|
|
import os |
|
|
import pyworld as pw |
|
|
import math |
|
|
import warnings |
|
|
import logging |
|
|
import torch |
|
|
import torchaudio |
|
|
import torch.nn.functional as F |
|
|
import torch.nn.init as init |
|
|
from torch import nn, Tensor |
|
|
|
|
|
import matplotlib.pyplot as plt |
|
|
from typing import Optional, Dict, Union, List, Tuple, Any |
|
|
import numpy as np |
|
|
from functools import partial |
|
|
from datetime import datetime |
|
|
from datasets import load_dataset, Audio |
|
|
from transformers.trainer_seq2seq import Seq2SeqTrainer |
|
|
from transformers.training_args_seq2seq import Seq2SeqTrainingArguments |
|
|
import transformers |
|
|
from dataclasses import dataclass |
|
|
from opimizer import MaxFactor |
|
|
from transformers.generation.configuration_utils import GenerationConfig |
|
|
torch.backends.cudnn.allow_tf32 = True |
|
|
torch.backends.cuda.matmul.allow_tf32 = True |
|
|
torch.set_float32_matmul_precision('high') |
|
|
transformers.utils.logging.set_verbosity_error() |
|
|
|
|
|
device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu") |
|
|
dtype = torch.float32 |
|
|
|
|
|
warnings.filterwarnings("ignore") |
|
|
logging.basicConfig(level=logging.ERROR) |
|
|
|
|
|
PATH = 'E:/hf' |
|
|
os.environ['HF_HOME'] = PATH |
|
|
os.environ['HF_DATASETS_CACHE'] = PATH |
|
|
os.environ['TORCH_HOME'] = PATH |
|
|
os.environ['TF_ENABLE_ONEDNN_OPTS'] = '0' |
|
|
os.environ["PYTORCH_CUDA_ALLOC_CONF"] = "expandable_segments:True" |
|
|
|
|
|
def get_activation(act: str) -> nn.Module: |
|
|
"""Get activation function by name.""" |
|
|
act_map = { |
|
|
"gelu": nn.GELU(), |
|
|
"relu": nn.ReLU(), |
|
|
"sigmoid": nn.Sigmoid(), |
|
|
"tanh": nn.Tanh(), |
|
|
"swish": nn.SiLU(), |
|
|
"tanhshrink": nn.Tanhshrink(), |
|
|
"softplus": nn.Softplus(), |
|
|
"softshrink": nn.Softshrink(), |
|
|
"leaky_relu": nn.LeakyReLU(), |
|
|
"elu": nn.ELU() |
|
|
} |
|
|
return act_map.get(act, nn.GELU()) |
|
|
|
|
|
@dataclass |
|
|
class Dimensions: |
|
|
vocab: int |
|
|
mels: int |
|
|
ctx: int |
|
|
dims: int |
|
|
head: int |
|
|
layer: int |
|
|
act: str |
|
|
debug: List[str] |
|
|
features: List[str] |
|
|
|
|
|
def get_generation_config(param): |
|
|
return GenerationConfig( |
|
|
max_length=param.text_ctx, |
|
|
pad_token_id=getattr(param, "pad_token_id", 0), |
|
|
bos_token_id=getattr(param, "bos_token_id", 1), |
|
|
eos_token_id=getattr(param, "eos_token_id", 2), |
|
|
do_sample=False, |
|
|
num_beams=1, |
|
|
early_stopping=False, |
|
|
length_penalty=1.0, |
|
|
no_repeat_ngram_size=0, |
|
|
repetition_penalty=1.0, |
|
|
temperature=1.0, |
|
|
decoder_start_token_id=1, |
|
|
is_multilingual=False, |
|
|
use_cache=False, |
|
|
return_timestamps=False) |
|
|
|
|
|
def plot_waveform(x=None, w=None, p=None, per=None, sample_idx=0, sr=16000, hop_length=160, |
|
|
title="", markers=None, marker_labels=None, |
|
|
show_voiced_regions=True, show_energy=False): |
|
|
num_plots = sum([x is not None, w is not None, p is not None, per is not None]) |
|
|
if num_plots == 0: |
|
|
raise ValueError("No data to plot. Please provide at least one input tensor.") |
|
|
t_spans = [] |
|
|
|
|
|
if w is not None: |
|
|
w_np = w[sample_idx].detach().cpu().numpy() |
|
|
if w_np.ndim > 1: |
|
|
w_np = w_np.squeeze() |
|
|
t_spans.append(len(w_np) / sr) |
|
|
if x is not None: |
|
|
x_np = x[sample_idx].detach().cpu().numpy() |
|
|
if x_np.shape[0] < x_np.shape[1]: |
|
|
x_np = x_np.T |
|
|
t_spans.append(x_np.shape[0] * hop_length / sr) |
|
|
if p is not None: |
|
|
p_np = p[sample_idx].detach().cpu().numpy() |
|
|
if p_np.ndim > 1: |
|
|
p_np = p_np.squeeze() |
|
|
t_spans.append(len(p_np) * hop_length / sr) |
|
|
if per is not None: |
|
|
per_np = per[sample_idx].detach().cpu().numpy() |
|
|
if per_np.ndim > 1: |
|
|
per_np = per_np.squeeze() |
|
|
t_spans.append(len(per_np) * hop_length / sr) |
|
|
max_t = max(t_spans) if t_spans else 0 |
|
|
fig, axs = plt.subplots(num_plots, 1, figsize=(14, 4*num_plots), sharex=True) |
|
|
if num_plots == 1: |
|
|
axs = [axs] |
|
|
if show_voiced_regions and per is not None: |
|
|
per_np = per[sample_idx].detach().cpu().numpy() |
|
|
if per_np.ndim > 1: |
|
|
per_np = per_np.squeeze() |
|
|
t_per = np.arange(len(per_np)) * hop_length / sr |
|
|
threshold = 0.5 |
|
|
for ax in axs: |
|
|
for i in range(len(per_np)-1): |
|
|
if per_np[i] > threshold: |
|
|
ax.axvspan(t_per[i], t_per[i+1], color='lightblue', alpha=0.2, zorder=0) |
|
|
cu_ax = 0 |
|
|
if w is not None: |
|
|
w_np = w[sample_idx].detach().cpu().numpy() |
|
|
if w_np.ndim > 1: |
|
|
w_np = w_np.squeeze() |
|
|
t = np.arange(len(w_np)) / sr |
|
|
axs[cu_ax].plot(t, w_np, color="tab:blue") |
|
|
|
|
|
if show_energy: |
|
|
frame_length = hop_length |
|
|
hop_length_energy = hop_length // 2 |
|
|
energy = [] |
|
|
for i in range(0, len(w_np)-frame_length, hop_length_energy): |
|
|
frame = w_np[i:i+frame_length] |
|
|
energy.append(np.sqrt(np.mean(frame**2))) |
|
|
energy = np.array(energy) |
|
|
energy = energy / np.max(energy) * 0.8 * max(abs(w_np.min()), abs(w_np.max())) |
|
|
t_energy = np.arange(len(energy)) * hop_length_energy / sr |
|
|
axs[cu_ax].plot(t_energy, energy, color="red", alpha=0.7, label="Energy") |
|
|
axs[cu_ax].legend(loc='upper right') |
|
|
axs[cu_ax].set_title("Waveform") |
|
|
axs[cu_ax].set_ylabel("Amplitude") |
|
|
axs[cu_ax].set_xlim([0, max_t]) |
|
|
axs[cu_ax].grid(True, axis='x', linestyle='--', alpha=0.3) |
|
|
cu_ax += 1 |
|
|
|
|
|
if x is not None: |
|
|
x_np = x[sample_idx].detach().cpu().numpy() |
|
|
if x_np.shape[0] < x_np.shape[1]: |
|
|
x_np = x_np.T |
|
|
axs[cu_ax].imshow(x_np.T, aspect="auto", origin="lower", cmap="magma", |
|
|
extent=[0, x_np.shape[0]*hop_length/sr, 0, x_np.shape[1]]) |
|
|
axs[cu_ax].set_title("Spectrogram") |
|
|
axs[cu_ax].set_ylabel("Mel Bin") |
|
|
axs[cu_ax].set_xlim([0, max_t]) |
|
|
axs[cu_ax].grid(True, axis='x', linestyle='--', alpha=0.3) |
|
|
cu_ax += 1 |
|
|
|
|
|
if p is not None: |
|
|
p_np = p[sample_idx].detach().cpu().numpy() |
|
|
if p_np.ndim > 1: |
|
|
p_np = p_np.squeeze() |
|
|
t_p = np.arange(len(p_np)) * hop_length / sr |
|
|
axs[cu_ax].plot(t_p, p_np, color="tab:green") |
|
|
axs[cu_ax].set_title("Pitch") |
|
|
axs[cu_ax].set_ylabel("Frequency (Hz)") |
|
|
