Upload folder using huggingface_hub

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:ea6cc0002ac99d442c85e1d2cdd9b7577e61ecc0fb78d4409421e6b92e3464ce
+oid sha256:6b0561791886e4939861077949e479c8783ed3ae4b2d376221059dff867a57d2
 size 10208852910

rng_state.pth

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:87ee2f2ae8de49581bfeb04ce113841200a5b6fdd727c9383b4d28dadbf204e2
+oid sha256:d0271b5c2fa4362a8f18ea19b1df28175d65cb4c4813cb8adefba0a2752315e9
 size 14645

scheduler.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:0e776e741fba5a953308d206a95e50abed2908736b9041a4ce113333636585ee
+oid sha256:b713d7c231b3ec6639ccb2ae4d4dc45f4c22877b54b3181bff176c54025c6a85
 size 1465

trainer_state.json

@@ -3,489 +3,562 @@
   "best_metric": null,
   "best_model_checkpoint": null,
   "epoch": 1.0,
-  "eval_steps": 143,
-  "global_step": 572,
+  "eval_steps": 53,
+  "global_step": 528,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
   "log_history": [
     {
-      "epoch": 0.0175054704595186,
-      "grad_norm": 1.09375,
+      "epoch": 0.018957345971563982,
+      "grad_norm": 1.484375,
       "learning_rate": 4.5e-06,
-      "loss": 0.3692042827606201,
+      "loss": 0.37855618000030516,
       "step": 10
     },
     {
-      "epoch": 0.0350109409190372,
-      "grad_norm": 0.93359375,
+      "epoch": 0.037914691943127965,
+      "grad_norm": 0.78125,
       "learning_rate": 9.5e-06,
-      "loss": 0.35977413654327395,
+      "loss": 0.374945330619812,
       "step": 20
     },
     {
-      "epoch": 0.0525164113785558,
-      "grad_norm": 1.046875,
+      "epoch": 0.05687203791469194,
+      "grad_norm": 0.81640625,
       "learning_rate": 1.45e-05,
-      "loss": 0.36324713230133054,
+      "loss": 0.362351393699646,
       "step": 30
     },
     {
-      "epoch": 0.0700218818380744,
-      "grad_norm": 1.0625,
+      "epoch": 0.07582938388625593,
+      "grad_norm": 0.94140625,
       "learning_rate": 1.9500000000000003e-05,
-      "loss": 0.33168764114379884,
+      "loss": 0.3529045104980469,
       "step": 40
     },
     {
-      "epoch": 0.087527352297593,
-      "grad_norm": 0.828125,
+      "epoch": 0.0947867298578199,
+      "grad_norm": 0.82421875,
       "learning_rate": 2.45e-05,
-      "loss": 0.33778228759765627,
+      "loss": 0.33787591457366944,
       "step": 50
     },
     {
-      "epoch": 0.1050328227571116,
-      "grad_norm": 1.078125,
+      "epoch": 0.1004739336492891,
+      "eval_text_loss": 1.4636504650115967,
+      "eval_text_model_preparation_time": 0.0155,
+      "eval_text_runtime": 4.8317,
+      "eval_text_samples_per_second": 12.418,
+      "eval_text_steps_per_second": 12.418,
+      "step": 53
+    },
+    {
+      "epoch": 0.1004739336492891,
+      "eval_audio_loss": 0.7069099545478821,
+      "eval_audio_model_preparation_time": 0.0155,
+      "eval_audio_runtime": 21.032,
+      "eval_audio_samples_per_second": 6.324,
+      "eval_audio_steps_per_second": 6.324,
+      "step": 53
+    },
+    {
+      "epoch": 0.11374407582938388,
+      "grad_norm": 0.99609375,
       "learning_rate": 2.95e-05,
-      "loss": 0.2982091665267944,
+      "loss": 0.33856768608093263,
       "step": 60
     },
     {
-      "epoch": 0.12253829321663019,
-      "grad_norm": 0.859375,
+      "epoch": 0.13270142180094788,
+      "grad_norm": 0.65234375,
       "learning_rate": 3.45e-05,
-      "loss": 0.3437821626663208,
