test

README.md

@@ -1,8 +1,6 @@
 language: sw  
 datasets:
-- OpenSLR - http://www.openslr.org/25/
-- TODO: add more datasets if you have used additional datasets. Make sure to use the exact same 
-dataset name as the one found [here](https://huggingface.co/datasets). If the dataset can not be found in the official datasets, just give it a new name
+- ALFFA (African Languages in the Field: speech Fundamentals and Automation) - [here](http://www.openslr.org/25/)
 metrics:
 - wer
 tags:
@@ -12,24 +10,23 @@ tags:
 - xlsr-fine-tuning-week
 license: apache-2.0
 model-index:
-- name: Swahili XLSR53 Wav2Vec2 Large
+- name: Swahili XLSR-53 Wav2Vec2.0 Large
   results:
   - task: 
       name: Speech Recognition
       type: automatic-speech-recognition
     dataset:
-      name: Common Voice {lang_id} #TODO: replace {lang_id} in your language code here. Make sure the code is one of the *ISO codes* of [this](https://huggingface.co/languages) site.
-      type: common_voice
-      args: {lang_id} #TODO: replace {lang_id} in your language code here. Make sure the code is one of the *ISO codes* of [this](https://huggingface.co/languages) site.
+      name: ALFFA sw
+      args: sw
     metrics:
        - name: Test WER
          type: wer
-         value: {wer_result_on_test} #TODO (IMPORTANT): replace {wer_result_on_test} with the WER error rate you achieved on the common_voice test set. It should be in the format XX.XX (don't add the % sign here). **Please** remember to fill out this value after you evaluated your model, so that your model appears on the leaderboard. If you fill out this model card before evaluating your model, please remember to edit the model card afterward to fill in your value
+         value: WIP
 ---
 
-# Wav2Vec2-Large-XLSR-53-{language} #TODO: replace language with your {language}, *e.g.* French
+# Wav2Vec2-Large-XLSR-53-{Swahili} 
 
-Fine-tuned [facebook/wav2vec2-large-xlsr-53](https://huggingface.co/facebook/wav2vec2-large-xlsr-53) on {language} using the [Common Voice](https://huggingface.co/datasets/common_voice), ... and ... dataset{s}. #TODO: replace {language} with your language, *e.g.* French and eventually add more datasets that were used and eventually remove common voice if model was not trained on common voice
+Fine-tuned [facebook/wav2vec2-large-xlsr-53](https://huggingface.co/facebook/wav2vec2-large-xlsr-53) on Swahili using the [ALFFA](http://www.openslr.org/25/), ... and ... dataset{s}.
 When using this model, make sure that your speech input is sampled at 16kHz.
 
 ## Usage
@@ -42,87 +39,39 @@ import torchaudio
 from datasets import load_dataset
 from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor
 
-test_dataset = load_dataset("common_voice", "{lang_id}", split="test[:2%]") #TODO: replace {lang_id} in your language code here. Make sure the code is one of the *ISO codes* of [this](https://huggingface.co/languages) site.
 
-processor = Wav2Vec2Processor.from_pretrained("{model_id}") #TODO: replace {model_id} with your model id. The model id consists of {your_username}/{your_modelname}, *e.g.* `elgeish/wav2vec2-large-xlsr-53-arabic`
-model = Wav2Vec2ForCTC.from_pretrained("{model_id}") #TODO: replace {model_id} with your model id. The model id consists of {your_username}/{your_modelname}, *e.g.* `elgeish/wav2vec2-large-xlsr-53-arabic`
+processor = Wav2Vec2Processor.from_pretrained("alokmatta/wav2vec2-large-xlsr-53-sw")
 
