Initial model

.ipynb_checkpoints/config-checkpoint.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": true,
+  "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": 0,
+  "transformers_version": "4.5.0.dev0",
+  "vocab_size": 40
+}

README.md

@@ -0,0 +1,349 @@
+---
+language: fa
+datasets:
+- common_voice
+tags:
+- audio
+- automatic-speech-recognition
+- speech
+- xlsr-fine-tuning-week
+license: apache-2.0
+widget:
+- label: Common Voice sample 4024
+  src: https://huggingface.co/m3hrdadfi/wav2vec2-large-xlsr-persian-v2/resolve/main/sample4024.flac
+- label: Common Voice sample 4084
+  src: https://huggingface.co/m3hrdadfi/wav2vec2-large-xlsr-persian-v2/resolve/main/sample4084.flac
+model-index:
+- name: XLSR Wav2Vec2 Persian (Farsi) V2 by Mehrdad Farahani
+  results:
+  - task: 
+      name: Speech Recognition
+      type: automatic-speech-recognition
+    dataset:
+      name: Common Voice fa
+      type: common_voice
+      args: fa  
+    metrics:
+       - name: Test WER
+         type: wer
+         value: 31.92
+        
+---
+
+# Wav2Vec2-Large-XLSR-53-Persian V2
+
+Fine-tuned [facebook/wav2vec2-large-xlsr-53](https://huggingface.co/facebook/wav2vec2-large-xlsr-53) in Persian (Farsi) using [Common Voice](https://huggingface.co/datasets/common_voice). When using this model, make sure that your speech input is sampled at 16kHz.
+
+## Usage
+The model can be used directly (without a language model) as follows:
+
+**Requirements**
+```bash
+# requirement packages
+!pip install git+https://github.com/huggingface/datasets.git
+!pip install git+https://github.com/huggingface/transformers.git
+!pip install torchaudio
+!pip install librosa
+!pip install jiwer
+!pip install hazm
+```
+
+
+**Prediction**
+```python
+import librosa
+import torch
+import torchaudio
+from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor
+from datasets import load_dataset
+
+import numpy as np
+import hazm
+import re
+import string
+
+import IPython.display as ipd
+
+_normalizer = hazm.Normalizer()
+
+chars_to_ignore = [
+    ",", "?", ".", "!", "-", ";", ":", '""', "%", "'", '"', "�",
+    "#", "!", "؟", "?", "«", "»", "،", "(", ")", "؛", "'ٔ", "٬",'ٔ', ",", "?", 
+    ".", "!", "-", ";", ":",'"',"“", "%", "‘", "”", "�", "–", "…", "_", "”", '“', '„',
+    'ā', 'š',
+    # "ء", 
+]
+
+# In case of farsi
+chars_to_ignore = chars_to_ignore + list(string.ascii_lowercase + string.digits)
+
+chars_to_mapping = {
+    'ك': 'ک', 'دِ': 'د', 'بِ': 'ب', 'زِ': 'ز', 'ذِ': 'ذ', 'شِ': 'ش', 'سِ': 'س', 'ى': 'ی',
+    'ي': 'ی', 'أ': 'ا', 'ؤ': 'و', "ے": "ی", "ۀ": "ه", "ﭘ": "پ", "ﮐ": "ک", "ﯽ": "ی",
+    "ﺎ": "ا", "ﺑ": "ب", "ﺘ": "ت", "ﺧ": "خ", "ﺩ": "د", "ﺱ": "س", "ﻀ": "ض", "ﻌ": "ع",
+    "ﻟ": "ل", "ﻡ": "م", "ﻢ": "م", "ﻪ": "ه", "ﻮ": "و", 'ﺍ': "ا", 'ة': "ه",
+    'ﯾ': "ی", 'ﯿ': "ی", 'ﺒ': "ب", 'ﺖ': "ت", 'ﺪ': "د", 'ﺮ': "ر", 'ﺴ': "س", 'ﺷ': "ش",
+    'ﺸ': "ش", 'ﻋ': "ع", 'ﻤ': "م", 'ﻥ': "ن", 'ﻧ': "ن", 'ﻭ': "و", 'ﺭ': "ر", "ﮔ": "گ",
+        
+    # "ها": "  ها", "ئ": "ی",
+        
+    "a": " ای ", "b": " بی ", "c": " سی ", "d": " دی ", "e": " ایی ", "f": " اف ",
+    "g": " جی ", "h": " اچ ", "i": " آی ", "j": " جی ", "k": " کی ", "l": " ال ",