axs[cu_ax].set_xlim([0, max_t]) |
|
|
axs[cu_ax].grid(True, axis='both', linestyle='--', alpha=0.3) |
|
|
axs[cu_ax].set_ylim([0, min(1000, p_np.max() * 1.2)]) |
|
|
cu_ax += 1 |
|
|
|
|
|
if per is not None: |
|
|
per_np = per[sample_idx].detach().cpu().numpy() |
|
|
if per_np.ndim > 1: |
|
|
per_np = per_np.squeeze() |
|
|
t_per = np.arange(len(per_np)) * hop_length / sr |
|
|
axs[cu_ax].plot(t_per, per_np, color="tab:red") |
|
|
axs[cu_ax].set_title("Period (Voice Activity)") |
|
|
axs[cu_ax].set_ylabel("periodocity") |
|
|
axs[cu_ax].set_xlim([0, max_t]) |
|
|
axs[cu_ax].grid(True, axis='both', linestyle='--', alpha=0.3) |
|
|
axs[cu_ax].set_ylim([-0.05, 1.05]) |
|
|
axs[cu_ax].axhline(y=0.5, color='k', linestyle='--', alpha=0.3) |
|
|
|
|
|
if markers is not None: |
|
|
for i, t in enumerate(markers): |
|
|
label = marker_labels[i] if marker_labels and i < len(marker_labels) else None |
|
|
for ax in axs: |
|
|
ax.axvline(x=t, color='k', linestyle='-', alpha=0.7, label=label if i == 0 else None) |
|
|
if marker_labels: |
|
|
axs[0].legend(loc='upper right', fontsize='small') |
|
|
axs[-1].set_xlabel("t (s)") |
|
|
fig.suptitle(title, fontsize=16) |
|
|
plt.tight_layout(rect=[0, 0, 1, 0.97]) |
|
|
plt.show() |
|
|
return fig |
|
|
|
|
|
def valid(default_value, *items): |
|
|
"""Get first non-None item""" |
|
|
for item in items: |
|
|
if item is not None: |
|
|
return item |
|
|
return default_value |
|
|
|
|
|
def dict_to(d, device, dtype=dtype): |
|
|
return {k: v.to(device, dtype) if isinstance(v, torch.Tensor) else v |
|
|
for k, v in d.items()} |
|
|
|
|
|
def exists(v): |
|
|
return v is not None |
|
|
|
|
|
def default(v, b): |
|
|
return v if exists(v) else b |
|
|
|
|
|
class Conv1d(nn.Conv1d): |
|
|
def _conv_forward( |
|
|
self, x: Tensor, weight: Tensor, bias) -> Tensor: |
|
|
return super()._conv_forward(x, weight.to(x.device, x.dtype), None if bias is None else bias.to(x.device, x.dtype)) |
|
|
|
|
|
class Conv2d(nn.Conv2d): |
|
|
def _conv_forward( |
|
|
self, x: Tensor, weight: Tensor, bias) -> Tensor: |
|
|
return super()._conv_forward(x, weight.to(x.device, x.dtype), None if bias is None else bias.to(x.device, x.dtype)) |
|
|
|
|
|
class Linear(nn.Module): |
|
|
def __init__(self, in_features: int, out_features: int, bias: bool = True) -> None: |
|
|
super(Linear, self).__init__() |
|
|
self.linear = nn.Linear(in_features, out_features, bias=bias) |
|
|
init.xavier_uniform_(self.linear.weight) |
|
|
if bias: |
|
|
init.zeros_(self.linear.bias) |
|
|
def forward(self, x: Tensor) -> Tensor: |
|
|
return self.linear(x) |
|
|
|
|
|
class RMSNorm(nn.Module): |
|
|
def __init__(self, dims: Union[int, Tensor, List, Tuple], |
|
|
eps = 1e-8, elementwise_affine = True): |
|
|
super(RMSNorm, self).__init__() |
|
|
if isinstance(dims, int): |
|
|
self.normalized_shape = (dims,) |
|
|
else: |
|
|
self.normalized_shape = tuple(dims) |
|
|
self.eps = eps |
|
|
self.elementwise_affine = elementwise_affine |
|
|
if self.elementwise_affine: |
|
|
self.weight = nn.Parameter(torch.empty(self.normalized_shape)) |
|
|
init.ones_(self.weight) |
|
|
else: |
|
|
self.register_parameter("weight", None) |
|
|
def forward(self, x): |
|
|
return F.rms_norm(x, self.normalized_shape, self.weight, self.eps) |
|
|
|
|
|
def LayerNorm(x: Tensor, normalized_shape: Union[int, Tensor, List, Tuple], |
|
|
weight: Optional[Tensor] = None, bias: Optional[Tensor] = None, |
|
|
eps: float = 1e-5) -> Tensor: |
|
|
return F.layer_norm(x, normalized_shape, weight, bias, eps) |
|
|
|
|
|
def get_device(): |
|
|
return torch.device("cuda:0" if torch.cuda.is_available() else "cpu") |
|
|
|
|
|
def get_dtype(): |
|
|
return torch.float32 if torch.cuda.is_available() else torch.float64 |
|
|
|
|
|
def tox(): |
|
|
return {"device": get_device(), "dtype": get_dtype()} |
|
|
|
|
|
class Sinusoids(nn.Module): |
|
|
def __init__(self, length, channels, max_tscale=10000): |
|
|
super().__init__() |
|
|
assert channels % 2 == 0 |
|
|
log_tscale_increment = np.log(max_tscale) / (channels // 2 - 1) |
|
|
inv_tscales = torch.exp(-log_tscale_increment * torch.arange(channels // 2)) |
|
|
scaled_t = torch.arange(length)[:, None] * inv_tscales[None, :] |
|
|
pos1 = torch.sin(scaled_t) |
|
|
pos2 = torch.cos(scaled_t) |
|
|
positions = torch.cat([pos1, pos2], dim=1) |
|
|
self.embedding = nn.Embedding.from_pretrained(positions, freeze=False) |
|
|
def forward(self, positions): |
|
|
return self.embedding(positions) |
|
|
|
|
|
def sinusoids(length, channels, max_tscale=10000): |
|
|
assert channels % 2 == 0 |
|
|
log_tscale_increment = np.log(max_tscale) / (channels // 2 - 1) |
|
|
inv_tscales = torch.exp(-log_tscale_increment * torch.arange(channels // 2)) |
|
|
scaled_t = torch.arange(length)[:, None] * inv_tscales[None, :] |
|
|
pos1 = torch.sin(scaled_t) |
|
|
pos2 = torch.cos(scaled_t) |
|
|
positions = torch.cat([pos1, pos2], dim=1) |
|
|
return nn.Parameter(positions.clone()) |
|
|
|
|
|
class rotary(nn.Module): |
|
|
def __init__(self, dims, head, max_ctx=1500, radii=True, debug: List[str] = [], use_pbias=False, use_2d_axial=False, spec_shape=None, use_true_2d_relative=False, freq_bins=None): |
|
|
super(rotary, self).__init__() |
|
|
self.use_pbias = use_pbias |
|
|
self.dims = dims |
|
|
self.head = head |
|
|
self.head_dim = dims // head |
|
|
self.radii = radii |
|
|
self.dim = self.head_dim |
|
|
self.debug = debug |
|
|
self.counter = 0 |
|
|
self.last_theta = None |
|
|
|
|
|
self.use_2d_axial = use_2d_axial |
|
|
if use_2d_axial and spec_shape is not None: |
|
|
time_frames, freq_bins = spec_shape |
|
|
self.time_frames = time_frames |
|
|
self.freq_bins = freq_bins |
|
|
time_theta = 50.0 |
|
|
time_freqs = 1.0 / (time_theta ** (torch.arange(0, self.head_dim, 2).float() / self.head_dim)) |
|
|
self.register_buffer('time_freqs', time_freqs) |
|
|
freq_theta = 100.0 |
|
|
freq_freqs = 1.0 / (freq_theta ** (torch.arange(0, self.head_dim, 2).float() / self.head_dim)) |
|
|
self.register_buffer('freq_freqs', freq_freqs) |
|
|
|
|
|
self.bias = nn.Parameter(torch.zeros(max_ctx, dims // 2), requires_grad=True if use_pbias else False) |
|
|
theta = (torch.tensor(10000, device=device, dtype=dtype)) |
|
|
self.theta = nn.Parameter(theta, requires_grad=True) |
|
|
self.theta_values = [] |
|
|
|
|
|