-resampler = torchaudio.transforms.Resample(48_000, 16_000)
-
-# Preprocessing the datasets.
-# We need to read the aduio files as arrays
-def speech_file_to_array_fn(batch):
-	speech_array, sampling_rate = torchaudio.load(batch["path"])
-	batch["speech"] = resampler(speech_array).squeeze().numpy()
-	return batch
-
-test_dataset = test_dataset.map(speech_file_to_array_fn)
-inputs = processor(test_dataset["speech"][:2], sampling_rate=16_000, return_tensors="pt", padding=True)
-
-with torch.no_grad():
-	logits = model(inputs.input_values, attention_mask=inputs.attention_mask).logits
-
-predicted_ids = torch.argmax(logits, dim=-1)
-
-print("Prediction:", processor.batch_decode(predicted_ids))
-print("Reference:", test_dataset["sentence"][:2])
-```
-
-
-## Evaluation
-
-The model can be evaluated as follows on the {language} test data of Common Voice.  # TODO: replace #TODO: replace language with your {language}, *e.g.* French
+model = Wav2Vec2ForCTC.from_pretrained("alokmatta/wav2vec2-large-xlsr-53-sw").to("cuda")
 
-
-```python
-import torch
-import torchaudio
-from datasets import load_dataset, load_metric
-from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor
-import re
-
-test_dataset = load_dataset("common_voice", "{lang_id}", split="test") #TODO: replace {lang_id} in your language code here. Make sure the code is one of the *ISO codes* of [this](https://huggingface.co/languages) site.
-wer = load_metric("wer")
-
-processor = Wav2Vec2Processor.from_pretrained("{model_id}") #TODO: replace {model_id} with your model id. The model id consists of {your_username}/{your_modelname}, *e.g.* `elgeish/wav2vec2-large-xlsr-53-arabic`
-model = Wav2Vec2ForCTC.from_pretrained("{model_id}") #TODO: replace {model_id} with your model id. The model id consists of {your_username}/{your_modelname}, *e.g.* `elgeish/wav2vec2-large-xlsr-53-arabic`
-model.to("cuda")
-
-chars_to_ignore_regex = '[\,\?\.\!\-\;\:\"\“]'  # TODO: adapt this list to include all special characters you removed from the data
 resampler = torchaudio.transforms.Resample(48_000, 16_000)
 
-# Preprocessing the datasets.
-# We need to read the aduio files as arrays
-def speech_file_to_array_fn(batch):
-	batch["sentence"] = re.sub(chars_to_ignore_regex, '', batch["sentence"]).lower()
-	speech_array, sampling_rate = torchaudio.load(batch["path"])
-	batch["speech"] = resampler(speech_array).squeeze().numpy()
-	return batch
-
-test_dataset = test_dataset.map(speech_file_to_array_fn)
-
-# Preprocessing the datasets.
-# We need to read the aduio files as arrays
-def evaluate(batch):
-	inputs = processor(batch["speech"], sampling_rate=16_000, return_tensors="pt", padding=True)
+resampler = torchaudio.transforms.Resample(orig_freq=48_000, new_freq=16_000)
 
-	with torch.no_grad():
-		logits = model(inputs.input_values.to("cuda"), attention_mask=inputs.attention_mask.to("cuda")).logits
+def load_file_to_data(file):
+    batch = {}
+    speech, _ = torchaudio.load(file)
+    batch["speech"] = resampler.forward(speech.squeeze(0)).numpy()
+    batch["sampling_rate"] = resampler.new_freq
+    return batch
 
-	pred_ids = torch.argmax(logits, dim=-1)
-	batch["pred_strings"] = processor.batch_decode(pred_ids)
-	return batch
 
-result = test_dataset.map(evaluate, batched=True, batch_size=8)
+def predict(data):
+    features = processor(data["speech"], sampling_rate=data["sampling_rate"], padding=True, return_tensors="pt")
+    input_values = features.input_values.to("cuda")
+    attention_mask = features.attention_mask.to("cuda")
+    with torch.no_grad():
+        logits = model(input_values, attention_mask=attention_mask).logits
+    pred_ids = torch.argmax(logits, dim=-1)
+    return processor.batch_decode(pred_ids)
 