+    "m": " ام ", "n": " ان ", "o": " او ", "p": " پی ", "q": " کیو ", "r": " آر ",
+    "s": " اس ", "t": " تی ", "u": " یو ", "v": " وی ", "w": " دبلیو ", "x": " اکس ",
+    "y": " وای ", "z": " زد ",
+    "\u200c": " ", "\u200d": " ", "\u200e": " ", "\u200f": " ", "\ufeff": " ",
+}
+
+def multiple_replace(text, chars_to_mapping):
+    pattern = "|".join(map(re.escape, chars_to_mapping.keys()))
+    return re.sub(pattern, lambda m: chars_to_mapping[m.group()], str(text))
+
+def remove_special_characters(text, chars_to_ignore_regex):
+    text = re.sub(chars_to_ignore_regex, '', text).lower() + " "
+    return text
+
+def normalizer(batch, chars_to_ignore, chars_to_mapping):
+    chars_to_ignore_regex = f"""[{"".join(chars_to_ignore)}]"""
+    text = batch["sentence"].lower().strip()
+    
+    text = _normalizer.normalize(text)
+    text = multiple_replace(text, chars_to_mapping)
+    text = remove_special_characters(text, chars_to_ignore_regex)
+    text = re.sub(" +", " ", text)
+    text = text.strip() + " "
+
+    batch["sentence"] = text
+    return batch
+
+
+def speech_file_to_array_fn(batch):
+    speech_array, sampling_rate = torchaudio.load(batch["path"])
+    speech_array = speech_array.squeeze().numpy()
+    speech_array = librosa.resample(np.asarray(speech_array), sampling_rate, 16_000)
+
+    batch["speech"] = speech_array
+    return batch
+
+
+def predict(batch):
+    features = processor(batch["speech"], sampling_rate=16_000, return_tensors="pt", padding=True)
+
+    input_values = features.input_values.to(device)
+    attention_mask = features.attention_mask.to(device)
+
+    with torch.no_grad():
+        logits = model(input_values, attention_mask=attention_mask).logits 
+        
+    pred_ids = torch.argmax(logits, dim=-1)
+
+    batch["predicted"] = processor.batch_decode(pred_ids)[0]
+    return batch
+
+
+device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+processor = Wav2Vec2Processor.from_pretrained("m3hrdadfi/wav2vec2-large-xlsr-persian-v2")
+model = Wav2Vec2ForCTC.from_pretrained("m3hrdadfi/wav2vec2-large-xlsr-persian-v2").to(device)
+
+dataset = load_dataset("common_voice", "fa", split="test[:1%]")
+dataset = dataset.map(
+    normalizer, 
+    fn_kwargs={"chars_to_ignore": chars_to_ignore, "chars_to_mapping": chars_to_mapping},
+    remove_columns=list(set(dataset.column_names) - set(['sentence', 'path']))
+)
+
+dataset = dataset.map(speech_file_to_array_fn)
+result = dataset.map(predict)
+
+max_items = np.random.randint(0, len(result), 20).tolist()
+for i in max_items:
+    reference, predicted =  result["sentence"][i], result["predicted"][i]
+    print("reference:", reference)
+    print("predicted:", predicted)
+    print('---')
+```
+
+**Output:**
+```text
+reference: عجم زنده کردم بدین پارسی 
+predicted: عجم زنده کردم بدین پارسی
+---
+reference: لباس هایم کی آماده خواهند شد 
+predicted: لباس خایم کی آماده خواهند شد
+---
+reference: با مهان همنشین شدم 
+predicted: با مهان همنشین شدم
+---
+reference: یکی از بهترین فیلم هایی بود که در این سال ها دیدم 
+predicted: یکی از بهترین فیلمهایی بود که در این سالها دیدم
