mirror of
https://github.com/storytold/storyteller-ml.git
synced 2026-10-28 09:08:54 +00:00
Introduce code from Rust monorepo. This is a mess
This commit is contained in:
Brandon Thomas
@@ -0,0 +1,428 @@
|
||||
# This config contains the default values for training a Citrinet model with CTC loss and BPE-based vocabulary.
|
||||
# Default learning parameters in this config are set for effective batch size of 1k on 32 GPUs.
|
||||
# To train it with smaller batch sizes, you may need to re-tune the learning parameters or use higher accumulate_grad_batches.
|
||||
# If training for a short time, you can also reduce weight decay to 0.
|
||||
|
||||
# Training Recipe
|
||||
# This model can be trained using the default settings in this config with FP32 precision.
|
||||
# When training under AMP, increase `warmup_steps` to 5000 for stable training.
|
||||
# In order to create Citrinet-C, find-replace `filters: 384` with `filters: C`.
|
||||
# When reducing the receptive field of these models, it is advised to reduce the amount of augmentation
|
||||
# for larger models from 10x time masking to 5x or 2x time masking.
|
||||
# For further details regarding Citrinet, visit - https://docs.nvidia.com/deeplearning/nemo/user-guide/docs/en/main/asr/configs.html#citrinet
|
||||
|
||||
name: &name "Citrinet-384-8x-Stride"
|
||||
|
||||
model:
|
||||
sample_rate: &sample_rate 16000
|
||||
|
||||
train_ds:
|
||||
manifest_filepath: ???
|
||||
sample_rate: 16000
|
||||
batch_size: 32
|
||||
trim_silence: false
|
||||
max_duration: 16.7
|
||||
shuffle: true
|
||||
is_tarred: false
|
||||
tarred_audio_filepaths: null
|
||||
use_start_end_token: false
|
||||
|
||||
validation_ds:
|
||||
manifest_filepath: ???
|
||||
sample_rate: 16000
|
||||
batch_size: 32
|
||||
shuffle: false
|
||||
use_start_end_token: false
|
||||
|
||||
test_ds:
|
||||
manifest_filepath: null
|
||||
sample_rate: 16000
|
||||
batch_size: 32
|
||||
shuffle: false
|
||||
use_start_end_token: false
|
||||
|
||||
model_defaults:
|
||||
repeat: 5
|
||||
dropout: 0.0
|
||||
separable: true
|
||||
se: true
|
||||
se_context_size: -1
|
||||
kernel_size_factor: 1.0
|
||||
|
||||
tokenizer:
|
||||
dir: ??? # path to directory which contains either tokenizer.model (bpe) or vocab.txt (for wpe)
|
||||
type: ??? # Can be either bpe or wpe
|
||||
|
||||
preprocessor:
|
||||
_target_: nemo.collections.asr.modules.AudioToMelSpectrogramPreprocessor
|
||||
sample_rate: *sample_rate
|
||||
normalize: "per_feature"
|
||||
window_size: 0.025
|
||||
window_stride: 0.01
|
||||
window: "hann"
|
||||
features: &n_mels 80
|
||||
n_fft: 512
|
||||
frame_splicing: 1
|
||||
dither: 0.00001
|
||||
pad_to: 16
|
||||
stft_conv: false
|
||||
|
||||
spec_augment:
|
||||
_target_: nemo.collections.asr.modules.SpectrogramAugmentation
|
||||
freq_masks: 2
|
||||
time_masks: 2
|
||||
freq_width: 27
|
||||
time_width: 0.05
|
||||
|
||||
encoder:
|
||||
_target_: nemo.collections.asr.modules.ConvASREncoder
|
||||
feat_in: *n_mels
|
||||
activation: relu
|
||||
conv_mask: true
|
||||
|
||||
jasper:
|
||||
- filters: 384
|
||||
repeat: 1
|
||||
kernel: [5]
|
||||
stride: [1]
|
||||
dilation: [1]
|
||||
dropout: 0.0
|
||||
residual: false
|
||||
separable: ${model.model_defaults.separable}
|
||||
se: ${model.model_defaults.se}
|
||||
se_context_size: ${model.model_defaults.se_context_size}
|
||||
|
||||
- filters: 384
|
||||
repeat: ${model.model_defaults.repeat}
|
||||
kernel: [11]
|
||||
stride: [2]
|
||||
dilation: [1]
|
||||
dropout: ${model.model_defaults.dropout}
|
||||
residual: true
|
||||
separable: ${model.model_defaults.separable}
|
||||
se: ${model.model_defaults.se}
|
||||
se_context_size: ${model.model_defaults.se_context_size}
|
||||
stride_last: true
|
||||
residual_mode: "stride_add"
|
||||
kernel_size_factor: ${model.model_defaults.kernel_size_factor}
|
||||
|
||||
- filters: 384
|
||||
repeat: ${model.model_defaults.repeat}
|
||||
kernel: [13]
|
||||
stride: [1]
|
||||
dilation: [1]
|
||||
dropout: ${model.model_defaults.dropout}
|
||||
residual: true
|
||||
separable: ${model.model_defaults.separable}
|
||||
se: ${model.model_defaults.se}
|
||||
se_context_size: ${model.model_defaults.se_context_size}
|
||||
kernel_size_factor: ${model.model_defaults.kernel_size_factor}
|
||||
|
||||
- filters: 384
|
||||
repeat: ${model.model_defaults.repeat}
|
||||
kernel: [15]
|
||||
stride: [1]
|
||||
dilation: [1]
|
||||
dropout: ${model.model_defaults.dropout}
|
||||
residual: true
|
||||
separable: ${model.model_defaults.separable}
|
||||
se: ${model.model_defaults.se}
|
||||
se_context_size: ${model.model_defaults.se_context_size}
|
||||