-print("WER: {:2f}".format(100 * wer.compute(predictions=result["pred_strings"], references=result["sentence"])))
+predict(load_file_to_data('./demo.wav'))
 ```
 
-**Test Result**: XX.XX %  # TODO: write output of print here. IMPORTANT: Please remember to also replace {wer_result_on_test} at the top of with this value here. tags.
+**Test Result**: WIP % 
 
 
 ## Training
 
-The Common Voice `train`, `validation`, and ... datasets were used for training as well as ... and ...  # TODO: adapt to state all the datasets that were used for training.
 
-The script used for training can be found [here](...) # TODO: fill in a link to your training script here. If you trained your model in a colab, simply fill in the link here. If you trained the model locally, it would be great if you could upload the training script on github and paste the link here.
+The script used for training can be found Here- Coming Soon!

config.json

@@ -0,0 +1,76 @@
+{
+  "_name_or_path": "facebook/wav2vec2-large-xlsr-53",
+  "activation_dropout": 0.0,
+  "apply_spec_augment": true,
+  "architectures": [
+    "Wav2Vec2ForCTC"
+  ],
+  "attention_dropout": 0.1,
+  "bos_token_id": 1,
+  "conv_bias": true,
+  "conv_dim": [
+    512,
+    512,
+    512,
+    512,
+    512,
+    512,
+    512
+  ],
+  "conv_kernel": [
+    10,
+    3,
+    3,
+    3,
+    3,
+    2,
+    2
+  ],
+  "conv_stride": [
+    5,
+    2,
+    2,
+    2,
+    2,
+    2,
+    2
+  ],
+  "ctc_loss_reduction": "mean",
+  "ctc_zero_infinity": false,
+  "do_stable_layer_norm": true,
+  "eos_token_id": 2,
+  "feat_extract_activation": "gelu",
+  "feat_extract_dropout": 0.0,
+  "feat_extract_norm": "layer",
+  "feat_proj_dropout": 0.0,
+  "final_dropout": 0.0,
+  "gradient_checkpointing": true,
+  "hidden_act": "gelu",
+  "hidden_dropout": 0.1,
+  "hidden_size": 1024,
+  "initializer_range": 0.02,
+  "intermediate_size": 4096,
+  "layer_norm_eps": 1e-05,
+  "layerdrop": 0.1,
+  "mask_channel_length": 10,
+  "mask_channel_min_space": 1,
+  "mask_channel_other": 0.0,
+  "mask_channel_prob": 0.0,
+  "mask_channel_selection": "static",
+  "mask_feature_length": 10,
+  "mask_feature_prob": 0.0,
+  "mask_time_length": 10,
+  "mask_time_min_space": 1,
+  "mask_time_other": 0.0,
+  "mask_time_prob": 0.05,
+  "mask_time_selection": "static",
+  "model_type": "wav2vec2",
+  "num_attention_heads": 16,
+  "num_conv_pos_embedding_groups": 16,
+  "num_conv_pos_embeddings": 128,
+  "num_feat_extract_layers": 7,
+  "num_hidden_layers": 24,
+  "pad_token_id": 40,
+  "transformers_version": "4.5.0.dev0",
+  "vocab_size": 41
+}

optimizer.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ce45012be135497e4b16ab8654d1a24b97bb14be98b5fabf07fdcff635dcf3e0
+size 1711

preprocessor_config.json

@@ -0,0 +1,8 @@
+{
+  "do_normalize": true,
+  "feature_size": 1,
+  "padding_side": "right",
+  "padding_value": 0.0,
+  "return_attention_mask": true,
+  "sampling_rate": 16000
+}

pytorch_model.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8a7772a1884e576f8ce8b03059b788e2f7a734edad5a45f3676945b1b37aba5f
+size 1262101912

scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:42d008c9c215de3fb87e964d070febd87668726621c0db21bca9ed9eda04b74d
+size 623

trainer_state.json

@@ -0,0 +1,316 @@
+{
+  "best_metric": 1.0,
+  "best_model_checkpoint": "../wav2vec2-large-xlsr-53-sw/checkpoint-154",
+  "epoch": 2.9967637540453076,
+  "global_step": 462,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00015,
+      "loss": Infinity,
+      "step": 10
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0003,
+      "loss": NaN,
+      "step": 20
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.000296,
+      "loss": NaN,
+      "step": 30
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.000292,
+      "loss": NaN,
+      "step": 40
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00028799999999999995,
+      "loss": NaN,
+      "step": 50
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00028399999999999996,
+      "loss": NaN,
+      "step": 60
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00028,
+      "loss": NaN,
+      "step": 70
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.000276,
+      "loss": NaN,
+      "step": 80
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00027199999999999994,
+      "loss": NaN,
+      "step": 90
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00026799999999999995,
+      "loss": NaN,
+      "step": 100
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00026399999999999997,
+      "loss": NaN,
+      "step": 110
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00026,