+---
+reference: اون خیلی بد ماساژ میده 
+predicted: اون خیلی بد ماساژ میده
+---
+reference: هنوزم بزرگترین دستاورد دولت روحانی اینه که رییسی رییسجمهور نشد 
+predicted: هنوزم بزرگترین دستآوردار دولت روانیاینه که ریسی ریسیومرو نشد
+---
+reference: واسه بدنسازی آماده ای 
+predicted: واسه بعدنسافی آماده ای
+---
+reference: خدای من شماها سالمین 
+predicted: خدای من شما ها سالمین
+---
+reference: بهشون ثابت میشه که دروغ نگفتم 
+predicted: بهشون ثابت میشه که دروغ مگفتم
+---
+reference: آیا ممکن است یک پتو برای من بیاورید 
+predicted: سف کمیتخ لظا
+---
+reference: نزدیک جلو 
+predicted: رزیک جلو
+---
+reference: شایعه پراکن دربارهاش دروغ و شایعه می سازد 
+predicted: شایه پراکن دربارهاش دروغ و شایعه می سازد
+---
+reference: وقتی نیاز است که یک چهره دوستانه بیابند 
+predicted: وقتی نیاز است یک چهره دوستانه بیابند
+---
+reference: ممکنه رادیواکتیوی چیزی باشه 
+predicted: ممکنه به آدیوتیوی چیزی باشه
+---
+reference: دهنتون رو ببندید 
+predicted: دهن جن رو ببندید
+---
+reference: پاشیم بریم قند و شکر و روغنمون رو بگیریم تا تموم نشده 
+predicted: پاشین بریم قند و شکر و روغنمون رو بگیریم تا تموم نشده
+---
+reference: اما قبل از تمام کردن بحث تاریخی باید ذکری هم از ناپیکس بکنیم 
+predicted: اما قبل از تمام کردن بحث تاریخی باید ذکری هم از نایپکس بکنیم
+---
+reference: لطفا کپی امضا شده قرارداد را بازگردانید 
+predicted: لطفا کپی امضال شده قرار داد را باز گردانید
+---
+reference: خیلی هم چیز مهمی نیست 
+predicted: خیلی هم چیز مهمی نیست
+---
+reference: شایعه پراکن دربارهاش دروغ و شایعه می سازد 
+predicted: شایه پراکن دربارهاش دروغ و شایعه می سازد
+---
+```
+
+## Evaluation
+
+The model can be evaluated as follows on the Persian (Farsi) test data of Common Voice.
+
+```python
+import librosa
+import torch
+import torchaudio
+from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor
+from datasets import load_dataset, load_metric
+
+import numpy as np
+import hazm
+import re
+import string
+
+_normalizer = hazm.Normalizer()
+
+chars_to_ignore = [
+    ",", "?", ".", "!", "-", ";", ":", '""', "%", "'", '"', "�",
+    "#", "!", "؟", "?", "«", "»", "،", "(", ")", "؛", "'ٔ", "٬",'ٔ', ",", "?", 
+    ".", "!", "-", ";", ":",'"',"“", "%", "‘", "”", "�", "–", "…", "_", "”", '“', '„',
+    'ā', 'š',
+    # "ء", 
+]
+
+# In case of farsi
+chars_to_ignore = chars_to_ignore + list(string.ascii_lowercase + string.digits)
+
+chars_to_mapping = {
+    'ك': 'ک', 'دِ': 'د', 'بِ': 'ب', 'زِ': 'ز', 'ذِ': 'ذ', 'شِ': 'ش', 'سِ': 'س', 'ى': 'ی',
+    'ي': 'ی', 'أ': 'ا', 'ؤ': 'و', "ے": "ی", "ۀ": "ه", "ﭘ": "پ", "ﮐ": "ک", "ﯽ": "ی",
+    "ﺎ": "ا", "ﺑ": "ب", "ﺘ": "ت", "ﺧ": "خ", "ﺩ": "د", "ﺱ": "س", "ﻀ": "ض", "ﻌ": "ع",
+    "ﻟ": "ل", "ﻡ": "م", "ﻢ": "م", "ﻪ": "ه", "ﻮ": "و", 'ﺍ': "ا", 'ة': "ه",
+    'ﯾ': "ی", 'ﯿ': "ی", 'ﺒ': "ب", 'ﺖ': "ت", 'ﺪ': "د", 'ﺮ': "ر", 'ﺴ': "س", 'ﺷ': "ش",
+    'ﺸ': "ش", 'ﻋ': "ع", 'ﻤ': "م", 'ﻥ': "ن", 'ﻧ': "ن", 'ﻭ': "و", 'ﺭ': "ر", "ﮔ": "گ",
+        
+    # "ها": "  ها", "ئ": "ی",
+        
+    "a": " ای ", "b": " بی ", "c": " سی ", "d": " دی ", "e": " ایی ", "f": " اف ",
+    "g": " جی ", "h": " اچ ", "i": " آی ", "j": " جی ", "k": " کی ", "l": " ال ",
+    "m": " ام ", "n": " ان ", "o": " او ", "p": " پی ", "q": " کیو ", "r": " آر ",
