Datasets:

ANANDHU-SCT

/

Speech-to-text

like 1

audiofile29.wav

this is the last sentence in this dataset so thank you

{'array': array([-0.00154497, -0.00154497, -0.00154497, ..., -0.00154497, -0.00154497, -0.00154497], dtype=float32), 'path': '/home/ba-anandhus/Documents/fine tuning/mono/audiofile29.wav', 'sampling_rate': 16000}

audiofile25.wav

we are using shared memory for matrix multiplication

{'array': array([-0.0013941, -0.0013941, -0.0013941, ..., -0.0013941, -0.0013941, -0.0013941], dtype=float32), 'path': '/home/ba-anandhus/Documents/fine tuning/mono/audiofile25.wav', 'sampling_rate': 16000}

audiofile13.wav

so i asked my friends also to record voice and send it to me

{'array': array([-0.00135984, -0.00135984, -0.00135984, ..., -0.00135984, -0.00135984, -0.00135984], dtype=float32), 'path': '/home/ba-anandhus/Documents/fine tuning/mono/audiofile13.wav', 'sampling_rate': 16000}

audiofile1.wav

good morning everyone

{'array': array([-9.7171287e-05, -9.7171287e-05, -9.7171287e-05, ..., -9.7171287e-05, 3.4426435e-04, 7.8569993e-04], dtype=float32), 'path': '/home/ba-anandhus/Documents/fine tuning/mono/audiofile1.wav', 'sampling_rate': 16000}

audiofile5.wav

train data is used to train the deep learning model

{'array': array([-0.00258282, -0.00258282, -0.00258282, ..., -0.00278775, -0.00258282, -0.00278775], dtype=float32), 'path': '/home/ba-anandhus/Documents/fine tuning/mono/audiofile5.wav', 'sampling_rate': 16000}

audiofile17.wav

in shared memory consider there are lot of threads in a thread block

{'array': array([-0.0011769, -0.0011769, -0.0011769, ..., -0.0011769, -0.0011769, -0.0011769], dtype=float32), 'path': '/home/ba-anandhus/Documents/fine tuning/mono/audiofile17.wav', 'sampling_rate': 16000}

audiofile6.wav

the validation dataset is used to fine tune the trained model and reduce error

{'array': array([-0.0019547, -0.0019547, -0.0019547, ..., -0.0019547, -0.0019547, -0.0019547], dtype=float32), 'path': '/home/ba-anandhus/Documents/fine tuning/mono/audiofile6.wav', 'sampling_rate': 16000}

audiofile14.wav

since it is not their sole duty im also indulging in dataset creation

{'array': array([6.8441186e-05, 6.8441186e-05, 6.8441186e-05, ..., 6.8441186e-05, 6.8441186e-05, 6.8441186e-05], dtype=float32), 'path': '/home/ba-anandhus/Documents/fine tuning/mono/audiofile14.wav', 'sampling_rate': 16000}

audiofile12.wav

model was working fine for the sample dataset but my accent was not recognising

{'array': array([-0.00075871, -0.00075871, -0.00075871, ..., -0.00075871, -0.00075871, -0.00075871], dtype=float32), 'path': '/home/ba-anandhus/Documents/fine tuning/mono/audiofile12.wav', 'sampling_rate': 16000}

audiofile23.wav

after that it will do commutation on shared memory

{'array': array([-0.00209466, -0.00209466, -0.00209466, ..., -0.00235505, -0.00235505, -0.00209466], dtype=float32), 'path': '/home/ba-anandhus/Documents/fine tuning/mono/audiofile23.wav', 'sampling_rate': 16000}

audiofile2.wav

it is always nice to meet you in a fresh mood

{'array': array([ 2.5861387e-05, 2.5861387e-05, 2.5861387e-05, ..., 2.6418168e-02, 1.2057647e-02, -8.1247035e-03], dtype=float32), 'path': '/home/ba-anandhus/Documents/fine tuning/mono/audiofile2.wav', 'sampling_rate': 16000}

audiofile3.wav

this is the custom dataset which is used to train the wave2vec model

{'array': array([-0.00024371, -0.00024371, -0.00024371, ..., -0.02418295, -0.02207067, -0.00024371], dtype=float32), 'path': '/home/ba-anandhus/Documents/fine tuning/mono/audiofile3.wav', 'sampling_rate': 16000}