+      "loss": 0.35529475212097167,
       "step": 70
     },
     {
-      "epoch": 0.1400437636761488,
-      "grad_norm": 1.015625,
+      "epoch": 0.15165876777251186,
+      "grad_norm": 0.87109375,
       "learning_rate": 3.9500000000000005e-05,
-      "loss": 0.29782021045684814,
+      "loss": 0.2936519384384155,
       "step": 80
     },
     {
-      "epoch": 0.1575492341356674,
-      "grad_norm": 0.6171875,
+      "epoch": 0.17061611374407584,
+      "grad_norm": 0.68359375,
       "learning_rate": 4.4500000000000004e-05,
-      "loss": 0.30974411964416504,
+      "loss": 0.318320107460022,
       "step": 90
     },
     {
-      "epoch": 0.175054704595186,
-      "grad_norm": 0.6796875,
+      "epoch": 0.1895734597156398,
+      "grad_norm": 0.7890625,
       "learning_rate": 4.9500000000000004e-05,
-      "loss": 0.31342487335205077,
+      "loss": 0.31594276428222656,
       "step": 100
     },
     {
-      "epoch": 0.1925601750547046,
-      "grad_norm": 0.62890625,
-      "learning_rate": 4.995515841989726e-05,
-      "loss": 0.2926024436950684,
+      "epoch": 0.2009478672985782,
+      "eval_text_loss": 1.329001545906067,
+      "eval_text_model_preparation_time": 0.0155,
+      "eval_text_runtime": 4.4604,
+      "eval_text_samples_per_second": 13.452,
+      "eval_text_steps_per_second": 13.452,
+      "step": 106
+    },
+    {
+      "epoch": 0.2009478672985782,
+      "eval_audio_loss": 0.7277711629867554,
+      "eval_audio_model_preparation_time": 0.0155,
+      "eval_audio_runtime": 20.1176,
+      "eval_audio_samples_per_second": 6.611,
+      "eval_audio_steps_per_second": 6.611,
+      "step": 106
+    },
+    {
+      "epoch": 0.20853080568720378,
+      "grad_norm": 0.8203125,
+      "learning_rate": 4.9945468268142656e-05,
+      "loss": 0.31812191009521484,
       "step": 110
     },
     {
-      "epoch": 0.2100656455142232,
-      "grad_norm": 0.5703125,
-      "learning_rate": 4.9800356999122746e-05,
-      "loss": 0.2794817924499512,
+      "epoch": 0.22748815165876776,
+      "grad_norm": 1.0,
+      "learning_rate": 4.975726891929585e-05,
+      "loss": 0.29132957458496095,
       "step": 120
     },
     {
-      "epoch": 0.2275711159737418,
-      "grad_norm": 0.98828125,
-      "learning_rate": 4.953572751052924e-05,
-      "loss": 0.32305901050567626,
+      "epoch": 0.24644549763033174,
+      "grad_norm": 0.609375,
+      "learning_rate": 4.943574200733625e-05,
+      "loss": 0.3463184595108032,
       "step": 130
     },
     {
-      "epoch": 0.24507658643326038,
-      "grad_norm": 0.99609375,
-      "learning_rate": 4.916244186246555e-05,
-      "loss": 0.29492759704589844,
+      "epoch": 0.26540284360189575,
+      "grad_norm": 0.59375,
+      "learning_rate": 4.89826190802705e-05,
+      "loss": 0.3099159479141235,
       "step": 140
     },
     {
-      "epoch": 0.25032822757111595,
-      "eval_text_loss": 1.3007618188858032,
-      "eval_text_model_preparation_time": 0.0138,
-      "eval_text_runtime": 5.3163,
-      "eval_text_samples_per_second": 11.286,
-      "eval_text_steps_per_second": 11.286,
-      "step": 143
+      "epoch": 0.2843601895734597,
+      "grad_norm": 0.9765625,
+      "learning_rate": 4.8400340382046866e-05,
+      "loss": 0.29767818450927735,
+      "step": 150
     },
     {
-      "epoch": 0.25032822757111595,
-      "eval_audio_loss": 0.7166803479194641,
-      "eval_audio_model_preparation_time": 0.0138,