+    "s": " اس ", "t": " تی ", "u": " یو ", "v": " وی ", "w": " دبلیو ", "x": " اکس ",
+    "y": " وای ", "z": " زد ",
+    "\u200c": " ", "\u200d": " ", "\u200e": " ", "\u200f": " ", "\ufeff": " ",
+}
+
+def multiple_replace(text, chars_to_mapping):
+    pattern = "|".join(map(re.escape, chars_to_mapping.keys()))
+    return re.sub(pattern, lambda m: chars_to_mapping[m.group()], str(text))
+
+def remove_special_characters(text, chars_to_ignore_regex):
+    text = re.sub(chars_to_ignore_regex, '', text).lower() + " "
+    return text
+
+def normalizer(batch, chars_to_ignore, chars_to_mapping):
+    chars_to_ignore_regex = f"""[{"".join(chars_to_ignore)}]"""
+    text = batch["sentence"].lower().strip()
+    
+    text = _normalizer.normalize(text)
+    text = multiple_replace(text, chars_to_mapping)
+    text = remove_special_characters(text, chars_to_ignore_regex)
+    text = re.sub(" +", " ", text)
+    text = text.strip() + " "
+
+    batch["sentence"] = text
+    return batch
+
+
+def speech_file_to_array_fn(batch):
+    speech_array, sampling_rate = torchaudio.load(batch["path"])
+    speech_array = speech_array.squeeze().numpy()
+    speech_array = librosa.resample(np.asarray(speech_array), sampling_rate, 16_000)
+
+    batch["speech"] = speech_array
+    return batch
+
+
+def predict(batch):
+    features = processor(batch["speech"], sampling_rate=16_000, return_tensors="pt", padding=True)
+
+    input_values = features.input_values.to(device)
+    attention_mask = features.attention_mask.to(device)
+
+    with torch.no_grad():
+        logits = model(input_values, attention_mask=attention_mask).logits 
+        
+    pred_ids = torch.argmax(logits, dim=-1)
+
+    batch["predicted"] = processor.batch_decode(pred_ids)[0]
+    return batch
+
+
+device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+processor = Wav2Vec2Processor.from_pretrained("m3hrdadfi/wav2vec2-large-xlsr-persian-v2")
+model = Wav2Vec2ForCTC.from_pretrained("m3hrdadfi/wav2vec2-large-xlsr-persian-v2").to(device)
+
+dataset = load_dataset("common_voice", "fa", split="test")
+dataset = dataset.map(
+    normalizer, 
+    fn_kwargs={"chars_to_ignore": chars_to_ignore, "chars_to_mapping": chars_to_mapping},
+    remove_columns=list(set(dataset.column_names) - set(['sentence', 'path']))
+)
+dataset = dataset.map(speech_file_to_array_fn)
+result = dataset.map(predict)
+
+wer = load_metric("wer")
+print("WER: {:.2f}".format(100 * wer.compute(predictions=result["predicted"], references=result["sentence"])))
+```
+
+**Test Result:** 
+- WER: 31.92%
+
+
+## Training
+The Common Voice `train`, `validation` datasets were used for training.
+
+You can see the training states [here](https://wandb.ai/m3hrdadfi/finetuned_wav2vec_xlsr_persian/reports/Fine-Tuning-for-Wav2Vec2-Large-XLSR-53-Persian--Vmlldzo1NjY1NjU?accessToken=pspukt0liicopnwe93wo1ipetqk0gzkuv8669g00wc6hcesk1fh0rfkbd0h46unk)
+
+The script used for training can be found [here](https://colab.research.google.com/github/m3hrdadfi/notebooks/blob/main/Fine_Tune_XLSR_Wav2Vec2_on_Persian_ASR_with_%F0%9F%A4%97_Transformers_ipynb.ipynb)

