End of training 60 epochs, 6240 steps

README.md

@@ -0,0 +1,69 @@
+---
+license: apache-2.0
+tags:
+- automatic-speech-recognition
+- openslr_SLR66
+- robust-speech-event
+- generated_from_trainer
+model-index:
+- name: ''
+  results: []
+---
+
+<!-- This model card has been generated automatically according to the information the Trainer had access to. You
+should probably proofread and complete it, then remove this comment. -->
+
+# 
+
+This model is a fine-tuned version of [facebook/wav2vec2-xls-r-2b](https://huggingface.co/facebook/wav2vec2-xls-r-2b) on the OPENSLR_SLR66 - NA dataset.
+It achieves the following results on the evaluation set:
+- Loss: 0.9243
+- Wer: 0.9293
+
+## Model description
+
+More information needed
+
+## Intended uses & limitations
+
+More information needed
+
+## Training and evaluation data
+
+More information needed
+
+## Training procedure
+
+### Training hyperparameters
+
+The following hyperparameters were used during training:
+- learning_rate: 7.5e-05
+- train_batch_size: 4
+- eval_batch_size: 4
+- seed: 42
+- gradient_accumulation_steps: 8
+- total_train_batch_size: 32
+- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
+- lr_scheduler_type: linear
+- lr_scheduler_warmup_steps: 2000
+- num_epochs: 60.0
+- mixed_precision_training: Native AMP
+
+### Training results
+
+| Training Loss | Epoch | Step | Validation Loss | Wer    |
+|:-------------:|:-----:|:----:|:---------------:|:------:|
+| 1.8141        | 9.61  | 1000 | 0.7609          | 0.7778 |
+| 2.2935        | 19.23 | 2000 | 1.1849          | 0.9657 |
+| 2.2848        | 28.84 | 3000 | 1.1789          | 0.9760 |
+| 2.3325        | 38.46 | 4000 | 1.0878          | 0.9595 |
+| 2.25          | 48.08 | 5000 | 1.0888          | 0.9586 |
+| 2.0757        | 57.69 | 6000 | 0.9504          | 0.9412 |
+
+
+### Framework versions
+
+- Transformers 4.16.0.dev0
+- Pytorch 1.10.1+cu102
+- Datasets 1.17.1.dev0
+- Tokenizers 0.11.0

all_results.json

@@ -1,14 +1,14 @@
 {
-    "epoch": 1.0,
-    "eval_loss": 0.8520588874816895,
-    "eval_runtime": 81.349,
+    "epoch": 60.0,
+    "eval_loss": 0.9243341088294983,
+    "eval_runtime": 80.7101,
     "eval_samples": 1112,
-    "eval_samples_per_second": 13.669,
-    "eval_steps_per_second": 3.417,
-    "eval_wer": 0.8527354260089686,
-    "train_loss": 5.8410720229148865,
-    "train_runtime": 1235.9298,
+    "eval_samples_per_second": 13.778,
+    "eval_steps_per_second": 3.444,
+    "eval_wer": 0.9293273542600897,
+    "train_loss": 2.2000570346147588,
+    "train_runtime": 38994.8678,
     "train_samples": 3336,
-    "train_samples_per_second": 2.699,
-    "train_steps_per_second": 0.084
+    "train_samples_per_second": 5.133,
+    "train_steps_per_second": 0.16
 }

eval_results.json

@@ -1,9 +1,9 @@
 {
-    "epoch": 1.0,
-    "eval_loss": 0.8520588874816895,
-    "eval_runtime": 81.349,
+    "epoch": 60.0,
+    "eval_loss": 0.9243341088294983,
+    "eval_runtime": 80.7101,
     "eval_samples": 1112,
-    "eval_samples_per_second": 13.669,
-    "eval_steps_per_second": 3.417,
-    "eval_wer": 0.8527354260089686
+    "eval_samples_per_second": 13.778,
+    "eval_steps_per_second": 3.444,
+    "eval_wer": 0.9293273542600897
 }

pytorch_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:7c53f4f0f2c0704e10cacf5b54e76302eae712ce8738043f489b591ec0b4a502
+oid sha256:6447b9bfe1ae69ddefa68b10c300469609be3bdc873e1d729bcd1461962117a9
 size 8637973773