-      "eval_audio_runtime": 23.1192,
-      "eval_audio_samples_per_second": 5.753,
-      "eval_audio_steps_per_second": 5.753,
-      "step": 143
+      "epoch": 0.3014218009478673,
+      "eval_text_loss": 1.2909588813781738,
+      "eval_text_model_preparation_time": 0.0155,
+      "eval_text_runtime": 4.8675,
+      "eval_text_samples_per_second": 12.327,
+      "eval_text_steps_per_second": 12.327,
+      "step": 159
     },
     {
-      "epoch": 0.26258205689277897,
-      "grad_norm": 1.0390625,
-      "learning_rate": 4.868215314571065e-05,
-      "loss": 0.31501500606536864,
-      "step": 150
+      "epoch": 0.3014218009478673,
+      "eval_audio_loss": 0.7182490825653076,
+      "eval_audio_model_preparation_time": 0.0155,
+      "eval_audio_runtime": 19.9605,
+      "eval_audio_samples_per_second": 6.663,
+      "eval_audio_steps_per_second": 6.663,
+      "step": 159
     },
     {
-      "epoch": 0.2800875273522976,
-      "grad_norm": 0.609375,
-      "learning_rate": 4.8096988312782174e-05,
-      "loss": 0.3073582649230957,
+      "epoch": 0.3033175355450237,
+      "grad_norm": 1.0625,
+      "learning_rate": 4.769204171088849e-05,
+      "loss": 0.3072782039642334,
       "step": 160
     },
     {
-      "epoch": 0.2975929978118162,
-      "grad_norm": 1.1171875,
-      "learning_rate": 4.74095387587539e-05,
-      "loss": 0.3264436960220337,
+      "epoch": 0.3222748815165877,
+      "grad_norm": 0.69921875,
+      "learning_rate": 4.6861537531795094e-05,
+      "loss": 0.30300824642181395,
       "step": 170
     },
     {
-      "epoch": 0.3150984682713348,
-      "grad_norm": 0.6796875,
-      "learning_rate": 4.662284884529549e-05,
-      "loss": 0.3189807653427124,
+      "epoch": 0.3412322274881517,
+      "grad_norm": 0.8046875,
+      "learning_rate": 4.591330043415909e-05,
+      "loss": 0.3216289758682251,
       "step": 180
     },
     {
-      "epoch": 0.33260393873085337,
-      "grad_norm": 0.87109375,
-      "learning_rate": 4.5740402418755246e-05,
-      "loss": 0.3061279535293579,
+      "epoch": 0.36018957345971564,
+      "grad_norm": 0.90234375,
+      "learning_rate": 4.485243704512474e-05,
+      "loss": 0.3087095499038696,
       "step": 190
     },
     {
-      "epoch": 0.350109409190372,
-      "grad_norm": 0.85546875,
-      "learning_rate": 4.476610738199053e-05,
-      "loss": 0.2863796234130859,
+      "epoch": 0.3791469194312796,
+      "grad_norm": 0.84765625,
+      "learning_rate": 4.368466052840636e-05,
+      "loss": 0.3263149976730347,
       "step": 200
     },
     {
-      "epoch": 0.3676148796498906,
-      "grad_norm": 0.81640625,
-      "learning_rate": 4.370427838826934e-05,
-      "loss": 0.30284264087677004,
+      "epoch": 0.3981042654028436,
+      "grad_norm": 1.46875,
+      "learning_rate": 4.2416259816670235e-05,
+      "loss": 0.29808921813964845,
       "step": 210
     },
     {
-      "epoch": 0.3851203501094092,
-      "grad_norm": 0.6796875,
-      "learning_rate": 4.2559617733882754e-05,
-      "loss": 0.28477931022644043,
+      "epoch": 0.4018957345971564,
+      "eval_text_loss": 1.2724202871322632,
+      "eval_text_model_preparation_time": 0.0155,
+      "eval_text_runtime": 4.8821,
+      "eval_text_samples_per_second": 12.29,
+      "eval_text_steps_per_second": 12.29,
+      "step": 212
+    },
+    {
+      "epoch": 0.4018957345971564,
+      "eval_audio_loss": 0.7229443788528442,