all_results.json

@@ -0,0 +1,24 @@
+{
+    "epoch": 30.0,
+    "eval_loss": 0.5013265609741211,
+    "eval_mem_cpu_alloc_delta": 447967023,
+    "eval_mem_cpu_peaked_delta": 40936880,
+    "eval_mem_gpu_alloc_delta": 0,
+    "eval_mem_gpu_peaked_delta": 6271776768,
+    "eval_runtime": 635.3557,
+    "eval_samples": 5208,
+    "eval_samples_per_second": 8.197,
+    "eval_wer": 0.32129213337144935,
+    "init_mem_cpu_alloc_delta": 20270917,
+    "init_mem_cpu_peaked_delta": 18306,
+    "init_mem_gpu_alloc_delta": 1261919232,
+    "init_mem_gpu_peaked_delta": 0,
+    "total_flos": 4.998078272539758e+19,
+    "train_mem_cpu_alloc_delta": 57067470,
+    "train_mem_cpu_peaked_delta": 487786762,
+    "train_mem_gpu_alloc_delta": 3793118208,
+    "train_mem_gpu_peaked_delta": 6813593600,
+    "train_runtime": 61300.1017,
+    "train_samples": 12730,
+    "train_samples_per_second": 0.111
+}

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": true,
+  "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": 0,
+  "transformers_version": "4.5.0.dev0",
+  "vocab_size": 40
+}

eval_results.json

@@ -0,0 +1,12 @@
+{
+    "epoch": 30.0,
+    "eval_loss": 0.5013265609741211,
+    "eval_mem_cpu_alloc_delta": 447967023,
+    "eval_mem_cpu_peaked_delta": 40936880,
+    "eval_mem_gpu_alloc_delta": 0,
+    "eval_mem_gpu_peaked_delta": 6271776768,
+    "eval_runtime": 635.3557,
+    "eval_samples": 5208,
+    "eval_samples_per_second": 8.197,
+    "eval_wer": 0.32129213337144935
+}