run_bnb.sh

@@ -5,7 +5,7 @@ python run_speech_recognition_ctc_bnb.py \
 --preprocessing_num_workers="8" \
 --output_dir="./" \
 --overwrite_output_dir \
---num_train_epochs="1" \
+--num_train_epochs="60" \
 --per_device_train_batch_size="4" \
 --per_device_eval_batch_size="4" \
 --gradient_accumulation_steps="8" \
@@ -15,8 +15,8 @@ python run_speech_recognition_ctc_bnb.py \
 --evaluation_strategy="steps" \
 --text_column_name="sentence" \
 --chars_to_ignore , ? . ! \- \; \: \" “ % ‘ ” � — ’ … – \
---save_steps="100" \
---eval_steps="100" \
+--save_steps="1000" \
+--eval_steps="1000" \
 --logging_steps="100" \
 --layerdrop="0.0" \
 --activation_dropout="0.1" \
@@ -31,5 +31,4 @@ python run_speech_recognition_ctc_bnb.py \
 --use_auth_token \
 --fp16 \
 --group_by_length \
---do_train --do_eval \
---push_to_hub
+--do_train --do_eval

special_tokens_map.json

@@ -1 +1 @@
-{"bos_token": "<s>", "eos_token": "</s>", "unk_token": "[UNK]", "pad_token": "[PAD]", "additional_special_tokens": [{"content": "<s>", "single_word": false, "lstrip": false, "rstrip": false, "normalized": true}, {"content": "</s>", "single_word": false, "lstrip": false, "rstrip": false, "normalized": true}, {"content": "<s>", "single_word": false, "lstrip": false, "rstrip": false, "normalized": true}, {"content": "</s>", "single_word": false, "lstrip": false, "rstrip": false, "normalized": true}, {"content": "<s>", "single_word": false, "lstrip": false, "rstrip": false, "normalized": true}, {"content": "</s>", "single_word": false, "lstrip": false, "rstrip": false, "normalized": true}, {"content": "<s>", "single_word": false, "lstrip": false, "rstrip": false, "normalized": true}, {"content": "</s>", "single_word": false, "lstrip": false, "rstrip": false, "normalized": true}, {"content": "<s>", "single_word": false, "lstrip": false, "rstrip": false, "normalized": true}, {"content": "</s>", "single_word": false, "lstrip": false, "rstrip": false, "normalized": true}]}
+{"bos_token": "<s>", "eos_token": "</s>", "unk_token": "[UNK]", "pad_token": "[PAD]", "additional_special_tokens": [{"content": "<s>", "single_word": false, "lstrip": false, "rstrip": false, "normalized": true}, {"content": "</s>", "single_word": false, "lstrip": false, "rstrip": false, "normalized": true}, {"content": "<s>", "single_word": false, "lstrip": false, "rstrip": false, "normalized": true}, {"content": "</s>", "single_word": false, "lstrip": false, "rstrip": false, "normalized": true}, {"content": "<s>", "single_word": false, "lstrip": false, "rstrip": false, "normalized": true}, {"content": "</s>", "single_word": false, "lstrip": false, "rstrip": false, "normalized": true}, {"content": "<s>", "single_word": false, "lstrip": false, "rstrip": false, "normalized": true}, {"content": "</s>", "single_word": false, "lstrip": false, "rstrip": false, "normalized": true}, {"content": "<s>", "single_word": false, "lstrip": false, "rstrip": false, "normalized": true}, {"content": "</s>", "single_word": false, "lstrip": false, "rstrip": false, "normalized": true}, {"content": "<s>", "single_word": false, "lstrip": false, "rstrip": false, "normalized": true}, {"content": "</s>", "single_word": false, "lstrip": false, "rstrip": false, "normalized": true}, {"content": "<s>", "single_word": false, "lstrip": false, "rstrip": false, "normalized": true}, {"content": "</s>", "single_word": false, "lstrip": false, "rstrip": false, "normalized": true}]}

train_results.json

@@ -1,8 +1,8 @@
 {
-    "epoch": 1.0,
-    "train_loss": 5.8410720229148865,
-    "train_runtime": 1235.9298,
+    "epoch": 60.0,
+    "train_loss": 2.2000570346147588,
+    "train_runtime": 38994.8678,
     "train_samples": 3336,
-    "train_samples_per_second": 2.699,
-    "train_steps_per_second": 0.084
+    "train_samples_per_second": 5.133,
+    "train_steps_per_second": 0.16
 }

