": 1.1144355535507202,
1763
+ "learning_rate": 5.050505050505051e-06,
1764
+ "loss": 6.8229,
1765
+ "step": 248
1766
+ },
1767
+ {
1768
+ "epoch": 2.515151515151515,
1769
+ "grad_norm": 0.6648267507553101,
1770
+ "learning_rate": 4.949494949494949e-06,
1771
+ "loss": 7.0238,
1772
+ "step": 249
1773
+ },
1774
+ {
1775
+ "epoch": 2.525252525252525,
1776
+ "grad_norm": 0.7869075536727905,
1777
+ "learning_rate": 4.848484848484849e-06,
1778
+ "loss": 7.0076,
1779
+ "step": 250
1780
+ },
1781
+ {
1782
+ "epoch": 2.5353535353535355,
1783
+ "grad_norm": 0.6390742659568787,
1784
+ "learning_rate": 4.747474747474747e-06,
1785
+ "loss": 7.0491,
1786
+ "step": 251
1787
+ },
1788
+ {
1789
+ "epoch": 2.5454545454545454,
1790
+ "grad_norm": 0.7328326106071472,
1791
+ "learning_rate": 4.646464646464646e-06,
1792
+ "loss": 7.0393,
1793
+ "step": 252
1794
+ },
1795
+ {
1796
+ "epoch": 2.5555555555555554,
1797
+ "grad_norm": 0.8217814564704895,
1798
+ "learning_rate": 4.5454545454545455e-06,
1799
+ "loss": 6.9342,
1800
+ "step": 253
1801
+ },
1802
+ {
1803
+ "epoch": 2.5656565656565657,
1804
+ "grad_norm": 1.3768435716629028,
1805
+ "learning_rate": 4.444444444444444e-06,
1806
+ "loss": 7.124,
1807
+ "step": 254
1808
+ },
1809
+ {
1810
+ "epoch": 2.5757575757575757,
1811
+ "grad_norm": 1.0091509819030762,
1812
+ "learning_rate": 4.343434343434344e-06,
1813
+ "loss": 6.8746,
1814
+ "step": 255
1815
+ },
1816
+ {
1817
+ "epoch": 2.5858585858585856,
1818
+ "grad_norm": 1.0249191522598267,
1819
+ "learning_rate": 4.242424242424242e-06,
1820
+ "loss": 6.8636,
1821
+ "step": 256
1822
+ },
1823
+ {
1824
+ "epoch": 2.595959595959596,
1825
+ "grad_norm": 0.9580966234207153,
1826
+ "learning_rate": 4.141414141414142e-06,
1827
+ "loss": 6.9923,
1828
+ "step": 257
1829
+ },
1830
+ {
1831
+ "epoch": 2.606060606060606,
1832
+ "grad_norm": 1.5487310886383057,
1833
+ "learning_rate": 4.04040404040404e-06,
1834
+ "loss": 6.8229,
1835
+ "step": 258
1836
+ },
1837
+ {
1838
+ "epoch": 2.616161616161616,
1839
+ "grad_norm": 0.6319442987442017,
1840
+ "learning_rate": 3.93939393939394e-06,
1841
+ "loss": 7.0224,
1842
+ "step": 259
1843
+ },
1844
+ {
1845
+ "epoch": 2.6262626262626263,
1846
+ "grad_norm": 0.7614464163780212,
1847
+ "learning_rate": 3.8383838383838385e-06,
1848
+ "loss": 6.9499,
1849
+ "step": 260
1850
+ },
1851
+ {
1852
+ "epoch": 2.6363636363636362,
1853
+ "grad_norm": 0.9656277298927307,
1854
+ "learning_rate": 3.7373737373737375e-06,
1855
+ "loss": 6.881,
1856
+ "step": 261
1857
+ },
1858
+ {
1859
+ "epoch": 2.6464646464646466,
1860
+ "grad_norm": 2.4072742462158203,
1861
+ "learning_rate": 3.6363636363636366e-06,
1862
+ "loss": 6.579,
1863
+ "step": 262
1864
+ },
1865
+ {
1866
+ "epoch": 2.6565656565656566,
1867
+ "grad_norm": 1.361374855041504,
1868
+ "learning_rate": 3.5353535353535356e-06,
1869
+ "loss": 7.1662,
1870
+ "step": 263
1871
+ },
1872
+ {
1873
+ "epoch": 2.6666666666666665,
1874
+ "grad_norm": 0.7471557855606079,
1875
+ "learning_rate": 3.4343434343434347e-06,
1876
+ "loss": 7.026,
1877
+ "step": 264
1878
+ },
1879
+ {
1880
+ "epoch": 2.676767676767677,
1881
+ "grad_norm": 0.9399821758270264,
1882
+ "learning_rate": 3.3333333333333333e-06,
1883
+ "loss": 6.8437,
1884
+ "step": 265
1885
+ },
1886
+ {
1887
+ "epoch": 2.686868686868687,
1888
+ "grad_norm": 0.960325300693512,
1889
+ "learning_rate": 3.2323232323232324e-06,
1890
+ "loss": 7.1221,
1891
+ "step": 266
1892