+      "eval_audio_model_preparation_time": 0.0155,
+      "eval_audio_runtime": 20.5668,
+      "eval_audio_samples_per_second": 6.467,
+      "eval_audio_steps_per_second": 6.467,
+      "step": 212
+    },
+    {
+      "epoch": 0.41706161137440756,
+      "grad_norm": 0.828125,
+      "learning_rate": 4.1054065743175786e-05,
+      "loss": 0.313119101524353,
       "step": 220
     },
     {
-      "epoch": 0.4026258205689278,
-      "grad_norm": 1.2265625,
-      "learning_rate": 4.13371945340852e-05,
-      "loss": 0.29639883041381837,
+      "epoch": 0.43601895734597157,
+      "grad_norm": 1.03125,
+      "learning_rate": 3.9605414255070396e-05,
+      "loss": 0.2832280874252319,
       "step": 230
     },
     {
-      "epoch": 0.4201312910284464,
-      "grad_norm": 0.7734375,
-      "learning_rate": 4.004242227458158e-05,
-      "loss": 0.30428614616394045,
+      "epoch": 0.4549763033175355,
+      "grad_norm": 0.70703125,
+      "learning_rate": 3.8078106906448683e-05,
+      "loss": 0.29762136936187744,
       "step": 240
     },
     {
-      "epoch": 0.437636761487965,
-      "grad_norm": 0.7890625,
-      "learning_rate": 3.868103483797414e-05,
-      "loss": 0.28317761421203613,
+      "epoch": 0.47393364928909953,
+      "grad_norm": 0.83203125,
+      "learning_rate": 3.648036884393646e-05,
+      "loss": 0.3122119665145874,
       "step": 250
     },
     {
-      "epoch": 0.4551422319474836,
-      "grad_norm": 0.78515625,
-      "learning_rate": 3.725906111133579e-05,
-      "loss": 0.2817823886871338,
+      "epoch": 0.4928909952606635,
+      "grad_norm": 0.94140625,
+      "learning_rate": 3.4820804511063496e-05,
+      "loss": 0.2904823780059814,
       "step": 260
     },
     {
-      "epoch": 0.4726477024070022,
-      "grad_norm": 0.5859375,
-      "learning_rate": 3.578279828735983e-05,
-      "loss": 0.2978217601776123,
-      "step": 270
+      "epoch": 0.5023696682464455,
+      "eval_text_loss": 1.224345088005066,
+      "eval_text_model_preparation_time": 0.0155,
+      "eval_text_runtime": 4.548,
+      "eval_text_samples_per_second": 13.193,
+      "eval_text_steps_per_second": 13.193,
+      "step": 265
     },
     {
-      "epoch": 0.49015317286652077,
-      "grad_norm": 0.77734375,
-      "learning_rate": 3.4258783977321615e-05,
-      "loss": 0.28665401935577395,
-      "step": 280
+      "epoch": 0.5023696682464455,
+      "eval_audio_loss": 0.7233917713165283,
+      "eval_audio_model_preparation_time": 0.0155,
+      "eval_audio_runtime": 20.862,
+      "eval_audio_samples_per_second": 6.375,
+      "eval_audio_steps_per_second": 6.375,
+      "step": 265
     },
     {
-      "epoch": 0.5006564551422319,
-      "eval_text_loss": 1.2351676225662231,
-      "eval_text_model_preparation_time": 0.0138,
-      "eval_text_runtime": 5.5155,
-      "eval_text_samples_per_second": 10.878,
-      "eval_text_steps_per_second": 10.878,
-      "step": 286
+      "epoch": 0.5118483412322274,
+      "grad_norm": 1.2890625,
+      "learning_rate": 3.3108351309974284e-05,
+      "loss": 0.3033822298049927,
+      "step": 270
     },
     {
-      "epoch": 0.5006564551422319,
-      "eval_audio_loss": 0.7075363397598267,
-      "eval_audio_model_preparation_time": 0.0138,
-      "eval_audio_runtime": 24.8383,
-      "eval_audio_samples_per_second": 5.355,
-      "eval_audio_steps_per_second": 5.355,
-      "step": 286