audiofile26.wav

there are two shared arrays since we are using two memory locations

{'array': array([-0.00214535, -0.00214535, -0.00214535, ..., -0.00240758, -0.00214535, -0.00240758], dtype=float32), 'path': '/home/ba-anandhus/Documents/fine tuning/mono/audiofile26.wav', 'sampling_rate': 16000}

audiofile4.wav

this consist of test train and validation data included

{'array': array([-0.00096769, -0.00096769, -0.00096769, ..., -0.00096769, -0.00096769, -0.00096769], dtype=float32), 'path': '/home/ba-anandhus/Documents/fine tuning/mono/audiofile4.wav', 'sampling_rate': 16000}

audiofile22.wav

it is necessory to preload a small block from input array

{'array': array([-0.00164573, -0.00164573, -0.00164573, ..., -0.00191057, -0.00191057, -0.00164573], dtype=float32), 'path': '/home/ba-anandhus/Documents/fine tuning/mono/audiofile22.wav', 'sampling_rate': 16000}

audiofile27.wav

one of our bro is installing the python libraries for doing the program along with the workshop

{'array': array([-0.00197866, -0.00197866, -0.00197866, ..., -0.00197866, -0.00197866, -0.00197866], dtype=float32), 'path': '/home/ba-anandhus/Documents/fine tuning/mono/audiofile27.wav', 'sampling_rate': 16000}

audiofile19.wav

we can tell gpu how much memory we can use as cache

{'array': array([-0.00142616, -0.00142616, -0.00142616, ..., -0.00142616, -0.00142616, -0.00142616], dtype=float32), 'path': '/home/ba-anandhus/Documents/fine tuning/mono/audiofile19.wav', 'sampling_rate': 16000}

audiofile30.wav

so this are the sentences that would turn the model into a better one

{'array': array([-0.00090364, -0.00090364, -0.00090364, ..., -0.00120979, -0.00120979, -0.00120979], dtype=float32), 'path': '/home/ba-anandhus/Documents/fine tuning/mono/audiofile30.wav', 'sampling_rate': 16000}

audiofile21.wav

memory is allocated once during the duration of the kernal

{'array': array([-0.00097256, -0.00097256, -0.00097256, ..., -0.00097256, -0.00128039, -0.00128039], dtype=float32), 'path': '/home/ba-anandhus/Documents/fine tuning/mono/audiofile21.wav', 'sampling_rate': 16000}

audiofile8.wav

these are recorded personally by me

{'array': array([-0.00147271, -0.00147271, -0.00147271, ..., -0.00147271, -0.00147271, -0.00147271], dtype=float32), 'path': '/home/ba-anandhus/Documents/fine tuning/mono/audiofile8.wav', 'sampling_rate': 16000}

audiofile11.wav

the model from facebok was identified after a lot of research and preperation

{'array': array([-0.00012905, -0.00012905, -0.00012905, ..., -0.00012905, -0.000616 , -0.000616 ], dtype=float32), 'path': '/home/ba-anandhus/Documents/fine tuning/mono/audiofile11.wav', 'sampling_rate': 16000}

audiofile15.wav

we are having a session based on cuda programming

{'array': array([-0.00074556, -0.00074556, -0.00074556, ..., -0.00074556, -0.00074556, -0.00074556], dtype=float32), 'path': '/home/ba-anandhus/Documents/fine tuning/mono/audiofile15.wav', 'sampling_rate': 16000}

audiofile20.wav

we are passing two arguments shape and time

{'array': array([-0.00225772, -0.00225772, -0.00225772, ..., -0.00225772, -0.00225772, -0.00225772], dtype=float32), 'path': '/home/ba-anandhus/Documents/fine tuning/mono/audiofile20.wav', 'sampling_rate': 16000}

audiofile7.wav

the test data is used to test the accuracy

{'array': array([-0.00150242, -0.00150242, -0.00150242, ..., -0.00150242, -0.00150242, -0.00150242], dtype=float32), 'path': '/home/ba-anandhus/Documents/fine tuning/mono/audiofile7.wav', 'sampling_rate': 16000}