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:c06a17254c2d51cc616567a8b7d0add932760e3c193d03c46676ee2441b322f0
+size 1262097815

result.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a47518a2653bffa224b006c91801824b302bac9dd29cda2a1360948b7837aa1d
+size 3183

special_tokens_map.json

@@ -0,0 +1 @@
+{"bos_token": "<s>", "eos_token": "</s>", "unk_token": "<unk>", "pad_token": "<pad>"}

tokenizer_config.json

@@ -0,0 +1 @@
+{"unk_token": "<unk>", "bos_token": "<s>", "eos_token": "</s>", "pad_token": "<pad>", "do_lower_case": false, "word_delimiter_token": "|"}

train_results.json

@@ -0,0 +1,15 @@
+{
+    "epoch": 30.0,
+    "init_mem_cpu_alloc_delta": 20270917,
+    "init_mem_cpu_peaked_delta": 18306,
+    "init_mem_gpu_alloc_delta": 1261919232,
+    "init_mem_gpu_peaked_delta": 0,
+    "total_flos": 4.998078272539758e+19,
+    "train_mem_cpu_alloc_delta": 57067470,
+    "train_mem_cpu_peaked_delta": 487786762,
+    "train_mem_gpu_alloc_delta": 3793118208,
+    "train_mem_gpu_peaked_delta": 6813593600,
+    "train_runtime": 61300.1017,
+    "train_samples": 12730,
+    "train_samples_per_second": 0.111
+}

