This project implements a method for faster and more memory-efficient RNN-T loss computation, called pruned rnnt.

Note: There is also a fast RNN-T loss implementation in k2 project, which shares the same code here. We make fast_rnnt a stand-alone project in case someone wants only this rnnt loss.

We first obtain pruning bounds for the RNN-T recursion using a simple joiner network that is just an addition of the encoder and decoder, then we use those pruning bounds to evaluate the full, non-linear joiner network.

The picture below display the gradients (obtained by rnnt_loss_simple with return_grad=true) of lattice nodes, at each time frame, only a small set of nodes have a non-zero gradient, which justifies the pruned RNN-T loss, i.e., putting a limit on the number of symbols per frame.

This picture is taken from here

You can install it via pip:

pip install fast_rnnt

You can also install from source:

git clone https://github.com/danpovey/fast_rnnt.git
cd fast_rnnt
python setup.py install

To check that fast_rnnt was installed successfully, please run

python3 -c "import fast_rnnt; print(fast_rnnt.__version__)"

which should print the version of the installed fast_rnnt, e.g., 1.0.

Use

pip install --verbose fast_rnnt

Use

export FT_MAKE_ARGS="-j"
pip install --verbose fast_rnnt

It will pass -j to make.

It has been tested on PyTorch >= 1.5.0.

Note: The cuda version of the Pytorch should be the same as the cuda version in your environment, or it will cause a compilation error.

Use

export FT_CMAKE_ARGS="-DCMAKE_BUILD_TYPE=Release -DFT_WITH_CUDA=OFF"
export FT_MAKE_ARGS="-j"
pip install --verbose fast_rnnt

It will pass -DCMAKE_BUILD_TYPE=Release -DFT_WITH_CUDA=OFF to cmake.

Please file an issue at https://github.com/danpovey/fast_rnnt/issues and describe your problem there.

This is a simple case of the RNN-T loss, where the joiner network is just addition.

Note: termination_symbol plays the role of blank in other RNN-T loss implementations, we call it termination_symbol as it terminates symbols of current frame.

am = torch.randn((B, T, C), dtype=torch.float32)
lm = torch.randn((B, S + 1, C), dtype=torch.float32)
symbols = torch.randint(0, C, (B, S))
termination_symbol = 0

boundary = torch.zeros((B, 4), dtype=torch.int64)
boundary[:, 2] = target_lengths
boundary[:, 3] = num_frames

loss = fast_rnnt.rnnt_loss_simple(
    lm=lm,
    am=am,
    symbols=symbols,
    termination_symbol=termination_symbol,
    boundary=boundary,
    reduction="sum",
)

The same as rnnt_loss_simple, except that it supports am_only & lm_only smoothing that allows you to make the loss-function one of the form:

      lm_only_scale * lm_probs +
      am_only_scale * am_probs +
      (1-lm_only_scale-am_only_scale) * combined_probs

where lm_probs and am_probs are the probabilities given the lm and acoustic model independently.

am = torch.randn((B, T, C), dtype=torch.float32)
lm = torch.randn((B, S + 1, C), dtype=torch.float32)
symbols = torch.randint(0, C, (B, S))
termination_symbol = 0

boundary = torch.zeros((B, 4), dtype=torch.int64)
boundary[:, 2] = target_lengths
boundary[:, 3] = num_frames

loss = fast_rnnt.rnnt_loss_smoothed(
    lm=lm,
    am=am,
    symbols=symbols,
    termination_symbol=termination_symbol,
    lm_only_scale=0.25,
    am_only_scale=0.0
    boundary=boundary,
    reduction="sum",
)

rnnt_loss_pruned can not be used alone, it needs the gradients returned by rnnt_loss_simple/rnnt_loss_smoothed to get pruning bounds.

am = torch.randn((B, T, C), dtype=torch.float32)
lm = torch.randn((B, S + 1, C), dtype=torch.float32)
symbols = torch.randint(0, C, (B, S))
termination_symbol = 0

boundary = torch.zeros((B, 4), dtype=torch.int64)
boundary[:, 2] = target_lengths
boundary[:, 3] = num_frames

# rnnt_loss_simple can be also rnnt_loss_smoothed
simple_loss, (px_grad, py_grad) = fast_rnnt.rnnt_loss_simple(
    lm=lm,
    am=am,
    symbols=symbols,
    termination_symbol=termination_symbol,
    boundary=boundary,
    reduction="sum",
    return_grad=True,
)
s_range = 5  # can be other values
ranges = fast_rnnt.get_rnnt_prune_ranges(
    px_grad=px_grad,
    py_grad=py_grad,
    boundary=boundary,
    s_range=s_range,
)

am_pruned, lm_pruned = fast_rnnt.do_rnnt_pruning(am=am, lm=lm, ranges=ranges)

logits = model.joiner(am_pruned, lm_pruned)
pruned_loss = fast_rnnt.rnnt_loss_pruned(
    logits=logits,
    symbols=symbols,
    ranges=ranges,
    termination_symbol=termination_symbol,
    boundary=boundary,
    reduction="sum",
)

You can also find recipes here that uses rnnt_loss_pruned to train a model.

The unprund rnnt_loss is the same as torchaudio rnnt_loss, it produces same output as torchaudio for the same input.

logits = torch.randn((B, S, T, C), dtype=torch.float32)
symbols = torch.randint(0, C, (B, S))
termination_symbol = 0

boundary = torch.zeros((B, 4), dtype=torch.int64)
boundary[:, 2] = target_lengths
boundary[:, 3] = num_frames

loss = fast_rnnt.rnnt_loss(
    logits=logits,
    symbols=symbols,
    termination_symbol=termination_symbol,
    boundary=boundary,
    reduction="sum",
)

The repo compares the speed and memory usage of several transducer losses, the summary in the following table is taken from there, you can check the repository for more details.

Note: As we declared above, fast_rnnt is also implemented in k2 project, so k2 and fast_rnnt are equivalent in the benchmarking.