+      "epoch": 0.5308056872037915,
+      "grad_norm": 0.6484375,
+      "learning_rate": 3.1352231470026584e-05,
+      "loss": 0.3055255651473999,
+      "step": 280
     },
     {
-      "epoch": 0.5076586433260394,
-      "grad_norm": 0.91015625,
-      "learning_rate": 3.2693767259349574e-05,
-      "loss": 0.30377969741821287,
+      "epoch": 0.5497630331753555,
+      "grad_norm": 0.90625,
+      "learning_rate": 2.956190238248425e-05,
+      "loss": 0.28981497287750246,
       "step": 290
     },
     {
-      "epoch": 0.5251641137855579,
-      "grad_norm": 0.72265625,
-      "learning_rate": 3.109467879021793e-05,
-      "loss": 0.2989386320114136,
+      "epoch": 0.5687203791469194,
+      "grad_norm": 1.1171875,
+      "learning_rate": 2.7747005668771293e-05,
+      "loss": 0.32351953983306886,
       "step": 300
     },
     {
-      "epoch": 0.5426695842450766,
-      "grad_norm": 0.765625,
-      "learning_rate": 2.9468600113020715e-05,
-      "loss": 0.2834346771240234,
+      "epoch": 0.5876777251184834,
+      "grad_norm": 0.71875,
+      "learning_rate": 2.591731525657454e-05,
+      "loss": 0.2949329614639282,
       "step": 310
     },
     {
-      "epoch": 0.5601750547045952,
-      "grad_norm": 0.80859375,
-      "learning_rate": 2.7822732296647548e-05,
-      "loss": 0.3039692401885986,
+      "epoch": 0.6028436018957346,
+      "eval_text_loss": 1.2085834741592407,
+      "eval_text_model_preparation_time": 0.0155,
+      "eval_text_runtime": 5.1662,
+      "eval_text_samples_per_second": 11.614,
+      "eval_text_steps_per_second": 11.614,
+      "step": 318
+    },
+    {
+      "epoch": 0.6028436018957346,
+      "eval_audio_loss": 0.7195463180541992,
+      "eval_audio_model_preparation_time": 0.0155,
+      "eval_audio_runtime": 20.8146,
+      "eval_audio_samples_per_second": 6.39,
+      "eval_audio_steps_per_second": 6.39,
+      "step": 318
+    },
+    {
+      "epoch": 0.6066350710900474,
+      "grad_norm": 0.7578125,
+      "learning_rate": 2.4082684743425458e-05,
+      "loss": 0.28318300247192385,
       "step": 320
     },
     {
-      "epoch": 0.5776805251641138,
-      "grad_norm": 0.8046875,
-      "learning_rate": 2.6164364045941042e-05,
-      "loss": 0.3042268753051758,
+      "epoch": 0.6255924170616114,
+      "grad_norm": 0.859375,
+      "learning_rate": 2.2252994331228713e-05,
+      "loss": 0.286310601234436,
       "step": 330
     },
     {
-      "epoch": 0.5951859956236324,
-      "grad_norm": 0.70703125,
-      "learning_rate": 2.450083942375939e-05,
-      "loss": 0.2903634548187256,
+      "epoch": 0.6445497630331753,
+      "grad_norm": 0.97265625,
+      "learning_rate": 2.0438097617515745e-05,
+      "loss": 0.28706789016723633,
       "step": 340
     },
     {
-      "epoch": 0.612691466083151,
-      "grad_norm": 0.6328125,
-      "learning_rate": 2.2839525327886768e-05,
-      "loss": 0.2882160425186157,
+      "epoch": 0.6635071090047393,
+      "grad_norm": 0.71875,
+      "learning_rate": 1.8647768529973425e-05,
+      "loss": 0.31264851093292234,
       "step": 350
     },
     {
-      "epoch": 0.6301969365426696,
-      "grad_norm": 0.98046875,
-      "learning_rate": 2.1187778866819505e-05,
-      "loss": 0.2965713024139404,
+      "epoch": 0.6824644549763034,
+      "grad_norm": 0.89453125,
+      "learning_rate": 1.6891648690025718e-05,
+      "loss": 0.2824721813201904,
       "step": 360