self.use_true_2d_relative = use_true_2d_relative |
|
|
self.freq_bins = freq_bins |
|
|
self.true2d_dim = (dims // head) // 2 |
|
|
self.omega_t = nn.Parameter(torch.randn(self.true2d_dim)) |
|
|
self.omega_f = nn.Parameter(torch.randn(self.true2d_dim)) |
|
|
|
|
|
def axial_freqs(self, seq_len): |
|
|
if not self.use_2d_axial: |
|
|
return None |
|
|
time_frames = self.time_frames |
|
|
freq_bins = self.freq_bins |
|
|
t = torch.arange(seq_len, device=device, dtype=dtype) |
|
|
t_x = (t % time_frames).float() |
|
|
t_y = torch.div(t, time_frames, rounding_mode='floor').float() |
|
|
freqs_x = torch.outer(t_x, self.time_freqs) |
|
|
freqs_y = torch.outer(t_y, self.freq_freqs) |
|
|
freqs_cis_x = torch.polar(torch.ones_like(freqs_x), freqs_x) |
|
|
freqs_cis_y = torch.polar(torch.ones_like(freqs_y), freqs_y) |
|
|
return torch.cat([freqs_cis_x, freqs_cis_y], dim=-1) |
|
|
|
|
|
def mel_scale_scalar(self, freq: float) -> float: |
|
|
return 1127.0 * math.log(1.0 + freq / 700.0) |
|
|
|
|
|
def mel_scale(self, freq: Tensor) -> Tensor: |
|
|
return 1127.0 * (1.0 + freq / 700.0).log() |
|
|
|
|
|
def pitch_bias(self, f0): |
|
|
if f0 is None: |
|
|
return None |
|
|
f0_flat = f0.squeeze().float() |
|
|
f0_norm = (f0_flat - f0_flat.mean()) / (f0_flat.std() + 1e-8) |
|
|
f0_sim = torch.exp(-torch.cdist(f0_norm.unsqueeze(1), |
|
|
f0_norm.unsqueeze(1))) |
|
|
return f0_sim.unsqueeze(0).unsqueeze(0) |
|
|
|
|
|
def theta_freqs(self, theta): |
|
|
if theta.dim() == 0: |
|
|
theta = theta.unsqueeze(0) |
|
|
freq = (theta.unsqueeze(-1) / 220.0) * 700 * ( |
|
|
torch.pow(10, torch.linspace(0, 2595 * torch.log10(torch.tensor(1 + 8000/700)), |
|
|
self.dim // 2, device=theta.device, dtype=theta.dtype) / 2595) - 1) / 1000 |
|
|
return freq |
|
|
|
|
|
def _apply_radii(self, freqs, f0, ctx): |
|
|
if self.radii and f0 is not None: |
|
|
radius = f0.to(device, dtype) |
|
|
L = radius.shape[0] |
|
|
if L != ctx: |
|
|
F = L / ctx |
|
|
idx = torch.arange(ctx, device=f0.device) |
|
|
idx = (idx * F).long().clamp(0, L - 1) |
|
|
radius = radius[idx] |
|
|
return torch.polar(radius.unsqueeze(-1), freqs), radius |
|
|
else: |
|
|
return torch.polar(radius.unsqueeze(-1), freqs), radius |
|
|
else: |
|
|
return torch.polar(torch.ones_like(freqs), freqs), None |
|
|
|
|
|
def check_f0(self, f0, f0t, ctx): |
|
|
if f0 is not None and f0.dim() == 2: |
|
|
f0 = f0.squeeze(0) |
|
|
if f0t is not None and f0t.dim() == 2: |
|
|
f0t = f0t.squeeze(0) |
|
|
|
|
|
if f0 is not None and f0.shape[0] == ctx: |
|
|
return f0 |
|
|
elif f0t is not None and f0t.shape[0] == ctx: |
|
|
return f0t |
|
|
else: |
|
|
return None |
|
|
|
|
|
def forward(self, x=None, enc=None, layer=None, feature=None) -> Tensor: |
|
|
ctx = x |
|
|
if self.use_2d_axial and feature == "spectrogram": |
|
|
freqs_2d = self.axial_freqs(ctx) |
|
|
if freqs_2d is not None: |
|
|
return freqs_2d.unsqueeze(0) |
|
|
|
|
|
f0 = enc.get("f0") if enc is not None else None |
|
|
f0t = enc.get("f0t") if enc is not None else None |
|
|
f0 = self.check_f0(f0, f0t, ctx) |
|
|
theta = f0 + self.theta if f0 is not None else self.theta |
|
|
freqs = self.theta_freqs(theta) |
|
|
t = torch.arange(ctx, device=device, dtype=dtype) |
|
|
freqs = t[:, None] * freqs |
|
|
freqs, radius = self._apply_radii(freqs, f0, ctx) |
|
|
|
|
|
if "radius" in self.debug and self.counter == 10: |
|
|
print(f" [{layer}] [Radius] {radius.shape if radius is not None else None} {radius.mean() if radius is not None else None} [Theta] {theta.mean() if theta is not None else None} [f0] {f0.shape if f0 is not None else None} [ctx] {ctx}") |
|
|
|
|
|
self.counter += 1 |
|
|
return freqs.unsqueeze(0) |
|
|
|
|
|
@staticmethod |
|
|
def apply_rotary(x, freqs): |
|
|
x1 = x[..., :freqs.shape[-1]*2] |
|
|
x2 = x[..., freqs.shape[-1]*2:] |
|
|
orig_shape = x1.shape |
|
|
if x1.ndim == 2: |
|
|
x1 = x1.unsqueeze(0) |
|
|
x1 = x1.float().reshape(*x1.shape[:-1], -1, 2).contiguous() |
|
|
x1 = torch.view_as_complex(x1) * freqs |
|
|
x1 = torch.view_as_real(x1).flatten(-2) |
|
|
x1 = x1.view(orig_shape) |
|
|
return torch.cat([x1.type_as(x), x2], dim=-1) |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
def parallel_slice(self, q, k, v, mask=None): |
|
|
batch, head, ctx, dims = q.shape |
|
|
head_dim = self.head_dim |
|
|
batch, ctx, dims = q.shape |
|
|
ctx_len = k.shape[1] |
|
|
head = dims // head_dim |
|
|
|
|
|
scores = torch.zeros(batch, head, ctx, ctx_len, device=q.device) |
|
|
|
|
|
for h in range(head): |
|
|
start_idx = h * head_dim |
|
|
end_idx = start_idx + head_dim |
|
|
q_h = q[:, :, start_idx:end_idx] |
|
|
k_h = k[:, :, start_idx:end_idx] |
|
|
|
|
|
scores[:, h] = torch.bmm(q_h, k_h.transpose(1, 2)) / math.sqrt(head_dim) |
|
|
|
|
|
if mask is not None: |
|
|
scores = scores + mask.unsqueeze(0).unsqueeze(0) |
|
|
|
|
|
attn_weights = F.softmax(scores, dim=-1) |
|
|
|
|
|
output = torch.zeros_like(q) |
|
|
for h in range(head): |
|
|
start_idx = h * head_dim |
|
|
end_idx = start_idx + head_dim |
|
|
v_h = v[:, :, start_idx:end_idx] |
|
|
output[:, :, start_idx:end_idx] = torch.bmm(attn_weights[:, h], v_h) |
|
|
return output |
|
|
|
|
|
class MultiheadA(nn.Module): |
|
|
def __init__(self, dims: int, head: int, rotary_emb: bool = True, |
|
|
zero_val: float = 1e-7, minz: float = 1e-8, maxz: float = 1e-6, debug: List[str] = [], optim_attn=False, use_pbias=False, use_true_2d_relative=False, freq_bins=None, radii=False, use_2d_axial=False, spec_shape=None, rbf=False): |
|
|
|
|
|
super(MultiheadA, self).__init__() |
|
|
self.dims = dims |
|
|
self.head = head |
|
|
self.head_dim = dims // head |
|
|
self.debug = debug |
|
|
self.counter = 0 |
|
|
self.use_pbias = use_pbias |
|
|
self.use_true_2d_relative = use_true_2d_relative |
|
|
self.freq_bins = freq_bins |
|
|
self.rbf = rbf |
|
|
|
|
|
self.q = nn.Linear(dims, dims).to(device, dtype) |
|
|
self.k = nn.Linear(dims, dims, bias=False).to(device, dtype) |
|
|
self.v = nn.Linear(dims, dims).to(device, dtype) |
|
|
self.o = nn.Linear(dims, dims).to(device, dtype) |
|
|
|
|
|
self.pad_token = 0 |
|
|
self.rotary_emb = rotary_emb |
|
|
self.minz = minz |
|
|
self.maxz = maxz |
|
|
self.zero_val = zero_val |
|
|
self.optim_attn = optim_attn |
|
|
self.fzero = nn.Parameter(torch.tensor(zero_val, device=device, dtype=dtype), requires_grad=False) |