trainer_state.json

@@ -0,0 +1,331 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 29.997802197802198,
+  "global_step": 6810,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 1.32,
+      "learning_rate": 0.00017999999999999998,
+      "loss": 11.6291,
+      "step": 300
+    },
+    {
+      "epoch": 1.32,
+      "eval_loss": 3.028644323348999,
+      "eval_runtime": 643.9816,
+      "eval_samples_per_second": 8.087,
+      "eval_wer": 1.0,
+      "step": 300
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 0.0002952456418383518,
+      "loss": 3.0796,
+      "step": 600
+    },
+    {
+      "epoch": 2.64,
+      "eval_loss": 2.9614171981811523,
+      "eval_runtime": 650.4972,
+      "eval_samples_per_second": 8.006,
+      "eval_wer": 1.0,
+      "step": 600
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 0.0002809825673534073,
+      "loss": 1.6756,
+      "step": 900
+    },
+    {
+      "epoch": 3.96,
+      "eval_loss": 0.6643654108047485,
+      "eval_runtime": 668.5248,
+      "eval_samples_per_second": 7.79,
+      "eval_wer": 0.5919852387016294,
+      "step": 900
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 0.00026671949286846274,
+      "loss": 0.5169,
+      "step": 1200
+    },
+    {
+      "epoch": 5.29,
+      "eval_loss": 0.48998841643333435,
+      "eval_runtime": 670.8355,
+      "eval_samples_per_second": 7.763,
+      "eval_wer": 0.4740247927440942,
+      "step": 1200
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 0.0002524564183835182,
+      "loss": 0.3643,
+      "step": 1500
+    },
+    {
+      "epoch": 6.61,
+      "eval_loss": 0.4465296268463135,
+      "eval_runtime": 685.664,
+      "eval_samples_per_second": 7.596,
+      "eval_wer": 0.43660178279061307,
+      "step": 1500
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 0.00023819334389857367,
+      "loss": 0.277,
+      "step": 1800
+    },
+    {
+      "epoch": 7.93,
+      "eval_loss": 0.4225097894668579,
+      "eval_runtime": 675.9712,
+      "eval_samples_per_second": 7.704,
+      "eval_wer": 0.4157592453026326,
+      "step": 1800
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 0.00022393026941362913,
+      "loss": 0.2271,
+      "step": 2100
+    },
+    {
+      "epoch": 9.25,
+      "eval_loss": 0.4406384825706482,
+      "eval_runtime": 685.3056,
+      "eval_samples_per_second": 7.6,
+      "eval_wer": 0.39260375789391616,
+      "step": 2100
+    },
+    {
+      "epoch": 10.57,
+      "learning_rate": 0.0002096671949286846,
+      "loss": 0.1895,
+      "step": 2400
+    },
+    {
+      "epoch": 10.57,
+      "eval_loss": 0.4478096067905426,
+      "eval_runtime": 689.1958,
+      "eval_samples_per_second": 7.557,
+      "eval_wer": 0.3899529613555446,
+      "step": 2400
+    },
+    {
+      "epoch": 11.89,
+      "learning_rate": 0.0001954041204437401,
+      "loss": 0.1653,
+      "step": 2700
+    },
+    {
+      "epoch": 11.89,
+      "eval_loss": 0.4374055862426758,
+      "eval_runtime": 693.8794,
+      "eval_samples_per_second": 7.506,
+      "eval_wer": 0.3792977987993451,
+      "step": 2700
+    },
+    {
+      "epoch": 13.22,
+      "learning_rate": 0.00018114104595879556,
+      "loss": 0.1366,
+      "step": 3000
+    },
+    {
+      "epoch": 13.22,
+      "eval_loss": 0.4575331211090088,
+      "eval_runtime": 703.7067,
+      "eval_samples_per_second": 7.401,
+      "eval_wer": 0.37355440629954,
+      "step": 3000
+    },
+    {
+      "epoch": 14.54,
+      "learning_rate": 0.000166877971473851,
+      "loss": 0.1219,
+      "step": 3300
+    },
+    {
+      "epoch": 14.54,
+      "eval_loss": 0.4847738444805145,
+      "eval_runtime": 705.6195,
+      "eval_samples_per_second": 7.381,
+      "eval_wer": 0.36983809350554847,
+      "step": 3300
+    },
+    {
+      "epoch": 15.86,
+      "learning_rate": 0.00015261489698890646,
+      "loss": 0.1102,
+      "step": 3600
+    },
+    {
+      "epoch": 15.86,
+      "eval_loss": 0.4509291350841522,
+      "eval_runtime": 713.2815,
+      "eval_samples_per_second": 7.301,
+      "eval_wer": 0.35738974505574467,
+      "step": 3600
+    },
+    {
+      "epoch": 17.18,