     },
     {
-      "epoch": 0.6477024070021882,
-      "grad_norm": 0.859375,
-      "learning_rate": 1.955291477890399e-05,
-      "loss": 0.2957930564880371,
+      "epoch": 0.7014218009478673,
+      "grad_norm": 0.84375,
+      "learning_rate": 1.5179195488936504e-05,
+      "loss": 0.2849812269210815,
       "step": 370
     },
     {
-      "epoch": 0.6652078774617067,
-      "grad_norm": 0.83203125,
-      "learning_rate": 1.7942173039110305e-05,
-      "loss": 0.299970006942749,
+      "epoch": 0.7033175355450237,
+      "eval_text_loss": 1.1984789371490479,
+      "eval_text_model_preparation_time": 0.0155,
+      "eval_text_runtime": 4.7136,
+      "eval_text_samples_per_second": 12.729,
+      "eval_text_steps_per_second": 12.729,
+      "step": 371
+    },
+    {
+      "epoch": 0.7033175355450237,
+      "eval_audio_loss": 0.719244122505188,
+      "eval_audio_model_preparation_time": 0.0155,
+      "eval_audio_runtime": 19.3073,
+      "eval_audio_samples_per_second": 6.889,
+      "eval_audio_steps_per_second": 6.889,
+      "step": 371
+    },
+    {
+      "epoch": 0.7203791469194313,
+      "grad_norm": 0.671875,
+      "learning_rate": 1.3519631156063539e-05,
+      "loss": 0.30515313148498535,
       "step": 380
     },
     {
-      "epoch": 0.6827133479212254,
-      "grad_norm": 0.70703125,
-      "learning_rate": 1.6362686796894435e-05,
-      "loss": 0.2794990062713623,
+      "epoch": 0.7393364928909952,
+      "grad_norm": 0.9296875,
+      "learning_rate": 1.1921893093551322e-05,
+      "loss": 0.298003101348877,
       "step": 390
     },
     {
-      "epoch": 0.700218818380744,
-      "grad_norm": 0.48046875,
-      "learning_rate": 1.4821450787136037e-05,
-      "loss": 0.28992691040039065,
+      "epoch": 0.7582938388625592,
+      "grad_norm": 0.57421875,
+      "learning_rate": 1.0394585744929605e-05,
+      "loss": 0.29361729621887206,
       "step": 400
     },
     {
-      "epoch": 0.7177242888402626,
-      "grad_norm": 0.640625,
-      "learning_rate": 1.332529035404327e-05,
-      "loss": 0.28732025623321533,
+      "epoch": 0.7772511848341233,
+      "grad_norm": 0.76171875,
+      "learning_rate": 8.945934256824218e-06,
+      "loss": 0.28779690265655516,
       "step": 410
     },
     {
-      "epoch": 0.7352297592997812,
-      "grad_norm": 0.828125,
-      "learning_rate": 1.1880831225202247e-05,
-      "loss": 0.2870940685272217,
+      "epoch": 0.7962085308056872,
+      "grad_norm": 0.7421875,
+      "learning_rate": 7.583740183329768e-06,
+      "loss": 0.27545139789581297,
       "step": 420
     },
     {
-      "epoch": 0.750984682713348,
-      "eval_text_loss": 1.21077561378479,
-      "eval_text_model_preparation_time": 0.0138,
-      "eval_text_runtime": 5.0709,
-      "eval_text_samples_per_second": 11.832,
-      "eval_text_steps_per_second": 11.832,
-      "step": 429
+      "epoch": 0.8037914691943128,
+      "eval_text_loss": 1.195946455001831,
+      "eval_text_model_preparation_time": 0.0155,
+      "eval_text_runtime": 5.0743,
+      "eval_text_samples_per_second": 11.824,
+      "eval_text_steps_per_second": 11.824,
+      "step": 424
     },
     {
-      "epoch": 0.750984682713348,
-      "eval_audio_loss": 0.698564350605011,
-      "eval_audio_model_preparation_time": 0.0138,
-      "eval_audio_runtime": 22.6597,
-      "eval_audio_samples_per_second": 5.869,
-      "eval_audio_steps_per_second": 5.869,