|
|
|
|
|
if rotary_emb: |
|
|
self.rope = rotary( |
|
|
dims=dims, |
|
|
head=head, |
|
|
debug=debug, |
|
|
radii=radii, |
|
|
use_true_2d_relative=use_true_2d_relative, |
|
|
freq_bins=freq_bins, |
|
|
) |
|
|
else: |
|
|
self.rope = None |
|
|
|
|
|
def cos_sim(self, q: Tensor, k: Tensor, v: Tensor, mask) -> Tensor: |
|
|
q_norm = torch.nn.functional.normalize(q, dim=-1, eps=1e-12) |
|
|
k_norm = torch.nn.functional.normalize(k, dim=-1, eps=1e-12) |
|
|
qk_cosine = torch.matmul(q_norm, k_norm.transpose(-1, -2)) |
|
|
qk_cosine = qk_cosine + mask |
|
|
weights = F.softmax(qk_cosine, dim=-1) |
|
|
out = torch.matmul(weights, v) |
|
|
return out |
|
|
|
|
|
def rbf_scores(self, q, k, rbf_sigma=1.0, rbf_ratio=0.0): |
|
|
scale = (self.dims // self.head) ** -0.25 |
|
|
dot_scores = torch.matmul(q, k.transpose(-1, -2)) * scale |
|
|
if rbf_ratio <= 0.0: |
|
|
return dot_scores |
|
|
q_norm = q.pow(2).sum(dim=-1, keepdim=True) |
|
|
k_norm = k.pow(2).sum(dim=-1, keepdim=True) |
|
|
qk = torch.matmul(q, k.transpose(-1, -2)) |
|
|
dist_sq = q_norm + k_norm.transpose(-1, -2) - 2 * qk |
|
|
rbf_scores = torch.exp(-dist_sq / (2 * rbf_sigma**2)) |
|
|
return (1 - rbf_ratio) * dot_scores + rbf_ratio * rbf_scores |
|
|
|
|
|
def forward(self, x: Tensor, xa = None, mask = None, enc = None, layer = None, feature=None) -> tuple: |
|
|
|
|
|
x = x.to(device, dtype) |
|
|
if xa is not None: |
|
|
xa = xa.to(device, dtype) |
|
|
scale = (self.dims // self.head) ** -0.25 |
|
|
|
|
|
z = default(xa, x).to(device, dtype) |
|
|
q = self.q(x) |
|
|
k = self.k(z) |
|
|
v = self.v(z) |
|
|
|
|
|
if self.rotary_emb: |
|
|
q = q.view(*q.shape[:2], self.head, -1).permute(0, 2, 1, 3) |
|
|
k = k.view(*k.shape[:2], self.head, -1).permute(0, 2, 1, 3) |
|
|
v = v.view(*v.shape[:2], self.head, -1).permute(0, 2, 1, 3) |
|
|
q2 = q.shape[2] |
|
|
k2 = k.shape[2] |
|
|
|
|
|
if self.use_true_2d_relative and feature == "spectrogram": |
|
|
seq_len = q2 |
|
|
freq_bins = self.freq_bins |
|
|
idxs = torch.arange(seq_len, device=q.device) |
|
|
t_idx = idxs // freq_bins |
|
|
f_idx = idxs % freq_bins |
|
|
angle = self.rope.true2d_relative_angle(t_idx, f_idx, t_idx, f_idx) |
|
|
q_rot, k_rot = self.rope.true2d_apply_rotary(q, k, angle) |
|
|
scale = (self.dims // self.head) ** -0.25 |
|
|
qk = (q_rot * scale * k_rot * scale).sum(-1) |
|
|
w = F.softmax(qk, dim=-1).to(q.dtype) |
|
|
wv = torch.einsum('bhij,bhjd->bhid', w, v.unsqueeze(2).expand(-1, -1, seq_len, -1, -1)) |
|
|
wv = wv.permute(0, 2, 1, 3).flatten(start_dim=2) |
|
|
return self.o(wv), qk |
|
|
else: |
|
|
q = self.rope.apply_rotary(q, (self.rope(x=q2, enc=enc, layer=layer))) |
|
|
k = self.rope.apply_rotary(k, (self.rope(x=k2, enc=enc, layer=layer))) |
|
|
else: |
|
|
q = q.view(*q.shape[:2], self.head, -1).permute(0, 2, 1, 3) |
|
|
k = k.view(*k.shape[:2], self.head, -1).permute(0, 2, 1, 3) |
|
|
v = v.view(*v.shape[:2], self.head, -1).permute(0, 2, 1, 3) |
|
|
|
|
|
qk = (q * scale) @ (k * scale).transpose(-1, -2) |
|
|
|
|
|
if self.rbf: |
|
|
qk = self.rbf_scores(q * scale, k * scale, rbf_sigma=1.0, rbf_ratio=0.3) |
|
|
if self.use_pbias: |
|
|
pbias = self.rope.pitch_bias(f0 = enc.get("f0", None) if enc is not None else None) |
|
|
if pbias is not None: |
|
|
qk = qk + pbias[:,:,:q2,:q2] |
|
|
|
|
|
if mask is not None: |
|
|
mask = mask[:q2, :q2] |
|
|
|
|
|
token_ids = k[:, :, :, 0] |
|
|
zscale = torch.ones_like(token_ids) |
|
|
fzero = torch.clamp(F.softplus(self.fzero), self.minz, self.maxz) |
|
|
zscale[token_ids.float() == self.pad_token] = fzero |
|
|
|
|
|
if xa is not None: |
|
|
qk = qk + mask * zscale.unsqueeze(-2).expand(qk.shape) |
|
|
qk = qk * zscale.unsqueeze(-2) |
|
|
w = F.softmax(qk, dim=-1).to(q.dtype) |
|
|
wv = (w @ v).permute(0, 2, 1, 3).flatten(start_dim=2) |
|
|
|
|
|
if "multihead" in self.debug and self.counter % 100 == 0: |
|
|
print(f"MHA: q={q.shape}, k={k.shape}, v={v.shape} - {qk.shape}, wv shape: {wv.shape}") |
|
|
self.counter += 1 |
|
|
return self.o(wv), qk |
|
|
|
|
|
class t_gate(nn.Module): |
|
|
def __init__(self, dims, num_types=4, enabled=True): |
|
|
super().__init__() |
|
|
self.enabled = enabled |
|
|
self.gate_projections = nn.ModuleList([ |
|
|
nn.Sequential(Linear(dims, 1), nn.Sigmoid()) |
|
|
for _ in range(num_types)]) |
|
|
self.type_classifier = nn.Sequential( |
|
|
Linear(dims, num_types), |
|
|
nn.Softmax(dim=-1)) |
|
|
def forward(self, x): |
|
|
if not self.enabled: |
|
|
return None |
|
|
type_probs = self.type_classifier(x) |
|
|
gates = torch.stack([gate(x) for gate in self.gate_projections], dim=-1) |
|
|
comb_gate = torch.sum(gates * type_probs.unsqueeze(2), dim=-1) |
|
|
return comb_gate |
|
|
|
|
|
class m_gate(nn.Module): |
|
|
def __init__(self, dims, mem_size=64, enabled=True): |
|
|
super().__init__() |
|
|
self.enabled = enabled |
|
|
if enabled: |
|
|
self.m_key = nn.Parameter(torch.randn(mem_size, dims)) |
|
|
self.m_val = nn.Parameter(torch.randn(mem_size, 1)) |
|
|
self.gate_proj = nn.Sequential(Linear(dims, dims//2), nn.SiLU(), Linear(dims//2, 1)) |
|
|
|
|
|
def forward(self, x): |
|
|
if not self.enabled: |
|
|
return None |
|
|
d_gate = torch.sigmoid(self.gate_proj(x)) |
|
|
attention = torch.matmul(x, self.m_key.transpose(0, 1)) |
|
|
attention = F.softmax(attention / math.sqrt(x.shape[-1]), dim=-1) |
|
|
m_gate = torch.matmul(attention, self.m_val) |
|
|
m_gate = torch.sigmoid(m_gate) |
|
|
return 0.5 * (d_gate + m_gate) |
|
|
|
|
|
class c_gate(nn.Module): |
|
|
def __init__(self, dims, enabled=True): |
|
|
super().__init__() |
|
|
self.enabled = enabled |
|
|
if enabled: |
|
|
self.s_gate = nn.Sequential(Linear(dims, 1), nn.Sigmoid()) |
|
|
self.w_gate = nn.Sequential(Linear(dims, 1), nn.Sigmoid()) |
|
|
self.p_gate = nn.Sequential(Linear(dims, 1), nn.Sigmoid()) |
|
|
self.e_gate = nn.Sequential(Linear(dims, 1), nn.Sigmoid()) |
|
|
self.ph_gate = nn.Sequential(Linear(dims, 1), nn.Sigmoid()) |
|
|
self.integ = Linear(dims*5, dims) |
|
|
|
|
|
def forward(self, x, features): |
|
|
if not self.enabled: |
|
|
return None |
|
|
s_feat = features.get("spectrogram", x) |
|
|
w_feat = features.get("waveform", x) |
|
|
p_feat = features.get("pitch", x) |
|
|
e_feat = features.get("envelope", x) |
|
|
ph_feat = features.get("phase", x) |
