File size: 8,432 Bytes
0b8c0e5 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 | #!/usr/bin/env python3
"""Export Parakeet Realtime EOU 120M RNNT components into CoreML."""
from __future__ import annotations
from dataclasses import dataclass
from pathlib import Path
from typing import Optional, Tuple
import coremltools as ct
import torch
@dataclass
class ExportSettings:
output_dir: Path
compute_units: ct.ComputeUnit
deployment_target: Optional[ct.target]
compute_precision: Optional[ct.precision]
max_audio_seconds: float
max_symbol_steps: int
class PreprocessorWrapper(torch.nn.Module):
"""Wrapper for the audio preprocessor (mel spectrogram extraction)."""
def __init__(self, module: torch.nn.Module) -> None:
super().__init__()
self.module = module
def forward(
self, audio_signal: torch.Tensor, length: torch.Tensor
) -> Tuple[torch.Tensor, torch.Tensor]:
mel, mel_length = self.module(
input_signal=audio_signal, length=length.to(dtype=torch.long)
)
return mel, mel_length
class EncoderWrapper(torch.nn.Module):
"""Wrapper for the cache-aware FastConformer encoder.
Note: For the realtime EOU model, the encoder is cache-aware which means
it can operate in a streaming fashion. For CoreML export, we export
without cache state for simplicity (full-context mode).
"""
def __init__(self, module: torch.nn.Module) -> None:
super().__init__()
self.module = module
def forward(
self, features: torch.Tensor, length: torch.Tensor
) -> Tuple[torch.Tensor, torch.Tensor, torch.Tensor]:
encoded, encoded_lengths = self.module(
audio_signal=features, length=length.to(dtype=torch.long)
)
# Synthesize per-frame timestamps (seconds) using the 80 ms encoder stride.
# Shape: [B, T_enc]
frame_times = (
torch.arange(encoded.shape[-1], device=encoded.device, dtype=torch.float32)
* 0.08
)
return encoded, encoded_lengths, frame_times
class DecoderWrapper(torch.nn.Module):
"""Wrapper for the RNNT prediction network (decoder)."""
def __init__(self, module: torch.nn.Module) -> None:
super().__init__()
self.module = module
def forward(
self,
targets: torch.Tensor,
target_lengths: torch.Tensor,
h_in: torch.Tensor,
c_in: torch.Tensor,
) -> Tuple[torch.Tensor, torch.Tensor, torch.Tensor]:
state = [h_in, c_in]
decoder_output, _, new_state = self.module(
targets=targets.to(dtype=torch.long),
target_length=target_lengths.to(dtype=torch.long),
states=state,
)
return decoder_output, new_state[0], new_state[1]
class JointWrapper(torch.nn.Module):
"""Wrapper for the RNNT joint network.
Note: Unlike Parakeet TDT v3, the realtime EOU model does NOT have
duration outputs (num_extra_outputs). The joint network outputs only
token logits over the vocabulary + blank.
"""
def __init__(self, module: torch.nn.Module) -> None:
super().__init__()
self.module = module
def forward(
self, encoder_outputs: torch.Tensor, decoder_outputs: torch.Tensor
) -> torch.Tensor:
# Input: encoder_outputs [B, D, T], decoder_outputs [B, D, U]
# Transpose to match what projection layers expect
encoder_outputs = encoder_outputs.transpose(1, 2) # [B, T, D]
decoder_outputs = decoder_outputs.transpose(1, 2) # [B, U, D]
# Apply projections
enc_proj = self.module.enc(encoder_outputs) # [B, T, joint_hidden]
dec_proj = self.module.pred(decoder_outputs) # [B, U, joint_hidden]
# Explicit broadcasting along T and U to avoid converter ambiguity
x = enc_proj.unsqueeze(2) + dec_proj.unsqueeze(1) # [B, T, U, joint_hidden]
x = self.module.joint_net[0](x) # ReLU
x = self.module.joint_net[1](x) # Dropout (no-op in eval)
out = self.module.joint_net[2](x) # Linear -> logits [B, T, U, vocab+blank]
return out
class MelEncoderWrapper(torch.nn.Module):
"""Fused wrapper: waveform -> mel -> encoder.