kernel_size_factor: ${model.model_defaults.kernel_size_factor}
|
||||
|
||||
- filters: 384
|
||||
repeat: ${model.model_defaults.repeat}
|
||||
kernel: [17]
|
||||
stride: [1]
|
||||
dilation: [1]
|
||||
dropout: ${model.model_defaults.dropout}
|
||||
residual: true
|
||||
separable: ${model.model_defaults.separable}
|
||||
se: ${model.model_defaults.se}
|
||||
se_context_size: ${model.model_defaults.se_context_size}
|
||||
kernel_size_factor: ${model.model_defaults.kernel_size_factor}
|
||||
|
||||
- filters: 384
|
||||
repeat: ${model.model_defaults.repeat}
|
||||
kernel: [19]
|
||||
stride: [1]
|
||||
dilation: [1]
|
||||
dropout: ${model.model_defaults.dropout}
|
||||
residual: true
|
||||
separable: ${model.model_defaults.separable}
|
||||
se: ${model.model_defaults.se}
|
||||
se_context_size: ${model.model_defaults.se_context_size}
|
||||
kernel_size_factor: ${model.model_defaults.kernel_size_factor}
|
||||
|
||||
- filters: 384
|
||||
repeat: ${model.model_defaults.repeat}
|
||||
kernel: [21]
|
||||
stride: [1]
|
||||
dilation: [1]
|
||||
dropout: ${model.model_defaults.dropout}
|
||||
residual: true
|
||||
separable: ${model.model_defaults.separable}
|
||||
se: ${model.model_defaults.se}
|
||||
se_context_size: ${model.model_defaults.se_context_size}
|
||||
kernel_size_factor: ${model.model_defaults.kernel_size_factor}
|
||||
|
||||
- filters: 384
|
||||
repeat: ${model.model_defaults.repeat}
|
||||
kernel: [13]
|
||||
stride: [2] # *stride
|
||||
dilation: [1]
|
||||
dropout: ${model.model_defaults.dropout}
|
||||
residual: true
|
||||
separable: ${model.model_defaults.separable}
|
||||
se: ${model.model_defaults.se}
|
||||
se_context_size: ${model.model_defaults.se_context_size}
|
||||
stride_last: true
|
||||
residual_mode: "stride_add"
|
||||
kernel_size_factor: ${model.model_defaults.kernel_size_factor}
|
||||
|
||||
- filters: 384
|
||||
repeat: ${model.model_defaults.repeat}
|
||||
kernel: [15]
|
||||
stride: [1]
|
||||
dilation: [1]
|
||||
dropout: ${model.model_defaults.dropout}
|
||||
residual: true
|
||||
separable: ${model.model_defaults.separable}
|
||||
se: ${model.model_defaults.se}
|
||||
se_context_size: ${model.model_defaults.se_context_size}
|
||||
kernel_size_factor: ${model.model_defaults.kernel_size_factor}
|
||||
|
||||
- filters: 384
|
||||
repeat: ${model.model_defaults.repeat}
|
||||
kernel: [17]
|
||||
stride: [1]
|
||||
dilation: [1]
|
||||
dropout: ${model.model_defaults.dropout}
|
||||
residual: true
|
||||
separable: ${model.model_defaults.separable}
|
||||
se: ${model.model_defaults.se}
|
||||
se_context_size: ${model.model_defaults.se_context_size}
|
||||
kernel_size_factor: ${model.model_defaults.kernel_size_factor}
|
||||
|
||||
- filters: 384
|
||||
repeat: ${model.model_defaults.repeat}
|
||||
kernel: [19]
|
||||
stride: [1]
|
||||
dilation: [1]
|
||||
dropout: ${model.model_defaults.dropout}
|
||||
residual: true
|
||||
separable: ${model.model_defaults.separable}
|
||||
se: ${model.model_defaults.se}
|
||||
se_context_size: ${model.model_defaults.se_context_size}
|
||||
kernel_size_factor: ${model.model_defaults.kernel_size_factor}
|
||||
|
||||
- filters: 384
|
||||
repeat: ${model.model_defaults.repeat}
|
||||
kernel: [21]
|
||||
stride: [1]
|
||||
dilation: [1]
|
||||
dropout: ${model.model_defaults.dropout}
|
||||
residual: true
|
||||
separable: ${model.model_defaults.separable}
|
||||
se: ${model.model_defaults.se}
|
||||
se_context_size: ${model.model_defaults.se_context_size}
|
||||
kernel_size_factor: ${model.model_defaults.kernel_size_factor}
|
||||
|
||||
- filters: 384
|
||||
repeat: ${model.model_defaults.repeat}
|
||||
kernel: [23]
|
||||
stride: [1]
|
||||
dilation: [1]
|
||||
dropout: ${model.model_defaults.dropout}
|
||||
residual: true
|
||||
separable: ${model.model_defaults.separable}
|
||||
se: ${model.model_defaults.se}
|
||||
se_context_size: ${model.model_defaults.se_context_size}
|
||||
kernel_size_factor: ${model.model_defaults.kernel_size_factor}
|
||||
|
||||
- filters: 384
|
||||
repeat: ${model.model_defaults.repeat}
|
||||
kernel: [25]
|
||||
stride: [1]
|
||||
dilation: [1]
|
||||
dropout: ${model.model_defaults.dropout}
|
||||
residual: true
|
||||
separable: ${model.model_defaults.separable}
|
||||
se: ${model.model_defaults.se}
|
||||
se_context_size: ${model.model_defaults.se_context_size}
|
||||
kernel_size_factor: ${model.model_defaults.kernel_size_factor}
|
||||
|
||||
- filters: 384
|
||||
repeat: ${model.model_defaults.repeat}
|
||||
kernel: [25]
|
||||
stride: [2] # stride
|
||||
dilation: [1]
|
||||
dropout: ${model.model_defaults.dropout}
|
||||
residual: true
|
||||