-      "step": 429
+      "epoch": 0.8037914691943128,
+      "eval_audio_loss": 0.7187640070915222,
+      "eval_audio_model_preparation_time": 0.0155,
+      "eval_audio_runtime": 20.0259,
+      "eval_audio_samples_per_second": 6.641,
+      "eval_audio_steps_per_second": 6.641,
+      "step": 424
     },
     {
-      "epoch": 0.7527352297592997,
-      "grad_norm": 0.78125,
-      "learning_rate": 1.049447016962597e-05,
-      "loss": 0.29113962650299074,
+      "epoch": 0.8151658767772512,
+      "grad_norm": 0.9375,
+      "learning_rate": 6.315339471593646e-06,
+      "loss": 0.2728050947189331,
       "step": 430
     },
     {
-      "epoch": 0.7702407002188184,
-      "grad_norm": 0.63671875,
-      "learning_rate": 9.172346669743626e-06,
-      "loss": 0.27481784820556643,
+      "epoch": 0.8341232227488151,
+      "grad_norm": 0.7421875,
+      "learning_rate": 5.147562954875268e-06,
+      "loss": 0.3111764907836914,
       "step": 440
     },
     {
-      "epoch": 0.787746170678337,
-      "grad_norm": 0.81640625,
-      "learning_rate": 7.920315732780126e-06,
-      "loss": 0.2930293560028076,
+      "epoch": 0.8530805687203792,
+      "grad_norm": 0.92578125,
+      "learning_rate": 4.086699565840915e-06,
+      "loss": 0.26120471954345703,
       "step": 450
     },
     {
-      "epoch": 0.8052516411378556,
-      "grad_norm": 0.578125,
-      "learning_rate": 6.743921961930549e-06,
-      "loss": 0.28920702934265136,
+      "epoch": 0.8720379146919431,
+      "grad_norm": 0.8125,
+      "learning_rate": 3.138462468204914e-06,
+      "loss": 0.2936476469039917,
       "step": 460
     },
     {
-      "epoch": 0.8227571115973742,
-      "grad_norm": 0.71875,
-      "learning_rate": 5.648375002154879e-06,
-      "loss": 0.28770692348480226,
+      "epoch": 0.8909952606635071,
+      "grad_norm": 0.66796875,
+      "learning_rate": 2.3079582891115144e-06,
+      "loss": 0.2918030023574829,
       "step": 470
     },
     {
-      "epoch": 0.8402625820568927,
-      "grad_norm": 0.56640625,
-      "learning_rate": 4.6385264693311305e-06,
-      "loss": 0.2829106330871582,
+      "epoch": 0.9042654028436019,
+      "eval_text_loss": 1.1954913139343262,
+      "eval_text_model_preparation_time": 0.0155,
+      "eval_text_runtime": 5.6311,
+      "eval_text_samples_per_second": 10.655,
+      "eval_text_steps_per_second": 10.655,
+      "step": 477
+    },
+    {
+      "epoch": 0.9042654028436019,
+      "eval_audio_loss": 0.7186709642410278,
+      "eval_audio_model_preparation_time": 0.0155,
+      "eval_audio_runtime": 21.3981,
+      "eval_audio_samples_per_second": 6.216,
+      "eval_audio_steps_per_second": 6.216,
+      "step": 477
+    },
+    {
+      "epoch": 0.909952606635071,
+      "grad_norm": 0.52734375,
+      "learning_rate": 1.5996596179531364e-06,
+      "loss": 0.29052374362945554,
       "step": 480
     },
     {
-      "epoch": 0.8577680525164114,
-      "grad_norm": 0.7421875,
-      "learning_rate": 3.7188484649356562e-06,
-      "loss": 0.29122676849365237,
+      "epoch": 0.9289099526066351,
+      "grad_norm": 0.625,
+      "learning_rate": 1.0173809197295075e-06,
+      "loss": 0.27139732837677,
       "step": 490
     },
     {
-      "epoch": 0.87527352297593,
-      "grad_norm": 0.8046875,
-      "learning_rate": 2.8934137713982454e-06,
-      "loss": 0.29016549587249757,
+      "epoch": 0.9478672985781991,