trainer_state.json

@@ -1,40 +1,451 @@
 {
   "best_metric": null,
   "best_model_checkpoint": null,
-  "epoch": 0.9976019184652278,
-  "global_step": 104,
+  "epoch": 59.99760191846523,
+  "global_step": 6240,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
   "log_history": [
     {
       "epoch": 0.96,
-      "learning_rate": 4.9000000000000005e-05,
-      "loss": 6.0022,
+      "learning_rate": 4.8e-05,
+      "loss": 6.3756,
       "step": 100
     },
     {
-      "epoch": 0.96,
-      "eval_loss": 0.8687445521354675,
-      "eval_runtime": 82.3614,
-      "eval_samples_per_second": 13.501,
-      "eval_steps_per_second": 3.375,
-      "eval_wer": 0.8968609865470852,
-      "step": 100
+      "epoch": 1.92,
+      "learning_rate": 9.800000000000001e-05,
+      "loss": 1.5856,
+      "step": 200
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 0.000148,
+      "loss": 1.5323,
+      "step": 300
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 0.00019800000000000002,
+      "loss": 1.5582,
+      "step": 400
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 0.000248,
+      "loss": 1.6612,
+      "step": 500
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 0.000298,
+      "loss": 1.6766,
+      "step": 600
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 0.000348,
+      "loss": 1.7327,
+      "step": 700
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 0.000398,
+      "loss": 1.7573,
+      "step": 800
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 0.000448,
+      "loss": 1.7926,
+      "step": 900
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 0.000498,
+      "loss": 1.8141,
+      "step": 1000
+    },
+    {
+      "epoch": 9.61,
+      "eval_loss": 0.7609468102455139,
+      "eval_runtime": 79.8366,
+      "eval_samples_per_second": 13.928,
+      "eval_steps_per_second": 3.482,
+      "eval_wer": 0.7777578475336323,
+      "step": 1000
+    },
+    {
+      "epoch": 10.58,
+      "learning_rate": 0.0005480000000000001,
+      "loss": 1.8648,
+      "step": 1100
+    },
+    {
+      "epoch": 11.54,
+      "learning_rate": 0.000598,
+      "loss": 1.9241,
+      "step": 1200
+    },
+    {
+      "epoch": 12.5,
+      "learning_rate": 0.000648,
+      "loss": 1.9235,
+      "step": 1300
+    },
+    {
+      "epoch": 13.46,
+      "learning_rate": 0.0006979999999999999,
+      "loss": 1.9089,
+      "step": 1400
+    },
+    {
+      "epoch": 14.42,
+      "learning_rate": 0.000748,
+      "loss": 1.9364,
+      "step": 1500
+    },
+    {
+      "epoch": 15.38,
+      "learning_rate": 0.0007980000000000001,
+      "loss": 2.0406,
+      "step": 1600
+    },
+    {
+      "epoch": 16.35,
+      "learning_rate": 0.000848,
+      "loss": 2.1109,
+      "step": 1700
+    },
+    {
+      "epoch": 17.31,
+      "learning_rate": 0.000898,
+      "loss": 2.2393,
+      "step": 1800
+    },
+    {
+      "epoch": 18.27,
+      "learning_rate": 0.000948,
+      "loss": 2.2755,
+      "step": 1900
+    },
+    {
+      "epoch": 19.23,
+      "learning_rate": 0.000998,
+      "loss": 2.2935,
+      "step": 2000
+    },
+    {
+      "epoch": 19.23,
+      "eval_loss": 1.1848976612091064,
+      "eval_runtime": 80.3457,
+      "eval_samples_per_second": 13.84,
+      "eval_steps_per_second": 3.46,
+      "eval_wer": 0.9657399103139014,
+      "step": 2000
+    },
+    {
+      "epoch": 20.19,
+      "learning_rate": 0.0009773584905660377,
+      "loss": 2.3135,
+      "step": 2100
+    },
+    {
+      "epoch": 21.15,
+      "learning_rate": 0.0009537735849056604,