separable: ${model.model_defaults.separable}
|
||||
se: ${model.model_defaults.se}
|
||||
se_context_size: ${model.model_defaults.se_context_size}
|
||||
stride_last: true
|
||||
residual_mode: "stride_add"
|
||||
kernel_size_factor: ${model.model_defaults.kernel_size_factor}
|
||||
|
||||
- filters: 384
|
||||
repeat: ${model.model_defaults.repeat}
|
||||
kernel: [27]
|
||||
stride: [1]
|
||||
dilation: [1]
|
||||
dropout: ${model.model_defaults.dropout}
|
||||
residual: true
|
||||
separable: ${model.model_defaults.separable}
|
||||
se: ${model.model_defaults.se}
|
||||
se_context_size: ${model.model_defaults.se_context_size}
|
||||
kernel_size_factor: ${model.model_defaults.kernel_size_factor}
|
||||
|
||||
- filters: 384
|
||||
repeat: ${model.model_defaults.repeat}
|
||||
kernel: [29]
|
||||
stride: [1]
|
||||
dilation: [1]
|
||||
dropout: ${model.model_defaults.dropout}
|
||||
residual: true
|
||||
separable: ${model.model_defaults.separable}
|
||||
se: ${model.model_defaults.se}
|
||||
se_context_size: ${model.model_defaults.se_context_size}
|
||||
kernel_size_factor: ${model.model_defaults.kernel_size_factor}
|
||||
|
||||
- filters: 384
|
||||
repeat: ${model.model_defaults.repeat}
|
||||
kernel: [31]
|
||||
stride: [1]
|
||||
dilation: [1]
|
||||
dropout: ${model.model_defaults.dropout}
|
||||
residual: true
|
||||
separable: ${model.model_defaults.separable}
|
||||
se: ${model.model_defaults.se}
|
||||
se_context_size: ${model.model_defaults.se_context_size}
|
||||
kernel_size_factor: ${model.model_defaults.kernel_size_factor}
|
||||
|
||||
- filters: 384
|
||||
repeat: ${model.model_defaults.repeat}
|
||||
kernel: [33]
|
||||
stride: [1]
|
||||
dilation: [1]
|
||||
dropout: ${model.model_defaults.dropout}
|
||||
residual: true
|
||||
separable: ${model.model_defaults.separable}
|
||||
se: ${model.model_defaults.se}
|
||||
se_context_size: ${model.model_defaults.se_context_size}
|
||||
kernel_size_factor: ${model.model_defaults.kernel_size_factor}
|
||||
|
||||
- filters: 384
|
||||
repeat: ${model.model_defaults.repeat}
|
||||
kernel: [35]
|
||||
stride: [1]
|
||||
dilation: [1]
|
||||
dropout: ${model.model_defaults.dropout}
|
||||
residual: true
|
||||
separable: ${model.model_defaults.separable}
|
||||
se: ${model.model_defaults.se}
|
||||
se_context_size: ${model.model_defaults.se_context_size}
|
||||
kernel_size_factor: ${model.model_defaults.kernel_size_factor}
|
||||
|
||||
- filters: 384
|
||||
repeat: ${model.model_defaults.repeat}
|
||||
kernel: [37]
|
||||
stride: [1]
|
||||
dilation: [1]
|
||||
dropout: ${model.model_defaults.dropout}
|
||||
residual: true
|
||||
separable: ${model.model_defaults.separable}
|
||||
se: ${model.model_defaults.se}
|
||||
se_context_size: ${model.model_defaults.se_context_size}
|
||||
kernel_size_factor: ${model.model_defaults.kernel_size_factor}
|
||||
|
||||
- filters: 384
|
||||
repeat: ${model.model_defaults.repeat}
|
||||
kernel: [39]
|
||||
stride: [1]
|
||||
dilation: [1]
|
||||
dropout: ${model.model_defaults.dropout}
|
||||
residual: true
|
||||
separable: ${model.model_defaults.separable}
|
||||
se: ${model.model_defaults.se}
|
||||
se_context_size: ${model.model_defaults.se_context_size}
|
||||
kernel_size_factor: ${model.model_defaults.kernel_size_factor}
|
||||
|
||||
- filters: &enc_final 640
|
||||
repeat: 1
|
||||
kernel: [41]
|
||||
stride: [1]
|
||||
dilation: [1]
|
||||
dropout: 0.0
|
||||
residual: false
|
||||
separable: ${model.model_defaults.separable}
|
||||
se: ${model.model_defaults.se}
|
||||
se_context_size: ${model.model_defaults.se_context_size}
|
||||
kernel_size_factor: ${model.model_defaults.kernel_size_factor}
|
||||
|
||||
|
||||
decoder:
|
||||
_target_: nemo.collections.asr.modules.ConvASRDecoder
|
||||
feat_in: *enc_final
|
||||
num_classes: -1 # filled with vocabulary size from tokenizer at runtime
|
||||
vocabulary: [] # filled with vocabulary from tokenizer at runtime
|
||||
|
||||
optim:
|
||||
name: novograd
|
||||
lr: 0.05
|
||||
|
||||
# optimizer arguments
|
||||
betas: [0.8, 0.25]
|
||||
weight_decay: 0.001
|
||||
|
||||
# scheduler setup
|
||||
sched:
|
||||
name: CosineAnnealing
|
||||
|
||||
# scheduler config override
|
||||
warmup_steps: 1000
|
||||
warmup_ratio: null
|
||||
min_lr: 1e-5
|
||||
last_epoch: -1
|
||||
|
||||
trainer:
|
||||
gpus: 0 # number of gpus
|
||||
max_epochs: 100
|
||||
max_steps: null # computed at runtime if not set
|
||||
num_nodes: 1
|
||||
accelerator: ddp
|
||||
accumulate_grad_batches: 1
|
||||
checkpoint_callback: false # Provided by exp_manager
|
||||
logger: false # Provided by exp_manager
|
||||
log_every_n_steps: 100 # Interval of logging.