+      "grad_norm": 0.67578125,
+      "learning_rate": 5.642579926637554e-07,
+      "loss": 0.29912357330322265,
       "step": 500
     },
     {
-      "epoch": 0.8927789934354485,
-      "grad_norm": 0.87890625,
-      "learning_rate": 2.1658778158365284e-06,
-      "loss": 0.29710874557495115,
+      "epoch": 0.966824644549763,
+      "grad_norm": 1.171875,
+      "learning_rate": 2.4273108070415607e-07,
+      "loss": 0.2772815465927124,
       "step": 510
     },
     {
-      "epoch": 0.9102844638949672,
-      "grad_norm": 1.0234375,
-      "learning_rate": 1.5394624820432079e-06,
-      "loss": 0.2932096004486084,
+      "epoch": 0.985781990521327,
+      "grad_norm": 0.60546875,
+      "learning_rate": 5.453173185734073e-08,
+      "loss": 0.29728262424468993,
       "step": 520
     },
-    {
-      "epoch": 0.9277899343544858,
-      "grad_norm": 0.84375,
-      "learning_rate": 1.01694184241429e-06,
-      "loss": 0.28654632568359373,
-      "step": 530
-    },
-    {
-      "epoch": 0.9452954048140044,
-      "grad_norm": 0.57421875,
-      "learning_rate": 6.006298730045601e-07,
-      "loss": 0.2893019676208496,
-      "step": 540
-    },
-    {
-      "epoch": 0.962800875273523,
-      "grad_norm": 0.70703125,
-      "learning_rate": 2.923702061138206e-07,
-      "loss": 0.2885005235671997,
-      "step": 550
-    },
-    {
-      "epoch": 0.9803063457330415,
-      "grad_norm": 0.5546875,
-      "learning_rate": 9.352796578453172e-08,
-      "loss": 0.30120275020599363,
-      "step": 560
-    },
-    {
-      "epoch": 0.9978118161925602,
-      "grad_norm": 0.5078125,
-      "learning_rate": 4.983722367229993e-09,
-      "loss": 0.2859893083572388,
-      "step": 570
-    },
     {
       "epoch": 1.0,
-      "eval_text_loss": 1.2087676525115967,
-      "eval_text_model_preparation_time": 0.0138,
-      "eval_text_runtime": 5.4583,
-      "eval_text_samples_per_second": 10.992,
-      "eval_text_steps_per_second": 10.992,
-      "step": 572
+      "eval_text_loss": 1.196468710899353,
+      "eval_text_model_preparation_time": 0.0155,
+      "eval_text_runtime": 4.6934,
+      "eval_text_samples_per_second": 12.784,
+      "eval_text_steps_per_second": 12.784,
+      "step": 528
     },
     {
       "epoch": 1.0,
-      "eval_audio_loss": 0.6977604627609253,
-      "eval_audio_model_preparation_time": 0.0138,
-      "eval_audio_runtime": 22.5218,
-      "eval_audio_samples_per_second": 5.905,
-      "eval_audio_steps_per_second": 5.905,
-      "step": 572
+      "eval_audio_loss": 0.7178842425346375,
+      "eval_audio_model_preparation_time": 0.0155,
+      "eval_audio_runtime": 20.4901,
+      "eval_audio_samples_per_second": 6.491,
+      "eval_audio_steps_per_second": 6.491,
+      "step": 528
     }
   ],
   "logging_steps": 10,
-  "max_steps": 572,
+  "max_steps": 528,
   "num_input_tokens_seen": 0,
   "num_train_epochs": 1,
-  "save_steps": 12,
+  "save_steps": 11,
   "stateful_callbacks": {
     "TrainerControl": {
       "args": {
@@ -498,7 +571,7 @@
       "attributes": {}
     }
   },
-  "total_flos": 6.96093256353648e+16,
+  "total_flos": 5.645359034142144e+16,
   "train_batch_size": 4,
   "trial_name": null,
   "trial_params": null

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:6906c2dd0f60d842f35c319522ede5e9054fb9019c882a25267a2d08f135ce09
+oid sha256:db80b36a03b61070e73a039eaa22ef4f319158fa5d76b4a8fe091f4aaa789aa0
 size 5905