+      "loss": 2.2768,
+      "step": 2200
+    },
+    {
+      "epoch": 22.12,
+      "learning_rate": 0.000930188679245283,
+      "loss": 2.2858,
+      "step": 2300
+    },
+    {
+      "epoch": 23.08,
+      "learning_rate": 0.0009066037735849056,
+      "loss": 2.2804,
+      "step": 2400
+    },
+    {
+      "epoch": 24.04,
+      "learning_rate": 0.0008830188679245283,
+      "loss": 2.2852,
+      "step": 2500
+    },
+    {
+      "epoch": 25.0,
+      "learning_rate": 0.0008594339622641509,
+      "loss": 2.2629,
+      "step": 2600
+    },
+    {
+      "epoch": 25.96,
+      "learning_rate": 0.0008358490566037736,
+      "loss": 2.3056,
+      "step": 2700
+    },
+    {
+      "epoch": 26.92,
+      "learning_rate": 0.0008122641509433963,
+      "loss": 2.2643,
+      "step": 2800
+    },
+    {
+      "epoch": 27.88,
+      "learning_rate": 0.000788679245283019,
+      "loss": 2.3471,
+      "step": 2900
+    },
+    {
+      "epoch": 28.84,
+      "learning_rate": 0.0007650943396226416,
+      "loss": 2.2848,
+      "step": 3000
+    },
+    {
+      "epoch": 28.84,
+      "eval_loss": 1.178926944732666,
+      "eval_runtime": 80.2197,
+      "eval_samples_per_second": 13.862,
+      "eval_steps_per_second": 3.465,
+      "eval_wer": 0.9759641255605381,
+      "step": 3000
+    },
+    {
+      "epoch": 29.81,
+      "learning_rate": 0.0007415094339622641,
+      "loss": 2.3196,
+      "step": 3100
+    },
+    {
+      "epoch": 30.77,
+      "learning_rate": 0.0007179245283018868,
+      "loss": 2.3493,
+      "step": 3200
+    },
+    {
+      "epoch": 31.73,
+      "learning_rate": 0.0006943396226415094,
+      "loss": 2.4006,
+      "step": 3300
+    },
+    {
+      "epoch": 32.69,
+      "learning_rate": 0.0006707547169811321,
+      "loss": 2.3034,
+      "step": 3400
+    },
+    {
+      "epoch": 33.65,
+      "learning_rate": 0.0006474056603773585,
+      "loss": 2.3289,
+      "step": 3500
+    },
+    {
+      "epoch": 34.61,
+      "learning_rate": 0.0006240566037735848,
+      "loss": 2.2975,
+      "step": 3600
+    },
+    {
+      "epoch": 35.58,
+      "learning_rate": 0.0006004716981132075,
+      "loss": 2.2996,
+      "step": 3700
+    },
+    {
+      "epoch": 36.54,
+      "learning_rate": 0.0005768867924528301,
+      "loss": 2.327,
+      "step": 3800
+    },
+    {
+      "epoch": 37.5,
+      "learning_rate": 0.0005533018867924528,
+      "loss": 2.3041,
+      "step": 3900
+    },
+    {
+      "epoch": 38.46,
+      "learning_rate": 0.0005297169811320755,
+      "loss": 2.3325,
+      "step": 4000
+    },
+    {
+      "epoch": 38.46,
+      "eval_loss": 1.087767481803894,
+      "eval_runtime": 79.9787,
+      "eval_samples_per_second": 13.904,
+      "eval_steps_per_second": 3.476,
+      "eval_wer": 0.9594618834080717,
+      "step": 4000
+    },
+    {
+      "epoch": 39.42,
+      "learning_rate": 0.000506367924528302,
+      "loss": 2.3006,
+      "step": 4100
+    },
+    {
+      "epoch": 40.38,
+      "learning_rate": 0.0004827830188679245,
+      "loss": 2.3251,
+      "step": 4200
+    },
+    {
+      "epoch": 41.35,
+      "learning_rate": 0.0004591981132075472,
+      "loss": 2.2754,
+      "step": 4300
+    },
+    {
+      "epoch": 42.31,
+      "learning_rate": 0.00043561320754716983,
+      "loss": 2.2903,
+      "step": 4400
+    },
+    {
+      "epoch": 43.27,
+      "learning_rate": 0.00041202830188679247,
+      "loss": 2.2868,
+      "step": 4500
+    },
+    {
+      "epoch": 44.23,
+      "learning_rate": 0.0003884433962264151,