|
|
s = self.s_gate(x) * s_feat |
|
|
w = self.w_gate(x) * w_feat |
|
|
p = self.p_gate(x) * p_feat |
|
|
e = self.e_gate(x) * e_feat |
|
|
ph = self.ph_gate(x) * ph_feat |
|
|
comb = torch.cat([s, w, p, e, ph], dim=-1) |
|
|
return self.integ(comb) |
|
|
|
|
|
class mlp_gate(nn.Module): |
|
|
def __init__(self, dims, enabled=True): |
|
|
super().__init__() |
|
|
self.enabled = enabled |
|
|
if enabled: |
|
|
self.gate = nn.Sequential(Linear(dims, 1), nn.Sigmoid()) |
|
|
|
|
|
def forward(self, x): |
|
|
if not self.enabled: |
|
|
return None |
|
|
return self.gate(x) |
|
|
|
|
|
class Residual(nn.Module): |
|
|
_seen = set() |
|
|
def __init__(self, ctx, dims, head, act, debug: List[str] = [], |
|
|
tgate=True, mgate=False, cgate=False, mem_size=512, features=None): |
|
|
super().__init__() |
|
|
|
|
|
self.dims = dims |
|
|
self.head = head |
|
|
self.ctx = ctx |
|
|
self.head_dim = dims // head |
|
|
self.features = features |
|
|
self.debug = debug |
|
|
self.counter = 0 |
|
|
self.dropout = 0.01 |
|
|
|
|
|
self.blend = nn.Parameter(torch.tensor(0.5)) |
|
|
act_fn = get_activation(act) |
|
|
self.attn = MultiheadA(dims, head, rotary_emb=True, debug=debug) |
|
|
|
|
|
if not any([tgate, mgate, cgate]): |
|
|
self.mlp_gate = nn.Sequential(Linear(dims, 1), nn.Sigmoid()) |
|
|
else: |
|
|
self.mlp_gate = None |
|
|
|
|
|
mlp = dims * 4 |
|
|
self.mlp = nn.Sequential(Linear(dims, mlp), act_fn, Linear(mlp, dims)) |
|
|
|
|
|
self.t_gate = t_gate(dims=dims, num_types=4*2, enabled=tgate) |
|
|
self.m_gate = m_gate(dims=dims, mem_size=mem_size, enabled=mgate) |
|
|
self.c_gate = c_gate(dims=dims, enabled=cgate) |
|
|
self.mlp_gate = mlp_gate(dims=dims, enabled=not any([tgate, mgate, cgate])) |
|
|
|
|
|
self.lna = RMSNorm(dims) |
|
|
self.lnb = RMSNorm(dims) |
|
|
self.lnc = RMSNorm(dims) |
|
|
|
|
|
def forward(self, x, xa=None, mask=None, enc=None, layer=None, feature=None) -> Tensor: |
|
|
|
|
|
b = torch.sigmoid(self.blend) |
|
|
ax = x + self.attn(self.lna(x), xa=xa, mask=mask, enc=enc, layer=layer, feature=feature)[0] |
|
|
bx = b * ax + (1 - b) * x |
|
|
cx = self.lnb(bx) |
|
|
dx = self.mlp(cx) |
|
|
ex = self.t_gate(cx) if not None else self.default(self.m_gate(cx), self.mlp_gate(cx)) |
|
|
fx = x + ex + dx |
|
|
gx = self.lnc(fx) |
|
|
return gx |
|
|
|
|
|
class FEncoder(nn.Module): |
|
|
def __init__(self, input_dims, dims, head, layer, kernel_size, act, stride=1, use_rope=False, spec_shape=None): |
|
|
super().__init__() |
|
|
|
|
|
self.head = head |
|
|
self.head_dim = dims // head |
|
|
self.dropout = 0.01 |
|
|
self.use_rope = use_rope |
|
|
self.dims = dims |
|
|
|
|
|
act_fn = get_activation(act) |
|
|
|
|
|
self.encoder = nn.Sequential( |
|
|
Conv1d(input_dims, dims, kernel_size=kernel_size, stride=stride, padding=kernel_size//2), act_fn, |
|
|
Conv1d(dims, dims, kernel_size=5, padding=2), act_fn, |
|
|
Conv1d(dims, dims, kernel_size=3, padding=1, groups=dims), act_fn) |
|
|
|
|
|
if use_rope: |
|
|
if spec_shape is not None: |
|
|
self.rope = rotary( |
|
|
dims=dims, |
|
|
head=head, |
|
|
use_2d_axial=True, |
|
|
spec_shape=spec_shape, debug=[]) |
|
|
else: |
|
|
self.rope = rotary( |
|
|
dims=dims, |
|
|
head=head, |
|
|
use_2d_axial=False, debug=[]) |
|
|
else: |
|
|
self.rope = None |
|
|
self.sinusoid_pos = lambda length, dims: sinusoids(length, dims, max_tscale=10000) |
|
|
|
|
|
self.norm = RMSNorm(dims) |
|
|
|
|
|
def apply_rope_to_features(self, x, layer=None, feature=None): |
|
|
batch, ctx, dims = x.shape |
|
|
x = x.view(batch, ctx, self.head, self.head_dim).permute(0, 2, 1, 3) |
|
|
if feature == "spectrogram" and self.rope is not None: |
|
|
rope_freqs = self.rope(ctx, layer=layer, feature="spectrogram") |
|
|
else: |
|
|
rope_freqs = self.rope(ctx, layer=layer, feature="audio") |
|
|
x = self.rope.apply_rotary(x, rope_freqs) |
|
|
x = x.permute(0, 2, 1, 3).contiguous().view(batch, ctx, dims) |
|
|
return x |
|
|
|
|
|
def forward(self, x, enc=None, layer=None, feature=None): |
|
|
x = self.encoder(x).permute(0, 2, 1) |
|
|
if self.use_rope: |
|
|
x = self.apply_rope_to_features(x, layer=layer, feature=feature) |
|
|
else: |
|
|
x = x + self.sinusoid_pos(x.shape[1], x.shape[-1]).to(x.device, x.dtype) |
|
|
x = nn.functional.dropout(x, p=self.dropout, training=self.training) |
|
|
return self.norm(x) |
|
|
|
|
|
class WEncoder(nn.Module): |
|
|
def __init__(self, input_dims, dims, head, layer, kernel_size, act, use_rope=False): |
|
|
super().__init__() |
|
|
|
|
|
self.head = head |
|
|
self.head_dim = dims // head |
|
|
self.dropout = 0.01 |
|
|
self.use_rope = use_rope |
|
|
self.dims = dims |
|
|
|
|
|
act_fn = get_activation(act) |
|
|
self.downsample = nn.Sequential( |
|
|
Conv1d(input_dims, dims//8, kernel_size=15, stride=8, padding=7), act_fn, |
|
|
Conv1d(dims//8, dims//4, kernel_size=7, stride=4, padding=3), act_fn, |
|
|
Conv1d(dims//4, dims, kernel_size=9, stride=5, padding=4), act_fn) |
|
|
|
|
|
self.encoder = nn.Sequential( |
|
|
Conv1d(dims, dims, kernel_size=3, padding=1, groups=dims//8), act_fn, |
|
|
Conv1d(dims, dims, kernel_size=1), act_fn) |
|
|
if use_rope: |
|
|
self.rope = rotary( |
|
|
dims=self.head_dim, |
|
|
head=self.head, |
|
|
debug=[]) |
|
|
else: |
|
|
self.rope = None |
|
|
self.sinusoid_pos = lambda length, dims: sinusoids(length, dims, max_tscale=10000) |
|
|
self.norm = RMSNorm(dims) |
|
|
|
|
|
def apply_rope_to_features(self, x, layer=None, feature=None): |
|
|
if not self.use_rope or self.rope is None: |
|
|
return x |
|
|
batch, ctx, dims = x.shape |
|
|
x = x.view(batch, ctx, self.head, self.head_dim).permute(0, 2, 1, 3) |
|
|
rope_freqs = self.rope(ctx, layer=layer, feature=feature) |
|
|
x = self.rope.apply_rotary(x, rope_freqs) |
|
|
x = x.permute(0, 2, 1, 3).contiguous().view(batch, ctx, dims) |
|
|
return x |
|
|
|
|
|
def forward(self, x, enc=None, layer=None, feature=None): |
|
|
x = self.downsample(x) |
|
|
x = self.encoder(x) |
|
|
x = x.permute(0, 2, 1) |
|
|
if self.use_rope: |
|
|
x = self.apply_rope_to_features(x, layer=layer) |
|
|
else: |
|
|
x = x + self.sinusoid_pos(x.shape[1], x.shape[-1]).to(x.device, x.dtype) |
|
|
x = nn.functional.dropout(x, p=self.dropout, training=self.training) |
|
|
return self.norm(x) |
|
|
|
|
|
class PEncoder(nn.Module): |
|
|