+ },
1893
+ {
1894
+ "epoch": 2.6969696969696972,
1895
+ "grad_norm": 0.7455516457557678,
1896
+ "learning_rate": 3.1313131313131314e-06,
1897
+ "loss": 6.8993,
1898
+ "step": 267
1899
+ },
1900
+ {
1901
+ "epoch": 2.707070707070707,
1902
+ "grad_norm": 0.7760592103004456,
1903
+ "learning_rate": 3.0303030303030305e-06,
1904
+ "loss": 6.9543,
1905
+ "step": 268
1906
+ },
1907
+ {
1908
+ "epoch": 2.717171717171717,
1909
+ "grad_norm": 0.7888494729995728,
1910
+ "learning_rate": 2.9292929292929295e-06,
1911
+ "loss": 7.0662,
1912
+ "step": 269
1913
+ },
1914
+ {
1915
+ "epoch": 2.7272727272727275,
1916
+ "grad_norm": 1.3352677822113037,
1917
+ "learning_rate": 2.8282828282828286e-06,
1918
+ "loss": 6.801,
1919
+ "step": 270
1920
+ },
1921
+ {
1922
+ "epoch": 2.7373737373737375,
1923
+ "grad_norm": 0.7915419936180115,
1924
+ "learning_rate": 2.7272727272727272e-06,
1925
+ "loss": 7.0028,
1926
+ "step": 271
1927
+ },
1928
+ {
1929
+ "epoch": 2.7474747474747474,
1930
+ "grad_norm": 0.5938275456428528,
1931
+ "learning_rate": 2.6262626262626263e-06,
1932
+ "loss": 6.9275,
1933
+ "step": 272
1934
+ },
1935
+ {
1936
+ "epoch": 2.757575757575758,
1937
+ "grad_norm": 1.24167001247406,
1938
+ "learning_rate": 2.5252525252525253e-06,
1939
+ "loss": 6.7646,
1940
+ "step": 273
1941
+ },
1942
+ {
1943
+ "epoch": 2.7676767676767677,
1944
+ "grad_norm": 1.3691089153289795,
1945
+ "learning_rate": 2.4242424242424244e-06,
1946
+ "loss": 7.0922,
1947
+ "step": 274
1948
+ },
1949
+ {
1950
+ "epoch": 2.7777777777777777,
1951
+ "grad_norm": 1.1850454807281494,
1952
+ "learning_rate": 2.323232323232323e-06,
1953
+ "loss": 7.0915,
1954
+ "step": 275
1955
+ },
1956
+ {
1957
+ "epoch": 2.787878787878788,
1958
+ "grad_norm": 1.693428874015808,
1959
+ "learning_rate": 2.222222222222222e-06,
1960
+ "loss": 6.7004,
1961
+ "step": 276
1962
+ },
1963
+ {
1964
+ "epoch": 2.797979797979798,
1965
+ "grad_norm": 0.5935156345367432,
1966
+ "learning_rate": 2.121212121212121e-06,
1967
+ "loss": 6.9309,
1968
+ "step": 277
1969
+ },
1970
+ {
1971
+ "epoch": 2.808080808080808,
1972
+ "grad_norm": 0.6563141942024231,
1973
+ "learning_rate": 2.02020202020202e-06,
1974
+ "loss": 7.0097,
1975
+ "step": 278
1976
+ },
1977
+ {
1978
+ "epoch": 2.8181818181818183,
1979
+ "grad_norm": 0.9683242440223694,
1980
+ "learning_rate": 1.9191919191919192e-06,
1981
+ "loss": 7.0719,
1982
+ "step": 279
1983
+ },
1984
+ {
1985
+ "epoch": 2.8282828282828283,
1986
+ "grad_norm": 1.0628868341445923,
1987
+ "learning_rate": 1.8181818181818183e-06,
1988
+ "loss": 7.0084,
1989
+ "step": 280
1990
+ },
1991
+ {
1992
+ "epoch": 2.8383838383838382,
1993
+ "grad_norm": 0.7137049436569214,
1994
+ "learning_rate": 1.7171717171717173e-06,
1995
+ "loss": 6.9748,
1996
+ "step": 281
1997
+ },
1998
+ {
1999
+ "epoch": 2.8484848484848486,
2000
+ "grad_norm": 1.1555520296096802,
2001
+ "learning_rate": 1.6161616161616162e-06,
2002
+ "loss": 6.8182,
2003
+ "step": 282
2004
+ },
2005
+ {
2006
+ "epoch": 2.8585858585858586,
2007
+ "grad_norm": 1.3852341175079346,
2008
+ "learning_rate": 1.5151515151515152e-06,
2009
+ "loss": 7.1022,
2010
+ "step": 283
2011
+ },
2012
+ {
2013
+ "epoch": 2.8686868686868685,
2014
+ "grad_norm": 0.6663222908973694,
2015
+ "learning_rate": 1.4141414141414143e-06,
2016
+ "loss": 7.0444,
2017
+ "step": 284
2018
+ },
2019
+ {
2020
+ "epoch": 2.878787878787879,
2021
+ "grad_norm": 0.9903120994567871,
2022