+      "learning_rate": 0.00013835182250396196,
+      "loss": 0.1017,
+      "step": 3900
+    },
+    {
+      "epoch": 17.18,
+      "eval_loss": 0.45042145252227783,
+      "eval_runtime": 714.3616,
+      "eval_samples_per_second": 7.29,
+      "eval_wer": 0.34715039372125056,
+      "step": 3900
+    },
+    {
+      "epoch": 18.5,
+      "learning_rate": 0.00012408874801901742,
+      "loss": 0.0937,
+      "step": 4200
+    },
+    {
+      "epoch": 18.5,
+      "eval_loss": 0.4522610008716583,
+      "eval_runtime": 719.9965,
+      "eval_samples_per_second": 7.233,
+      "eval_wer": 0.3469164999090413,
+      "step": 4200
+    },
+    {
+      "epoch": 19.82,
+      "learning_rate": 0.00010982567353407289,
+      "loss": 0.0829,
+      "step": 4500
+    },
+    {
+      "epoch": 19.82,
+      "eval_loss": 0.4488724172115326,
+      "eval_runtime": 720.1162,
+      "eval_samples_per_second": 7.232,
+      "eval_wer": 0.3452012786195067,
+      "step": 4500
+    },
+    {
+      "epoch": 21.15,
+      "learning_rate": 9.556259904912835e-05,
+      "loss": 0.0752,
+      "step": 4800
+    },
+    {
+      "epoch": 21.15,
+      "eval_loss": 0.4762090742588043,
+      "eval_runtime": 730.1963,
+      "eval_samples_per_second": 7.132,
+      "eval_wer": 0.33800254684373293,
+      "step": 4800
+    },
+    {
+      "epoch": 22.47,
+      "learning_rate": 8.129952456418383e-05,
+      "loss": 0.0698,
+      "step": 5100
+    },
+    {
+      "epoch": 22.47,
+      "eval_loss": 0.480989009141922,
+      "eval_runtime": 727.7194,
+      "eval_samples_per_second": 7.157,
+      "eval_wer": 0.33597546713791937,
+      "step": 5100
+    },
+    {
+      "epoch": 23.79,
+      "learning_rate": 6.70364500792393e-05,
+      "loss": 0.0642,
+      "step": 5400
+    },
+    {
+      "epoch": 23.79,
+      "eval_loss": 0.49236759543418884,
+      "eval_runtime": 737.7861,
+      "eval_samples_per_second": 7.059,
+      "eval_wer": 0.3343382104524546,
+      "step": 5400
+    },
+    {
+      "epoch": 25.11,
+      "learning_rate": 5.2773375594294765e-05,
+      "loss": 0.058,
+      "step": 5700
+    },
+    {
+      "epoch": 25.11,
+      "eval_loss": 0.4960506558418274,
+      "eval_runtime": 740.7836,
+      "eval_samples_per_second": 7.03,
+      "eval_wer": 0.3302840510408275,
+      "step": 5700
+    },
+    {
+      "epoch": 26.43,
+      "learning_rate": 3.851030110935023e-05,
+      "loss": 0.0553,
+      "step": 6000
+    },
+    {
+      "epoch": 26.43,
+      "eval_loss": 0.4915299415588379,
+      "eval_runtime": 748.2447,
+      "eval_samples_per_second": 6.96,
+      "eval_wer": 0.32506042256815404,
+      "step": 6000
+    },
+    {
+      "epoch": 27.75,
+      "learning_rate": 2.4247226624405704e-05,
+      "loss": 0.0545,
+      "step": 6300
+    },
+    {
+      "epoch": 27.75,
+      "eval_loss": 0.5011223554611206,
+      "eval_runtime": 762.4791,
+      "eval_samples_per_second": 6.83,
+      "eval_wer": 0.3210062631565269,
+      "step": 6300
+    },
+    {
+      "epoch": 29.07,
+      "learning_rate": 9.984152139461172e-06,
+      "loss": 0.0523,
+      "step": 6600
+    },
+    {
+      "epoch": 29.07,
+      "eval_loss": 0.49885958433151245,
+      "eval_runtime": 743.9824,
+      "eval_samples_per_second": 7.0,
+      "eval_wer": 0.3229034018555576,
+      "step": 6600
+    },
+    {
+      "epoch": 30.0,
+      "step": 6810,
+      "total_flos": 4.998078272539758e+19,
+      "train_runtime": 61300.1017,
+      "train_samples_per_second": 0.111
+    }
+  ],
+  "max_steps": 6810,
+  "num_train_epochs": 30,
+  "total_flos": 4.998078272539758e+19,
+  "trial_name": null,
+  "trial_params": null
+}

training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5dbe2c70e4a0a1ea1b9c9db9b598b453f20cf687a2938fc2858cb547981ea9e0
+size 2351

vocab.json

@@ -0,0 +1 @@
+{"<pad>": 0, "<s>": 1, "</s>": 2, "<unk>": 3, "|": 4, "ء": 5, "آ": 6, "ئ": 7, "ا": 8, "ب": 9, "ت": 10, "ث": 11, "ج": 12, "ح": 13, "خ": 14, "د": 15, "ذ": 16, "ر": 17, "ز": 18, "س": 19, "ش": 20, "ص": 21, "ض": 22, "ط": 23, "ظ": 24, "ع": 25, "غ": 26, "ف": 27, "ق": 28, "ل": 29, "م": 30, "ن": 31, "ه": 32, "و": 33, "پ": 34, "چ": 35, "ژ": 36, "ک": 37, "گ": 38, "ی": 39}