def __init__(self, input_dims, dims, head, layer, kernel_size, act, use_rope=False): |
|
|
super().__init__() |
|
|
|
|
|
self.head = head |
|
|
self.head_dim = dims // head |
|
|
self.dropout = 0.01 |
|
|
self.use_rope = use_rope |
|
|
self.dims = dims |
|
|
|
|
|
act_fn = get_activation(act) |
|
|
|
|
|
self.encoder = nn.Sequential( |
|
|
Conv1d(input_dims, dims//4, kernel_size=7, stride=8, padding=3), act_fn, |
|
|
Conv1d(dims//4, dims//2, kernel_size=5, stride=4, padding=2), act_fn, |
|
|
Conv1d(dims//2, dims, kernel_size=5, stride=5, padding=2), act_fn) |
|
|
|
|
|
if use_rope: |
|
|
self.rope = rotary( |
|
|
dims=self.head_dim, |
|
|
head=self.head, |
|
|
debug=[]) |
|
|
else: |
|
|
self.rope = None |
|
|
self.sinusoid_pos = lambda length, dims: sinusoids(length, dims, max_tscale=10000) |
|
|
self.norm = RMSNorm(dims) |
|
|
|
|
|
def apply_rope_to_features(self, x, layer=None, feature=None): |
|
|
if not self.use_rope or self.rope is None: |
|
|
return x |
|
|
batch, ctx, dims = x.shape |
|
|
x = x.view(batch, ctx, self.head, self.head_dim).permute(0, 2, 1, 3) |
|
|
rope_freqs = self.rope(ctx, layer=layer, feature=feature) |
|
|
x = self.rope.apply_rotary(x, rope_freqs) |
|
|
x = x.permute(0, 2, 1, 3).contiguous().view(batch, ctx, dims) |
|
|
return x |
|
|
|
|
|
def forward(self, x, enc=None, layer=None, feature=None): |
|
|
x = self.encoder(x).permute(0, 2, 1) |
|
|
if self.use_rope: |
|
|
x = self.apply_rope_to_features(x, layer=layer) |
|
|
else: |
|
|
x = x + self.sinusoid_pos(x.shape[1], x.shape[-1]).to(x.device, x.dtype) |
|
|
x = nn.functional.dropout(x, p=self.dropout, training=self.training) |
|
|
return self.norm(x) |
|
|
|
|
|
class theBridge(nn.Module): |
|
|
def __init__(self, vocab: int, mels: int, ctx: int, dims: int, head: int, layer: int, |
|
|
debug: List[str], features: List[str], act: str = "gelu"): |
|
|
super(theBridge, self).__init__() |
|
|
|
|
|
self.debug = debug |
|
|
self.counter = 0 |
|
|
self.dropout = 0.01 |
|
|
self.features = features |
|
|
self.do_blend = "no_blend" not in self.debug |
|
|
self.sequential = "sequential" in self.debug |
|
|
|
|
|
self.token = nn.Embedding(vocab, dims, device=device, dtype=dtype) |
|
|
self.positional = nn.Parameter(torch.empty(ctx, dims, device=device, dtype=dtype), requires_grad=True) |
|
|
self.blend = nn.Parameter(torch.tensor(0.5, device=device, dtype=dtype), requires_grad=True) |
|
|
self.sinusoid_pos = lambda length, dims: sinusoids(length, dims, max_tscale=10000) |
|
|
|
|
|
with torch.no_grad(): |
|
|
self.token.weight[0].zero_() |
|
|
|
|
|
self.block = nn.ModuleList([ |
|
|
Residual(ctx=ctx, dims=dims, head=head, act=act, debug=debug, features=features) |
|
|
for _ in range(layer)]) |
|
|
|
|
|
self.cross_attn = nn.ModuleList([ |
|
|
Residual(ctx=ctx, dims=dims, head=head, act=act, debug=debug, features=features) |
|
|
for _ in range(layer)]) |
|
|
|
|
|
self.cross_modal = nn.ModuleList([ |
|
|
Residual(ctx=ctx, dims=dims, head=head, act=act, debug=debug, features=features) |
|
|
for _ in range(layer)]) |
|
|
|
|
|
mask = torch.tril(torch.ones(ctx, ctx), diagonal=0).unsqueeze(0).unsqueeze(0) |
|
|
self.register_buffer("mask", mask, persistent=False) |
|
|
|
|
|
act_fn = get_activation(act) |
|
|
if features == ["spectrogram", "waveform", "pitch"]: |
|
|
cgate=True |
|
|
else: |
|
|
cgate = False |
|
|
|
|
|
self.blocks = nn.ModuleDict({ |
|
|
"spectrogram": nn.ModuleList( |
|
|
[FEncoder(input_dims=mels, dims=dims, head=head, layer=layer, kernel_size=3, act=act_fn)] + |
|
|
[Residual(ctx=ctx, dims=dims, head=head, act=act, debug=debug, features=features, cgate=cgate) for _ in range(layer)] if "spectrogram" in features else None), |
|
|
"waveform": nn.ModuleList( |
|
|
[WEncoder(input_dims=1, dims=dims, head=head, layer=layer, kernel_size=11, act=act_fn)] + |
|
|
[Residual(ctx=ctx, dims=dims, head=head, act=act, debug=debug, features=features, cgate=cgate) for _ in range(layer)] if "waveform" in features else None), |
|
|
"pitch": nn.ModuleList( |
|
|
[FEncoder(input_dims=1, dims=dims, head=head, layer=layer, kernel_size=9, act=act, stride=2)] + |
|
|
[Residual(ctx=ctx, dims=dims, head=head, act=act, debug=debug, features=features, cgate=cgate) for _ in range(layer)] if "pitch" in features else None), |
|
|
"envelope": nn.ModuleList( |
|
|
[FEncoder(input_dims=mels, dims=dims, head=head, layer=layer, kernel_size=3, act=act_fn)] + |
|
|
[Residual(ctx=ctx, dims=dims, head=head, act=act, debug=debug, features=features, cgate=cgate) for _ in range(layer)] if "envelope" in features else None), |
|
|
"phase": nn.ModuleList( |
|
|
[FEncoder(input_dims=mels, dims=dims, head=head, layer=layer, kernel_size=3, act=act_fn)] + |
|
|
[Residual(ctx=ctx, dims=dims, head=head, act=act, debug=debug, features=features, cgate=cgate) for _ in range(layer)] if "phase" in features else None)}) |
|
|
|
|
|
self.norm = RMSNorm(dims) |
|
|
|
|
|
def forward(self, x, enc, layer='decoder', feature=None) -> Tensor: |
|
|
out = {} |
|
|
out.update(enc) |
|
|
enc = dict_to(enc, device, dtype) |
|
|
_text_len = x.shape[1] |
|
|
|
|
|
x = self.token(x) + self.positional[:x.shape[1]] |
|
|
|
|
|
for f in enc: |
|
|
if f in self.features: |
|
|
xa = enc[f] |
|
|
for block in self.blocks[f]: |
|
|
xa = block(xa, enc=enc, layer=layer, feature=feature) |
|
|
xa = xa + self.sinusoid_pos(xa.shape[1], xa.shape[-1]).to(xa.device, xa.dtype) |
|
|
out[f] = xa |
|
|
|
|
|
for block in self.block: |
|
|
x = block(x, xa=None, mask=self.mask, enc=enc, layer=layer) |
|
|
if f in self.features: |
|
|
|
|
|
out = block(x, xa=xa, mask=self.mask, enc=enc, layer=layer) |
|
|
if self.sequential: |
|
|
x = out |
|
|
else: |
|
|
a = torch.sigmoid(self.blend) |
|
|
x = a * out + (1 - a) * x |
|
|
x = self.token(x) + self.positional[:x.shape[1]] |
|
|
out[f] = x |
|
|
|
|
|
for block in self.cross_attn: |
|
|
if f in self.features: |
|
|
x = block(x, xa=xa, mask=self.mask, enc=enc, layer=layer) |
|
|
xa = block(xa, xa=x, mask=self.mask, enc=enc, layer=layer) |
|
|
out = block(x, xa=xa, mask=self.mask, enc=enc, layer=layer) |
|
|
if self.sequential: |
|
|
x = out |
|
|
else: |
|
|
a = torch.sigmoid(self.blend) |
|
|
x = a * out + (1 - a) * x |
|
|
x = self.token(x) + self.positional[:x.shape[1]] |
|
|
out[f] = x |
|
|
|
|
|
for block in self.cross_modal: |