+ "learning_rate": 1.3131313131313131e-06,
2023
+ "loss": 6.7819,
2024
+ "step": 285
2025
+ },
2026
+ {
2027
+ "epoch": 2.888888888888889,
2028
+ "grad_norm": 1.510622262954712,
2029
+ "learning_rate": 1.2121212121212122e-06,
2030
+ "loss": 7.1841,
2031
+ "step": 286
2032
+ },
2033
+ {
2034
+ "epoch": 2.898989898989899,
2035
+ "grad_norm": 1.2110823392868042,
2036
+ "learning_rate": 1.111111111111111e-06,
2037
+ "loss": 7.0199,
2038
+ "step": 287
2039
+ },
2040
+ {
2041
+ "epoch": 2.909090909090909,
2042
+ "grad_norm": 1.0056835412979126,
2043
+ "learning_rate": 1.01010101010101e-06,
2044
+ "loss": 6.9958,
2045
+ "step": 288
2046
+ },
2047
+ {
2048
+ "epoch": 2.919191919191919,
2049
+ "grad_norm": 0.8939002156257629,
2050
+ "learning_rate": 9.090909090909091e-07,
2051
+ "loss": 7.0829,
2052
+ "step": 289
2053
+ },
2054
+ {
2055
+ "epoch": 2.929292929292929,
2056
+ "grad_norm": 0.6934316158294678,
2057
+ "learning_rate": 8.080808080808081e-07,
2058
+ "loss": 7.015,
2059
+ "step": 290
2060
+ },
2061
+ {
2062
+ "epoch": 2.9393939393939394,
2063
+ "grad_norm": 0.7522878050804138,
2064
+ "learning_rate": 7.070707070707071e-07,
2065
+ "loss": 7.0191,
2066
+ "step": 291
2067
+ },
2068
+ {
2069
+ "epoch": 2.9494949494949494,
2070
+ "grad_norm": 1.2762123346328735,
2071
+ "learning_rate": 6.060606060606061e-07,
2072
+ "loss": 6.7045,
2073
+ "step": 292
2074
+ },
2075
+ {
2076
+ "epoch": 2.9595959595959593,
2077
+ "grad_norm": 1.0797396898269653,
2078
+ "learning_rate": 5.05050505050505e-07,
2079
+ "loss": 7.0543,
2080
+ "step": 293
2081
+ },
2082
+ {
2083
+ "epoch": 2.9696969696969697,
2084
+ "grad_norm": 1.322259783744812,
2085
+ "learning_rate": 4.0404040404040405e-07,
2086
+ "loss": 6.7815,
2087
+ "step": 294
2088
+ },
2089
+ {
2090
+ "epoch": 2.9797979797979797,
2091
+ "grad_norm": 0.63960200548172,
2092
+ "learning_rate": 3.0303030303030305e-07,
2093
+ "loss": 6.9333,
2094
+ "step": 295
2095
+ },
2096
+ {
2097
+ "epoch": 2.98989898989899,
2098
+ "grad_norm": 0.8398324847221375,
2099
+ "learning_rate": 2.0202020202020202e-07,
2100
+ "loss": 7.0499,
2101
+ "step": 296
2102
+ },
2103
+ {
2104
+ "epoch": 3.0,
2105
+ "grad_norm": 1.3194384574890137,
2106
+ "learning_rate": 1.0101010101010101e-07,
2107
+ "loss": 7.1508,
2108
+ "step": 297
2109
+ },
2110
+ {
2111
+ "epoch": 3.0,
2112
+ "eval_audio_cosine_sim": 0.32165560126304626,
2113
+ "eval_loss": 5.101603984832764,
2114
+ "eval_runtime": 153.0389,
2115
+ "eval_samples_per_second": 0.059,
2116
+ "eval_steps_per_second": 0.02,
2117
+ "eval_text_cosine_sim": 0.11808153986930847,
2118
+ "step": 297
2119
+ },
2120
+ {
2121
+ "epoch": 3.0,
2122
+ "step": 297,
2123
+ "total_flos": 6235958582540448.0,
2124
+ "train_loss": 7.227042491989907,
2125
+ "train_runtime": 3530.6424,
2126
+ "train_samples_per_second": 1.343,
2127
+ "train_steps_per_second": 0.084
2128
+ }
2129
+ ],
2130
+ "logging_steps": 1.0,
2131
+ "max_steps": 297,
2132
+ "num_input_tokens_seen": 0,
2133
+ "num_train_epochs": 3,
2134
+ "save_steps": 500,
2135
+ "stateful_callbacks": {
2136
+ "TrainerControl": {
2137
+ "args": {
2138
+ "should_epoch_stop": false,
2139
+ "should_evaluate": false,
2140
+ "should_log": false,
2141
+ "should_save": false,
2142
+ "should_training_stop": false
2143
+ },
2144
+ "attributes": {}
2145
+ }
2146
+ },
2147
+ "total_flos": 6235958582540448.0,
2148
+ "train_batch_size": 16,
2149
+ "trial_name": null,
2150
+ "trial_params": null
2151
+ }