Inputs:
- audio_signal: [B, S]
- audio_length: [B]
Outputs:
- encoder: [B, D, T_enc]
- encoder_length: [B]
- frame_times: [T_enc]
"""
def __init__(
self, preprocessor: PreprocessorWrapper, encoder: EncoderWrapper
) -> None:
super().__init__()
self.preprocessor = preprocessor
self.encoder = encoder
def forward(
self, audio_signal: torch.Tensor, audio_length: torch.Tensor
) -> Tuple[torch.Tensor, torch.Tensor, torch.Tensor]:
mel, mel_length = self.preprocessor(audio_signal, audio_length)
encoded, enc_len, frame_times = self.encoder(mel, mel_length.to(dtype=torch.int32))
return encoded, enc_len, frame_times
class JointDecisionWrapper(torch.nn.Module):
"""Joint + decision head: outputs label id and label prob.
Unlike Parakeet TDT v3, this model does NOT have duration outputs.
Inputs:
- encoder_outputs: [B, D, T]
- decoder_outputs: [B, D, U]
Returns:
- token_id: [B, T, U] int32
- token_prob: [B, T, U] float32
"""
def __init__(self, joint: JointWrapper, vocab_size: int) -> None:
super().__init__()
self.joint = joint
self.vocab_with_blank = int(vocab_size) + 1
def forward(self, encoder_outputs: torch.Tensor, decoder_outputs: torch.Tensor):
logits = self.joint(encoder_outputs, decoder_outputs)
# Token selection
token_ids = torch.argmax(logits, dim=-1).to(dtype=torch.int32)
token_probs_all = torch.softmax(logits, dim=-1)
# gather expects int64 (long) indices; cast only for gather
token_prob = torch.gather(
token_probs_all, dim=-1, index=token_ids.long().unsqueeze(-1)
).squeeze(-1)
return token_ids, token_prob
class JointDecisionSingleStep(torch.nn.Module):
"""Single-step variant for streaming: encoder_step -> token decision.
Inputs:
- encoder_step: [B=1, D, T=1]
- decoder_step: [B=1, D, U=1]
Returns:
- token_id: [1, 1, 1] int32
- token_prob: [1, 1, 1] float32
- top_k_ids: [1, 1, 1, K] int32
- top_k_logits: [1, 1, 1, K] float32
"""
def __init__(self, joint: JointWrapper, vocab_size: int, top_k: int = 64) -> None:
super().__init__()
self.joint = joint
self.vocab_with_blank = int(vocab_size) + 1
self.top_k = int(top_k)
def forward(self, encoder_step: torch.Tensor, decoder_step: torch.Tensor):
# Reuse JointWrapper which expects [B, D, T] and [B, D, U]
logits = self.joint(encoder_step, decoder_step) # [1, 1, 1, V+blank]
token_ids = torch.argmax(logits, dim=-1, keepdim=False).to(dtype=torch.int32)
token_probs_all = torch.softmax(logits, dim=-1)
token_prob = torch.gather(
token_probs_all, dim=-1, index=token_ids.long().unsqueeze(-1)
).squeeze(-1)
# Also expose top-K candidates for host-side processing
topk_logits, topk_ids_long = torch.topk(
logits, k=min(self.top_k, logits.shape[-1]), dim=-1
)
topk_ids = topk_ids_long.to(dtype=torch.int32)
return token_ids, token_prob, topk_ids, topk_logits
def _coreml_convert(
traced: torch.jit.ScriptModule,
inputs,
outputs,
settings: ExportSettings,
compute_units_override: Optional[ct.ComputeUnit] = None,
compute_precision: Optional[ct.precision] = None,
) -> ct.models.MLModel:
cu = (
compute_units_override
if compute_units_override is not None
else settings.compute_units
)
kwargs = {
"convert_to": "mlprogram",
"inputs": inputs,
"outputs": outputs,
"compute_units": cu,
}
print("Converting:", traced.__class__.__name__)
print("Conversion kwargs:", kwargs)
if settings.deployment_target is not None:
kwargs["minimum_deployment_target"] = settings.deployment_target
# Priority: explicit argument > settings
if compute_precision is not None:
kwargs["compute_precision"] = compute_precision
elif settings.compute_precision is not None:
kwargs["compute_precision"] = settings.compute_precision
return ct.convert(traced, **kwargs)
|