|
|
if f in self.features: |
|
|
xcat = torch.cat([x, xa], dim=1) |
|
|
x = block(xcat, xa=None, mask=self.mask, enc=enc, layer=layer) |
|
|
x = x[:, :_text_len] |
|
|
out[f] = x |
|
|
if self.counter < 1 and "encoder" in self.debug: |
|
|
shapes = {k: v.shape for k, v in enc.items()} |
|
|
print(f"Step {self.counter}: mode: {list(enc.keys()) }: shapes: {shapes}") |
|
|
self.counter += 1 |
|
|
|
|
|
x = x @ torch.transpose(self.token.weight.to(dtype), 0, 1).float() |
|
|
x = self.norm(x) |
|
|
return x, out |
|
|
|
|
|
class Echo(nn.Module): |
|
|
def __init__(self, param: Dimensions): |
|
|
super().__init__() |
|
|
self.param = param |
|
|
|
|
|
self.processor = theBridge( |
|
|
vocab=param.vocab, |
|
|
mels=param.mels, |
|
|
ctx=param.ctx, |
|
|
dims=param.dims, |
|
|
head=param.head, |
|
|
layer=param.layer, |
|
|
features=param.features, |
|
|
act=param.act, |
|
|
debug=param.debug, |
|
|
) |
|
|
|
|
|
def forward(self, |
|
|
labels=None, |
|
|
input_ids=None, |
|
|
waveform: Optional[torch.Tensor]=None, |
|
|
spectrogram: Optional[torch.Tensor]=None, |
|
|
pitch: Optional[torch.Tensor]=None, |
|
|
f0: Optional[torch.Tensor]=None, |
|
|
f0t: Optional[torch.Tensor]=None, |
|
|
harmonic: Optional[torch.Tensor]=None, |
|
|
aperiodic: Optional[torch.Tensor]=None, |
|
|
) -> Dict[str, Optional[torch.Tensor]]: |
|
|
|
|
|
enc = {} |
|
|
if spectrogram is not None: |
|
|
enc["spectrogram"] = spectrogram |
|
|
feature = "spectrogram" |
|
|
if waveform is not None: |
|
|
enc["waveform"] = waveform |
|
|
feature = "waveform" |
|
|
if pitch is not None: |
|
|
enc["pitch"] = pitch |
|
|
feature = "pitch" |
|
|
if f0 is not None: |
|
|
enc["f0"] = f0 |
|
|
if f0t is not None: |
|
|
enc["f0t"] = f0t |
|
|
if harmonic is not None: |
|
|
enc["harmonic"] = harmonic |
|
|
if aperiodic is not None: |
|
|
enc["aperiodic"] = aperiodic |
|
|
if input_ids is not None: |
|
|
enc["input_ids"] = input_ids |
|
|
feature = "input_ids" |
|
|
else: |
|
|
feature = "spectrogram" |
|
|
|
|
|
out, logits = self.processor(input_ids, enc, feature) |
|
|
self.out=out |
|
|
|
|
|
loss = None |
|
|
if labels is not None: |
|
|
loss = F.cross_entropy( |
|
|
logits.view(-1, logits.shape[-1]), labels.view(-1), ignore_index=0) |
|
|
|
|
|
return {"logits": logits, "loss": loss} |
|
|
|
|
|
@property |
|
|
def device(self): |
|
|
return next(self.parameters()).device |
|
|
@property |
|
|
def dtype(self): |
|
|
return next(self.parameters()).dtype |
|
|
|
|
|
def _init_weights(self, module): |
|
|
std = 0.02 |
|
|
self.init_counts = { |
|
|
"Linear": 0, "Conv1d": 0, "LayerNorm": 0, "RMSNorm": 0, |
|
|
"Conv2d": 0, "SEBlock": 0, "SpeechTransformer": 0, |
|
|
"Residual": 0, "MultiheadA": 0, |
|
|
"MultiheadC": 0, "MultiheadD": 0, "FEncoder": 0, |
|
|
"WEncoder": 0, "PEncoder": 0} |
|
|
|
|
|
for name, module in self.named_modules(): |
|
|
if isinstance(module, RMSNorm): |
|
|
nn.init.ones_(module.weight) |
|
|
self.init_counts["RMSNorm"] += 1 |
|
|
elif isinstance(module, nn.Linear): |
|
|
if module.weight is not None: |
|
|
nn.init.xavier_uniform_(module.weight) |
|
|
if module.bias is not None: |
|
|
nn.init.zeros_(module.bias) |
|
|
self.init_counts["Linear"] += 1 |
|
|
elif isinstance(module, Conv1d): |
|
|
nn.init.normal_(module.weight, mean=0.0, std=std) |
|
|
if module.bias is not None: |
|
|
nn.init.zeros_(module.bias) |
|
|
self.init_counts["Conv1d"] += 1 |
|
|
elif isinstance(module, Conv2d): |
|
|
nn.init.normal_(module.weight, mean=0.0, std=std) |
|
|
if module.bias is not None: |
|
|
nn.init.zeros_(module.bias) |
|
|
self.init_counts["Conv2d"] += 1 |
|
|
elif isinstance(module, MultiheadA): |
|
|
self.init_counts["MultiheadA"] += 1 |
|
|
elif isinstance(module, Residual): |
|
|
self.init_counts["Residual"] += 1 |
|
|
|
|
|
def init_weights(self): |
|
|
print("Initializing model weights...") |
|
|
self.apply(self._init_weights) |
|
|
print("Initialization summary:") |
|
|
for module_type, count in self.init_counts.items(): |
|
|
if count > 0: |
|
|
print(f"{module_type}: {count}") |
|
|
|
|
|
def generate(self, input_ids=None, spectrogram=None, waveform=None, pitch=None, f0=None, |
|
|
envelope=None, phase=None, tokenizer=None, max_length=128, min_length=1, device=None, **kwargs): |
|
|
if device is None: |
|
|
device = self.device |
|
|
pad_token_id = getattr(tokenizer, "pad_token_id", 0) |
|
|
bos_token_id = getattr(tokenizer, "bos_token_id", 1) |
|
|
eos_token_id = getattr(tokenizer, "eos_token_id", 2) |
|
|
batch_size = 1 |
|
|
for x in [spectrogram, waveform, pitch, f0, envelope, phase]: |
|
|
if x is not None: |
|
|
batch_size = x.shape[0] |
|
|
break |
|
|
ids = torch.full((batch_size, 1), bos_token_id, dtype=torch.long, device=device) |
|
|
feature = {} |
|
|
if spectrogram is not None: |
|
|
feature["spectrogram"] = spectrogram |
|
|
if waveform is not None: |
|
|
feature["waveform"] = waveform |
|
|
if pitch is not None: |
|
|
feature["pitch"] = pitch |
|
|
if envelope is not None: |
|
|
feature["envelope"] = envelope |
|
|
if phase is not None: |
|
|
feature["phase"] = phase |
|
|
if f0 is not None: |
|
|
feature["f0"] = f0 |
|
|
|
|
|
for i in range(max_length - 1): |
|
|
with torch.no_grad(): |
|
|
feature["input_ids"] = ids |
|
|
logits = self.SpeechTransformer(feature) |
|
|
next_token_logits = logits[:, -1, :] |
|
|
if i < min_length: |
|
|
next_token_logits[:, eos_token_id] = 0 |
|
|
next_tokens = torch.argmax(next_token_logits, dim=-1, keepdim=True) |
|
|
ids = torch.cat([ids, next_tokens], dim=1) |
|
|
if (next_tokens == eos_token_id).all() and i >= min_length: |
|
|
break |
|
|
return ids |
|
|
|
|
|
@property |
|
|
def config(self): |
|
|
class Config: |
|
|
pad_token_id = getattr(self.param, "pad_token_id", 0) |
|
|
bos_token_id = getattr(self.param, "bos_token_id", 1) |
|
|
eos_token_id = getattr(self.param, "eos_token_id", 2) |
|
|
def to_json_string(self): |
|
|
import json |
|
|
return json.dumps({ |
|
|
"pad_token_id": self.pad_token_id, |
|
|
"bos_token_id": self.bos_token_id, |
|
|
"eos_token_id": self.eos_token_id, |
|
|
}) |
|
|
return Config() |
|
|
|
|
|
def setup_tokenizer(token: str): |
|
|
from tokenizers import Tokenizer |
|
|
tokenizer = Tokenizer.from_file("./tokenizer.json") |
|
|
orig_encode = tokenizer.encode |
|
|
def enc(text, add_special_tokens=True): |