|
||||
val_check_interval: 1.0 # Set to 0.25 to check 4 times per epoch, or an int for number of iterations
|
||||
check_val_every_n_epoch: 1
|
||||
precision: 32
|
||||
sync_batchnorm: false
|
||||
benchmark: false
|
||||
|
||||
exp_manager:
|
||||
exp_dir: null
|
||||
name: *name
|
||||
create_tensorboard_logger: true
|
||||
create_checkpoint_callback: true
|
||||
checkpoint_callback_params:
|
||||
monitor: "val_wer"
|
||||
mode: "min"
|
||||
save_top_k: 3
|
||||
create_wandb_logger: false
|
||||
wandb_logger_kwargs:
|
||||
name: null
|
||||
project: null
|
||||
entity: null
|
||||
resume_if_exists: false
|
||||
resume_ignore_no_checkpoint: false
|
||||
|
||||
hydra:
|
||||
run:
|
||||
dir: .
|
||||
job_logging:
|
||||
root:
|
||||
handlers: null
|
||||
@@ -0,0 +1,428 @@
|
||||
# This config contains the default values for training a Citrinet model with CTC loss and BPE-based vocabulary.
|
||||
# Default learning parameters in this config are set for effective batch size of 1k on 32 GPUs.
|
||||
# To train it with smaller batch sizes, you may need to re-tune the learning parameters or use higher accumulate_grad_batches.
|
||||
# If training for a short time, you can also reduce weight decay to 0.
|
||||
|
||||
# Training Recipe
|
||||
# This model can be trained using the default settings in this config with FP32 precision.
|
||||
# When training under AMP, increase `warmup_steps` to 5000 for stable training.
|
||||
# In order to create Citrinet-C, find-replace `filters: 384` with `filters: C`.
|
||||
# When reducing the receptive field of these models, it is advised to reduce the amount of augmentation
|
||||
# for larger models from 10x time masking to 5x or 2x time masking.
|
||||
# For further details regarding Citrinet, visit - https://docs.nvidia.com/deeplearning/nemo/user-guide/docs/en/main/asr/configs.html#citrinet
|
||||
|
||||
name: &name "Citrinet-512-8x-Stride"
|
||||
|
||||
model:
|
||||
sample_rate: &sample_rate 16000
|
||||
|
||||
train_ds:
|
||||
manifest_filepath: ???
|
||||
sample_rate: 16000
|
||||
batch_size: 32
|
||||
trim_silence: false
|
||||
max_duration: 16.7
|
||||
shuffle: true
|
||||
is_tarred: false
|
||||
tarred_audio_filepaths: null
|
||||
use_start_end_token: false
|
||||
|
||||
validation_ds:
|
||||
manifest_filepath: ???