+      "loss": NaN,
+      "step": 120
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.000256,
+      "loss": NaN,
+      "step": 130
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00025199999999999995,
+      "loss": NaN,
+      "step": 140
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00024799999999999996,
+      "loss": NaN,
+      "step": 150
+    },
+    {
+      "epoch": 1.0,
+      "eval_loss": Infinity,
+      "eval_runtime": 358.7961,
+      "eval_samples_per_second": 5.549,
+      "eval_wer": 1.0,
+      "step": 154
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 0.000244,
+      "loss": NaN,
+      "step": 160
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 0.00023999999999999998,
+      "loss": NaN,
+      "step": 170
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 0.00023599999999999996,
+      "loss": NaN,
+      "step": 180
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 0.00023199999999999997,
+      "loss": NaN,
+      "step": 190
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 0.00022799999999999999,
+      "loss": NaN,
+      "step": 200
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 0.000224,
+      "loss": NaN,
+      "step": 210
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 0.00021999999999999995,
+      "loss": NaN,
+      "step": 220
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 0.00021599999999999996,
+      "loss": NaN,
+      "step": 230
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 0.00021199999999999998,
+      "loss": NaN,
+      "step": 240
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 0.000208,
+      "loss": NaN,
+      "step": 250
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 0.000204,
+      "loss": NaN,
+      "step": 260
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 0.00019999999999999998,
+      "loss": NaN,
+      "step": 270
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 0.00019599999999999997,
+      "loss": NaN,
+      "step": 280
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 0.00019199999999999998,
+      "loss": NaN,
+      "step": 290
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 0.000188,
+      "loss": NaN,
+      "step": 300
+    },
+    {
+      "epoch": 2.0,
+      "eval_loss": Infinity,
+      "eval_runtime": 374.8395,
+      "eval_samples_per_second": 5.312,
+      "eval_wer": 1.0,
+      "step": 308
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 0.00018399999999999997,
+      "loss": NaN,
+      "step": 310
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 0.00017999999999999998,
+      "loss": NaN,
+      "step": 320
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 0.000176,
+      "loss": NaN,
+      "step": 330
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 0.000172,
+      "loss": NaN,
+      "step": 340
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 0.000168,
+      "loss": NaN,
+      "step": 350
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 0.00016399999999999997,
+      "loss": NaN,
+      "step": 360
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 0.00015999999999999999,
+      "loss": NaN,
+      "step": 370
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 0.000156,
+      "loss": NaN,
+      "step": 380
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 0.000152,
+      "loss": NaN,
+      "step": 390
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 0.000148,
+      "loss": NaN,
+      "step": 400
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 0.00014399999999999998,
+      "loss": NaN,
+      "step": 410
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 0.00014,
+      "loss": NaN,
+      "step": 420
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 0.00013599999999999997,
+      "loss": NaN,
+      "step": 430
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 0.00013199999999999998,
+      "loss": NaN,
+      "step": 440
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 0.000128,
+      "loss": NaN,
+      "step": 450
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 0.00012399999999999998,
+      "loss": NaN,
+      "step": 460
+    },
+    {
+      "epoch": 3.0,
+      "eval_loss": Infinity,
+      "eval_runtime": 334.0794,
+      "eval_samples_per_second": 5.96,
+      "eval_wer": 1.0,
+      "step": 462
+    }
+  ],
+  "max_steps": 770,
+  "num_train_epochs": 5,
+  "total_flos": 1.7250623020466376e+18,
+  "trial_name": null,
+  "trial_params": null
+}

training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5fcccbfe1b211984cf07a9455560cac8fbc9c69a011acf7c165d7e5331248598
+size 2351