+      "loss": 2.2541,
+      "step": 4600
+    },
+    {
+      "epoch": 45.19,
+      "learning_rate": 0.00036485849056603774,
+      "loss": 2.2362,
+      "step": 4700
+    },
+    {
+      "epoch": 46.15,
+      "learning_rate": 0.00034127358490566043,
+      "loss": 2.2357,
+      "step": 4800
+    },
+    {
+      "epoch": 47.12,
+      "learning_rate": 0.00031768867924528306,
+      "loss": 2.2935,
+      "step": 4900
+    },
+    {
+      "epoch": 48.08,
+      "learning_rate": 0.00029410377358490565,
+      "loss": 2.25,
+      "step": 5000
+    },
+    {
+      "epoch": 48.08,
+      "eval_loss": 1.088808536529541,
+      "eval_runtime": 80.2673,
+      "eval_samples_per_second": 13.854,
+      "eval_steps_per_second": 3.463,
+      "eval_wer": 0.9585650224215246,
+      "step": 5000
+    },
+    {
+      "epoch": 49.04,
+      "learning_rate": 0.0002705188679245283,
+      "loss": 2.2023,
+      "step": 5100
+    },
+    {
+      "epoch": 50.0,
+      "learning_rate": 0.0002469339622641509,
+      "loss": 2.2132,
+      "step": 5200
+    },
+    {
+      "epoch": 50.96,
+      "learning_rate": 0.00022334905660377358,
+      "loss": 2.1842,
+      "step": 5300
+    },
+    {
+      "epoch": 51.92,
+      "learning_rate": 0.00019976415094339622,
+      "loss": 2.1644,
+      "step": 5400
+    },
+    {
+      "epoch": 52.88,
+      "learning_rate": 0.00017617924528301888,
+      "loss": 2.1449,
+      "step": 5500
+    },
+    {
+      "epoch": 53.84,
+      "learning_rate": 0.00015259433962264152,
+      "loss": 2.1539,
+      "step": 5600
+    },
+    {
+      "epoch": 54.81,
+      "learning_rate": 0.00012900943396226418,
+      "loss": 2.1193,
+      "step": 5700
+    },
+    {
+      "epoch": 55.77,
+      "learning_rate": 0.0001054245283018868,
+      "loss": 2.091,
+      "step": 5800
+    },
+    {
+      "epoch": 56.73,
+      "learning_rate": 8.183962264150944e-05,
+      "loss": 2.0871,
+      "step": 5900
+    },
+    {
+      "epoch": 57.69,
+      "learning_rate": 5.825471698113208e-05,
+      "loss": 2.0757,
+      "step": 6000
+    },
+    {
+      "epoch": 57.69,
+      "eval_loss": 0.950368344783783,
+      "eval_runtime": 79.8206,
+      "eval_samples_per_second": 13.931,
+      "eval_steps_per_second": 3.483,
+      "eval_wer": 0.9411659192825113,
+      "step": 6000
+    },
+    {
+      "epoch": 58.65,
+      "learning_rate": 3.4669811320754716e-05,
+      "loss": 2.056,
+      "step": 6100
+    },
+    {
+      "epoch": 59.61,
+      "learning_rate": 1.1084905660377358e-05,
+      "loss": 2.0488,
+      "step": 6200
     },
     {
-      "epoch": 1.0,
-      "step": 104,
-      "total_flos": 3.20825392163127e+18,
-      "train_loss": 5.8410720229148865,
-      "train_runtime": 1235.9298,
-      "train_samples_per_second": 2.699,
-      "train_steps_per_second": 0.084
+      "epoch": 60.0,
+      "step": 6240,
+      "total_flos": 1.9322773224714512e+20,
+      "train_loss": 2.2000570346147588,
+      "train_runtime": 38994.8678,
+      "train_samples_per_second": 5.133,
+      "train_steps_per_second": 0.16
     }
   ],
-  "max_steps": 104,
-  "num_train_epochs": 1,
-  "total_flos": 3.20825392163127e+18,
+  "max_steps": 6240,
+  "num_train_epochs": 60,
+  "total_flos": 1.9322773224714512e+20,
   "trial_name": null,
   "trial_params": null
 }

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:7967a25f5d5506dbbe9b4e4e4efee7fc2ebcba52c0602cf5e0ce4cf99f7ec2b4
+oid sha256:e671975c9fc29d251830e4bd90003705e709ea3d562726308fb5fc129d6750fe
 size 2991