|
||||
sample_rate: 16000
|
||||
batch_size: 32
|
||||
shuffle: false
|
||||
use_start_end_token: false
|
||||
|
||||
test_ds:
|
||||
manifest_filepath: null
|
||||
sample_rate: 16000
|
||||
batch_size: 32
|
||||
shuffle: false
|
||||
use_start_end_token: false
|
||||
|
||||
model_defaults:
|
||||
repeat: 5
|
||||
dropout: 0.1
|
||||
separable: true
|
||||
se: true
|
||||
se_context_size: -1
|
||||
kernel_size_factor: 1.0
|
||||
|
||||
tokenizer:
|
||||
dir: ??? # path to directory which contains either tokenizer.model (bpe) or vocab.txt (for wpe)
|
||||
type: ??? # Can be either bpe or wpe
|
||||
|
||||
preprocessor:
|
||||
_target_: nemo.collections.asr.modules.AudioToMelSpectrogramPreprocessor
|
||||
sample_rate: *sample_rate
|
||||
normalize: "per_feature"
|
||||
window_size: 0.025
|
||||
window_stride: 0.01
|
||||
window: "hann"
|
||||
features: &n_mels 80
|
||||
n_fft: 512
|
||||
frame_splicing: 1
|
||||
dither: 0.00001
|
||||
pad_to: 16
|
||||
stft_conv: false
|
||||
|
||||
spec_augment:
|
||||
_target_: nemo.collections.asr.modules.SpectrogramAugmentation
|
||||
freq_masks: 2
|
||||
time_masks: 10
|
||||
freq_width: 27
|
||||
time_width: 0.05
|
||||
|
||||
encoder:
|
||||
_target_: nemo.collections.asr.modules.ConvASREncoder
|
||||
feat_in: *n_mels
|
||||
activation: relu
|
||||
conv_mask: true
|
||||
|
||||
jasper:
|
||||
- filters: 512
|
||||
repeat: 1
|
||||
kernel: [5]
|
||||
stride: [1]
|
||||
dilation: [1]
|
||||
dropout: 0.0
|
||||
residual: false
|
||||
separable: ${model.model_defaults.separable}
|
||||
se: ${model.model_defaults.se}
|
||||
se_context_size: ${model.model_defaults.se_context_size}
|
||||
|
||||
- filters: 512
|
||||
repeat: ${model.model_defaults.repeat}
|
||||
kernel: [11]
|
||||
stride: [2]
|
||||
dilation: [1]
|
||||
dropout: ${model.model_defaults.dropout}
|
||||
residual: true
|
||||
separable: ${model.model_defaults.separable}
|
||||
se: ${model.model_defaults.se}
|
||||
se_context_size: ${model.model_defaults.se_context_size}
|
||||
stride_last: true
|
||||
residual_mode: "stride_add"
|
||||
kernel_size_factor: ${model.model_defaults.kernel_size_factor}
|
||||
|
||||
- filters: 512
|
||||
repeat: ${model.model_defaults.repeat}
|
||||
kernel: [13]
|
||||
stride: [1]
|
||||
dilation: [1]
|
||||
dropout: ${model.model_defaults.dropout}
|
||||
residual: true
|
||||
separable: ${model.model_defaults.separable}
|
||||
se: ${model.model_defaults.se}
|
||||
se_context_size: ${model.model_defaults.se_context_size}
|
||||
kernel_size_factor: ${model.model_defaults.kernel_size_factor}
|
||||
|
||||
- filters: 512
|
||||
repeat: ${model.model_defaults.repeat}
|
||||
kernel: [15]
|
||||
stride: [1]
|
||||
dilation: [1]
|
||||
dropout: ${model.model_defaults.dropout}
|
||||
residual: true
|
||||
separable: ${model.model_defaults.separable}
|
||||
se: ${model.model_defaults.se}
|
||||
se_context_size: ${model.model_defaults.se_context_size}
|
||||
kernel_size_factor: ${model.model_defaults.kernel_size_factor}
|
||||
|
||||
- filters: 512
|
||||
repeat: ${model.model_defaults.repeat}
|
||||
kernel: [17]
|
||||
stride: [1]
|
||||
dilation: [1]
|
||||
dropout: ${model.model_defaults.dropout}
|
||||
residual: true
|
||||
separable: ${model.model_defaults.separable}
|
||||
se: ${model.model_defaults.se}
|
||||
se_context_size: ${model.model_defaults.se_context_size}
|
||||
kernel_size_factor: ${model.model_defaults.kernel_size_factor}
|
||||
|
||||
- filters: 512
|
||||
repeat: ${model.model_defaults.repeat}
|
||||
kernel: [19]
|
||||
stride: [1]
|
||||
dilation: [1]
|
||||
dropout: ${model.model_defaults.dropout}
|
||||
residual: true
|
||||
separable: ${model.model_defaults.separable}
|
||||
se: ${model.model_defaults.se}
|
||||
se_context_size: ${model.model_defaults.se_context_size}
|
||||
kernel_size_factor: ${model.model_defaults.kernel_size_factor}
|
||||
|
||||
- filters: 512
|
||||
repeat: ${model.model_defaults.repeat}
|
||||
kernel: [21]
|
||||
stride: [1]
|
||||
dilation: [1]
|
||||
dropout: ${model.model_defaults.dropout}
|
||||
residual: true
|
||||
separable: ${model.model_defaults.separable}
|
||||
se: ${model.model_defaults.se}
|
||||
se_context_size: ${model.model_defaults.se_context_size}
|
||||
kernel_size_factor: ${model.model_defaults.kernel_size_factor}
|
||||
|
||||
- filters: 512
|
||||
repeat: ${model.model_defaults.repeat}
|
||||
kernel: [13]
|
||||
stride: [2] # *stride
|
||||
dilation: [1]
|
||||
dropout: ${model.model_defaults.dropout}
|
||||
residual: true
|
||||
separable: ${model.model_defaults.separable}
|
||||
se: ${model.model_defaults.se}
|
||||
se_context_size: ${model.model_defaults.se_context_size}
|
||||
stride_last: true
|
||||
residual_mode: "stride_add"
|
||||
kernel_size_factor: ${model.model_defaults.kernel_size_factor}
|
||||
|
||||
- filters: 512
|
||||
repeat: ${model.model_defaults.repeat}
|
||||
kernel: [15]
|
||||
stride: [1]
|
||||
dilation: [1]
|
||||
dropout: ${model.model_defaults.dropout}
|
||||
residual: true
|
||||
separable: ${model.model_defaults.separable}
|
||||
se: ${model.model_defaults.se}
|
||||
se_context_size: ${model.model_defaults.se_context_size}
|
||||
kernel_size_factor: ${model.model_defaults.kernel_size_factor}
|
||||
|
||||
- filters: 512
|
||||
repeat: ${model.model_defaults.repeat}
|
||||
kernel: [17]
|
||||
stride: [1]
|
||||
dilation: [1]
|
||||
dropout: ${model.model_defaults.dropout}
|
||||
residual: true
|
||||
separable: ${model.model_defaults.separable}
|
||||
se: ${model.model_defaults.se}
|
||||
se_context_size: ${model.model_defaults.se_context_size}
|
||||
kernel_size_factor: ${model.model_defaults.kernel_size_factor}
|
||||
|
||||
- filters: 512
|
||||
repeat: ${model.model_defaults.repeat}
|
||||
kernel: [19]
|
||||
stride: [1]
|
||||
dilation: [1]
|
||||
dropout: ${model.model_defaults.dropout}
|
||||
residual: true
|
||||
separable: ${model.model_defaults.separable}
|
||||
se: ${model.model_defaults.se}
|
||||
se_context_size: ${model.model_defaults.se_context_size}
|
||||
kernel_size_factor: ${model.model_defaults.kernel_size_factor}
|
||||
|
||||
- filters: 512
|
||||
repeat: ${model.model_defaults.repeat}
|
||||
kernel: [21]
|
||||
stride: [1]
|
||||
dilation: [1]
|
||||
dropout: ${model.model_defaults.dropout}
|
||||
residual: true
|
||||
separable: ${model.model_defaults.separable}
|
||||
se: ${model.model_defaults.se}
|
||||
se_context_size: ${model.model_defaults.se_context_size}
|
||||
kernel_size_factor: ${model.model_defaults.kernel_size_factor}
|
||||
|
||||
- filters: 512
|
||||
repeat: ${model.model_defaults.repeat}
|
||||
kernel: [23]
|
||||
stride: [1]
|
||||
dilation: [1]
|
||||
dropout: ${model.model_defaults.dropout}
|
||||
residual: true
|
||||
separable: ${model.model_defaults.separable}
|
||||
se: ${model.model_defaults.se}
|
||||
se_context_size: ${model.model_defaults.se_context_size}
|
||||
kernel_size_factor: ${model.model_defaults.kernel_size_factor}
|
||||
|
||||
- filters: 512
|
||||
repeat: ${model.model_defaults.repeat}
|
||||
kernel: [25]
|
||||
stride: [1]
|
||||
dilation: [1]
|
||||
dropout: ${model.model_defaults.dropout}
|
||||
residual: true
|
||||
separable: ${model.model_defaults.separable}
|
||||
se: ${model.model_defaults.se}
|
||||
se_context_size: ${model.model_defaults.se_context_size}
|
||||
kernel_size_factor: ${model.model_defaults.kernel_size_factor}
|
||||
|
||||
- filters: 512
|
||||
repeat: ${model.model_defaults.repeat}
|
||||
kernel: [25]
|
||||
stride: [2] # stride
|
||||
dilation: [1]
|
||||
dropout: ${model.model_defaults.dropout}
|
||||
residual: true
|
||||
separable: ${model.model_defaults.separable}
|
||||
se: ${model.model_defaults.se}
|
||||
se_context_size: ${model.model_defaults.se_context_size}
|
||||
stride_last: true
|
||||
residual_mode: "stride_add"
|
||||
kernel_size_factor: ${model.model_defaults.kernel_size_factor}
|
||||
|
||||
- filters: 512
|
||||
repeat: ${model.model_defaults.repeat}
|
||||
kernel: [27]
|
||||
stride: [1]
|
||||
dilation: [1]
|
||||
dropout: ${model.model_defaults.dropout}
|
||||
residual: true
|
||||
separable: ${model.model_defaults.separable}
|
||||
se: ${model.model_defaults.se}
|
||||
se_context_size: ${model.model_defaults.se_context_size}
|
||||
kernel_size_factor: ${model.model_defaults.kernel_size_factor}
|
||||
|
||||
- filters: 512
|
||||
repeat: ${model.model_defaults.repeat}
|
||||
kernel: [29]
|
||||
stride: [1]
|
||||
dilation: [1]
|
||||
dropout: ${model.model_defaults.dropout}
|
||||
residual: true
|
||||
separable: ${model.model_defaults.separable}
|
||||
se: ${model.model_defaults.se}
|
||||
se_context_size: ${model.model_defaults.se_context_size}
|
||||
kernel_size_factor: ${model.model_defaults.kernel_size_factor}
|
||||
|
||||
- filters: 512
|
||||
repeat: ${model.model_defaults.repeat}
|
||||
kernel: [31]
|
||||
stride: [1]
|
||||
dilation: [1]
|
||||
dropout: ${model.model_defaults.dropout}
|
||||
residual: true
|
||||
separable: ${model.model_defaults.separable}
|
||||
se: ${model.model_defaults.se}
|
||||
se_context_size: ${model.model_defaults.se_context_size}
|
||||
kernel_size_factor: ${model.model_defaults.kernel_size_factor}
|
||||
|
||||
- filters: 512
|
||||
repeat: ${model.model_defaults.repeat}
|
||||
kernel: [33]
|
||||
stride: [1]
|
||||
dilation: [1]
|
||||
dropout: ${model.model_defaults.dropout}
|
||||
residual: true
|
||||
separable: ${model.model_defaults.separable}
|
||||
se: ${model.model_defaults.se}
|
||||
se_context_size: ${model.model_defaults.se_context_size}
|
||||
kernel_size_factor: ${model.model_defaults.kernel_size_factor}
|
||||
|
||||
- filters: 512
|
||||
repeat: ${model.model_defaults.repeat}
|
||||
kernel: [35]
|
||||
stride: [1]
|
||||
dilation: [1]
|
||||
dropout: ${model.model_defaults.dropout}
|
||||
residual: true
|
||||
separable: ${model.model_defaults.separable}
|
||||
se: ${model.model_defaults.se}
|
||||
se_context_size: ${model.model_defaults.se_context_size}
|
||||
kernel_size_factor: ${model.model_defaults.kernel_size_factor}
|
||||
|
||||
- filters: 512
|
||||
repeat: ${model.model_defaults.repeat}
|
||||
kernel: [37]
|
||||
stride: [1]
|
||||
dilation: [1]
|
||||
dropout: ${model.model_defaults.dropout}
|
||||
residual: true
|
||||
separable: ${model.model_defaults.separable}
|
||||
se: ${model.model_defaults.se}
|
||||
se_context_size: ${model.model_defaults.se_context_size}
|
||||
kernel_size_factor: ${model.model_defaults.kernel_size_factor}
|
||||
|
||||
- filters: 512
|
||||
repeat: ${model.model_defaults.repeat}
|
||||
kernel: [39]
|
||||
stride: [1]
|
||||
dilation: [1]
|
||||
dropout: ${model.model_defaults.dropout}
|
||||
residual: true
|
||||
separable: ${model.model_defaults.separable}
|
||||
se: ${model.model_defaults.se}
|
||||
se_context_size: ${model.model_defaults.se_context_size}
|
||||
kernel_size_factor: ${model.model_defaults.kernel_size_factor}
|
||||
|
||||
- filters: &enc_final 640
|
||||
repeat: 1
|
||||
kernel: [41]
|
||||
stride: [1]
|
||||
dilation: [1]
|
||||
dropout: 0.0
|
||||
residual: false
|
||||
separable: ${model.model_defaults.separable}
|
||||
se: ${model.model_defaults.se}
|
||||
se_context_size: ${model.model_defaults.se_context_size}
|
||||
kernel_size_factor: ${model.model_defaults.kernel_size_factor}
|
||||
|
||||
|
||||
decoder:
|
||||
_target_: nemo.collections.asr.modules.ConvASRDecoder
|
||||
feat_in: *enc_final
|
||||
num_classes: -1 # filled with vocabulary size from tokenizer at runtime
|
||||
vocabulary: [] # filled with vocabulary from tokenizer at runtime
|
||||
|
||||
optim:
|
||||
name: novograd
|
||||
lr: 0.05
|
||||
|
||||
# optimizer arguments
|
||||
betas: [0.8, 0.25]
|
||||
weight_decay: 0.001
|
||||
|
||||
# scheduler setup
|
||||
sched:
|
||||
name: CosineAnnealing
|
||||
|
||||
# scheduler config override
|
||||
warmup_steps: 1000
|
||||
warmup_ratio: null
|
||||
min_lr: 1e-5
|
||||
last_epoch: -1
|
||||
|
||||
trainer:
|
||||
gpus: 0 # number of gpus
|
||||
max_epochs: 100
|
||||
max_steps: null # computed at runtime if not set
|
||||
num_nodes: 1
|
||||
accelerator: ddp
|
||||
accumulate_grad_batches: 1
|
||||
checkpoint_callback: false # Provided by exp_manager
|
||||
logger: false # Provided by exp_manager
|
||||
log_every_n_steps: 100 # Interval of logging.
|
||||
val_check_interval: 1.0 # Set to 0.25 to check 4 times per epoch, or an int for number of iterations
|
||||
check_val_every_n_epoch: 1
|
||||
precision: 32
|
||||
sync_batchnorm: false
|
||||
benchmark: false
|
||||
|
||||
exp_manager:
|
||||
exp_dir: null
|
||||
name: *name
|
||||
create_tensorboard_logger: true
|
||||
create_checkpoint_callback: true
|
||||
checkpoint_callback_params:
|
||||
monitor: "val_wer"
|
||||
mode: "min"
|
||||
save_top_k: 3
|
||||
create_wandb_logger: false
|
||||
wandb_logger_kwargs:
|
||||
name: null
|
||||
project: null
|
||||
entity: null
|
||||
resume_if_exists: false
|
||||
resume_ignore_no_checkpoint: false
|
||||
|
||||
hydra:
|
||||
run:
|
||||
dir: .
|
||||
job_logging:
|
||||
root:
|
||||
handlers: null
|
||||
@@ -0,0 +1,183 @@
|
||||
name: &name "ContextNet5x1"
|
||||
sample_rate: &sample_rate 16000
|
||||
repeat: &repeat 1
|
||||
dropout: &dropout 0.0
|
||||
separable: &separable true
|
||||
|
||||
model:
|
||||
train_ds:
|
||||
manifest_filepath: ???
|
||||
sample_rate: 16000
|
||||
batch_size: 32
|
||||
trim_silence: True
|
||||
max_duration: 16.7
|
||||
shuffle: True
|
||||
is_tarred: False
|
||||
tarred_audio_filepaths: null
|
||||
shard_strategy: "scatter"
|
||||
|
||||
validation_ds:
|
||||
manifest_filepath: ???
|
||||
sample_rate: 16000
|
||||
batch_size: 32
|
||||
shuffle: False
|
||||
|
||||
tokenizer:
|
||||
dir: ??? # path to directory which contains either tokenizer.model (bpe) or vocab.txt (for wpe)
|
||||
type: ??? # Can be either bpe or wpe
|
||||
|
||||
preprocessor:
|
||||
_target_: nemo.collections.asr.modules.AudioToMelSpectrogramPreprocessor
|
||||
normalize: "per_feature"
|
||||
window_size: 0.02
|
||||
sample_rate: *sample_rate
|
||||
window_stride: 0.01
|
||||
window: "hann"
|
||||
features: &n_mels 64
|
||||
n_fft: 512
|
||||
frame_splicing: 1
|
||||
dither: 0.00001
|
||||
|
||||
spec_augment:
|
||||
_target_: nemo.collections.asr.modules.SpectrogramAugmentation
|
||||
rect_freq: 50
|
||||
rect_masks: 5
|
||||
rect_time: 120
|
||||
|
||||
encoder:
|
||||
_target_: nemo.collections.asr.modules.ConvASREncoder
|
||||
feat_in: *n_mels
|
||||
activation: relu
|
||||
conv_mask: true
|
||||
|
||||
jasper:
|
||||
- filters: 128
|
||||
repeat: 1
|
||||
kernel: [11]
|
||||
stride: [1]
|
||||
dilation: [1]
|
||||
dropout: *dropout
|
||||
residual: true
|
||||
separable: *separable
|
||||
se: true
|
||||
se_context_size: -1
|
||||
|
||||
- filters: 256
|
||||
repeat: *repeat
|
||||
kernel: [13]
|
||||
stride: [1]
|
||||
dilation: [1]
|
||||
dropout: *dropout
|
||||
residual: true
|
||||
separable: *separable
|
||||
se: true
|
||||
se_context_size: -1
|
||||
|
||||
- filters: 256
|
||||
repeat: *repeat
|
||||
kernel: [15]
|
||||
stride: [1]
|
||||
dilation: [1]
|
||||
dropout: *dropout
|
||||
residual: true
|
||||
separable: *separable
|
||||
se: true
|
||||
se_context_size: -1
|
||||
|
||||
- filters: 256
|
||||
repeat: *repeat
|
||||
kernel: [17]
|
||||
stride: [1]
|
||||
dilation: [1]
|
||||
dropout: *dropout
|
||||
residual: true
|
||||
separable: *separable
|
||||
se: true
|
||||
se_context_size: -1
|
||||
|
||||
- filters: 256
|
||||
repeat: *repeat
|
||||
kernel: [19]
|
||||
stride: [1]
|
||||
dilation: [1]
|
||||
dropout: *dropout
|
||||
residual: true
|
||||
separable: *separable
|
||||
se: true
|
||||
se_context_size: -1
|
||||
|
||||
- filters: 256
|
||||
repeat: 1
|
||||
kernel: [21]
|
||||
stride: [1]
|
||||
dilation: [1]
|
||||
dropout: 0.0
|
||||
residual: false
|
||||
separable: *separable
|
||||
se: true
|
||||
se_context_size: -1
|
||||
|
||||
- filters: &enc_feat_out 1024
|
||||
repeat: 1
|
||||
kernel: [1]
|
||||
stride: [1]
|
||||
dilation: [1]
|
||||
dropout: 0.0
|
||||
residual: false
|
||||
separable: *separable
|
||||
se: true
|
||||
se_context_size: -1
|
||||
|
||||
decoder:
|
||||
_target_: nemo.collections.asr.modules.ConvASRDecoder
|
||||
feat_in: 1024
|
||||
num_classes: -1 # filled with vocabulary size from tokenizer at runtime
|
||||
vocabulary: [] # filled with vocabulary from tokenizer at runtime
|
||||
|
||||
optim:
|
||||
name: adam
|
||||
# _target_: nemo.core.optim.optimizers.Adam
|
||||
lr: .1
|
||||
|
||||
# optimizer arguments
|
||||
betas: [0.9, 0.999]
|
||||
weight_decay: 0.0001
|
||||
|
||||
# scheduler setup
|
||||
sched:
|
||||
name: CosineAnnealing
|
||||
|
||||
# scheduler config override
|
||||
warmup_steps: null
|
||||
warmup_ratio: 0.05
|
||||
min_lr: 1e-6
|
||||
last_epoch: -1
|
||||
|
||||
trainer:
|
||||
gpus: 0 # number of gpus
|
||||
max_epochs: 5
|
||||
max_steps: null # computed at runtime if not set
|
||||
num_nodes: 1
|
||||
accelerator: ddp
|
||||
accumulate_grad_batches: 1
|
||||
checkpoint_callback: False # Provided by exp_manager
|
||||
logger: False # Provided by exp_manager
|
||||
log_every_n_steps: 1 # Interval of logging.
|
||||
val_check_interval: 1.0 # Set to 0.25 to check 4 times per epoch, or an int for number of iterations
|
||||
|
||||
exp_manager:
|
||||
exp_dir: null
|
||||
name: *name
|
||||
create_tensorboard_logger: True
|
||||
create_checkpoint_callback: True
|
||||
create_wandb_logger: False
|
||||
wandb_logger_kwargs:
|
||||
name: null
|
||||
project: null
|
||||
|
||||
hydra:
|
||||
run:
|
||||
dir: .
|
||||
job_logging:
|
||||
root:
|
||||
handlers: null
|
||||
Reference in New Issue
Block a user