|
|
ids = orig_encode(text).ids |
|
|
if not add_special_tokens: |
|
|
sp_ids = [tokenizer.token_to_id(t) for t in ["<PAD>", "<BOS>", "<EOS>"]] |
|
|
ids = [id for id in ids if id not in sp_ids] |
|
|
return ids |
|
|
|
|
|
def bdec(ids_list, skip_special_tokens=True, pad_token_id=0, bos_token_id=1, eos_token_id=2): |
|
|
results = [] |
|
|
for ids in ids_list: |
|
|
if isinstance(ids, torch.Tensor): |
|
|
ids = ids.tolist() |
|
|
ids = [int(id) for id in ids if id != -100] |
|
|
if skip_special_tokens: |
|
|
ids = [id for id in ids if id not in (pad_token_id, bos_token_id, eos_token_id)] |
|
|
|
|
|
if ids and ids and ids[0] == bos_token_id: |
|
|
ids = ids[1:] |
|
|
while ids and ids[-1] == eos_token_id: |
|
|
ids = ids[:-1] |
|
|
results.append(tokenizer.decode(ids)) |
|
|
return results |
|
|
|
|
|
def save_pretrained(save_dir): |
|
|
os.makedirs(save_dir, exist_ok=True) |
|
|
tokenizer.save(f"{save_dir}/tokenizer.json") |
|
|
tokenizer.encode = enc |
|
|
tokenizer.batch_decode = bdec |
|
|
tokenizer.save_pretrained = save_pretrained |
|
|
tokenizer.pad_token_id = 0 |
|
|
tokenizer.bos_token_id = 1 |
|
|
tokenizer.eos_token_id = 2 |
|
|
return tokenizer |
|
|
|
|
|
def tokenize_pitch(pitch_features, target_length): |
|
|
pitch_len = pitch_features.shape[-1] |
|
|
token_len = target_length |
|
|
if pitch_len > token_len: |
|
|
pitch_tokens = F.adaptive_avg_pool1d(pitch_features, token_len) |
|
|
else: |
|
|
pitch_tokens = F.interpolate(pitch_features, token_len) |
|
|
return pitch_tokens |
|
|
|
|
|
def load_wave(wave_data, sample_rate): |
|
|
if isinstance(wave_data, str): |
|
|
waveform, sr = torchaudio.load(uri=wave_data, normalize=False) |
|
|
elif isinstance(wave_data, dict): |
|
|
waveform = torch.tensor(data=wave_data["array"]).float() |
|
|
sr = wave_data["sampling_rate"] |
|
|
else: |
|
|
raise TypeError("Invalid wave_data format.") |
|
|
|
|
|
return waveform |
|
|
|
|
|
def world_to_mel(sp, ap, sample_rate=16000, n_mels=128): |
|
|
import librosa |
|
|
mel_basis = librosa.filters.mel(sr=sample_rate, n_fft=1024, n_mels=n_mels) |
|
|
mel_basis = torch.from_numpy(mel_basis).float() |
|
|
|
|
|
sp_mel = torch.matmul(sp, mel_basis.T) |
|
|
ap_mel = torch.matmul(ap, mel_basis.T) |
|
|
|
|
|
return sp_mel, ap_mel |
|
|
|
|
|
def extract_features(batch, tokenizer, waveform=False, spec=True, f0=True, f0t=True, pitch=False, harmonics=False, sample_rate=16000, hop_length=256, mode="mean", debug=False, **dataset_config): |
|
|
dataset_config = { |
|
|
"hop_length": 256, |
|
|
"f_min": 150, |
|
|
"f_max": 2000, |
|
|
"n_mels": 128, |
|
|
"n_fft": 1024, |
|
|
"sample_rate": 16000, |
|
|
"pad_mode": "constant", |
|
|
"center": True, |
|
|
"power": 1.0, |
|
|
"window_fn": torch.hann_window, |
|
|
"mel_scale": "htk", |
|
|
"norm": None, |
|
|
"normalized": False, |
|
|
} |
|
|
|
|
|
audio = batch["audio"] |
|
|
sr = audio["sampling_rate"] |
|
|
labels = tokenizer.encode(batch["transcription"]) |
|
|
|
|
|
wav = wavnp = f0_np = t = None |
|
|
spectrogram = f0_tensor = f0t_tensor = harmonic = aperiodic = None |
|
|
|
|
|
if waveform or spec or f0 or f0t or harmonics: |
|
|
wav = load_wave(wave_data=audio, sample_rate=sr) |
|
|
wavnp = wav.numpy().astype(np.float64) |
|
|
|
|
|
if spec: |
|
|
transform = torchaudio.transforms.MelSpectrogram(**dataset_config) |
|
|
mel_spectrogram = transform(wav) |
|
|
log_mel = torch.clamp(mel_spectrogram, min=1e-10).log10() |
|
|
log_mel = torch.maximum(log_mel, log_mel.max() - 8.0) |
|
|
spectrogram = (log_mel + 4.0) / 4.0 |
|
|
spectrogram = torch.tensor(spectrogram) |
|
|
|
|
|
if f0 or f0t or harmonics: |
|
|
f0_np, t = pw.dio(wavnp, sample_rate, |
|
|
frame_period=hop_length / sample_rate * 1000, f0_ceil=500, f0_floor=71.1) |
|
|
f0_np = pw.stonemask(wavnp, f0_np, t, sample_rate) |
|
|
|
|
|
if f0: |
|
|
f0_tensor = torch.from_numpy(f0_np) |
|
|
f0_tensor = torch.where(f0_tensor == 0.0, torch.zeros_like(f0_tensor), (f0_tensor - 71.0) / (500.0 - 71.0)) |
|
|
|
|
|
if f0t: |
|
|
audio_duration = len(wavnp) / sample_rate |
|
|
T = len(labels) |
|
|
tok_dur_sec = audio_duration / T |
|
|
token_starts = np.arange(T) * tok_dur_sec |
|
|
token_ends = token_starts + tok_dur_sec |
|
|
start_idx = np.searchsorted(t, token_starts, side="left") |
|
|
end_idx = np.searchsorted(t, token_ends, side="right") |
|
|
pitch_tok = np.zeros(T, dtype=np.float32) |
|
|
for i in range(T): |
|
|
lo, hi = start_idx[i], max(start_idx[i]+1, end_idx[i]) |
|
|
segment = f0_np[lo:hi] |
|
|
if mode == "mean": |
|
|
pitch_tok[i] = segment.mean() |
|
|
elif mode == "median": |
|
|
pitch_tok[i] = np.median(segment) |
|
|
else: |
|
|
pitch_tok[i] = segment[-1] |
|
|
pitch_tok[pitch_tok < 100.0] = 0.0 |
|
|
bos_pitch = pitch_tok[0] if len(pitch_tok) > 0 else 0.0 |
|
|
f0t_tensor = torch.from_numpy(np.concatenate([[bos_pitch], pitch_tok])) |
|
|
f0t_tensor = torch.where(f0t_tensor == 0.0, torch.zeros_like(f0t_tensor), (f0t_tensor - 71.0) / (500.0 - 71.0)) |
|
|
|
|
|
if harmonics: |
|
|
spnp = pw.cheaptrick(wavnp, f0_np, t, sample_rate, fft_size=256) |
|
|
apnp = pw.d4c(wavnp, f0_np, t, sample_rate, fft_size=256) |
|
|
harmonic = torch.from_numpy(spnp) |
|
|
aperiodic = torch.from_numpy(apnp) |
|
|
harmonic = harmonic[:, :128].contiguous().T |
|
|
aperiodic = aperiodic[:, :128].contiguous().T |
|
|
harmonic = torch.where(harmonic == 0.0, torch.zeros_like(harmonic), harmonic / 1.0) |
|
|
aperiodic = torch.where(aperiodic == 0.0, torch.zeros_like(aperiodic), aperiodic / 1.0) |
|
|
|
|
|
if debug: |
|
|
print(f"['f0']: {f0_tensor.shape if f0_tensor is not None else None}") |
|
|
print(f"['f0t']: {f0t_tensor.shape if f0t_tensor is not None else None}") |
|
|
print(f"['harmonic']: {harmonic.shape if harmonic is not None else None}") |
|
|
print(f"['aperiodic']: {aperiodic.shape if aperiodic is not None else None}") |
|
|
print(f"['spectrogram']: {spectrogram.shape if spectrogram is not None else None}") |
|
|
print(f"['waveform']: {wav.shape if wav is not None else None}") |
|
|
print(f"['labels']: {len(labels) if labels is not None else None}") |
|
|
|
|
|
return { |
|
|
"waveform": wav if waveform else None, |
|
|
"spectrogram": spectrogram if spec else None, |
|
|
"f0": f0_tensor if f0 else None, |
|
|
"f0t": f0t_tensor if f0t else None, |
|
|
"harmonic": harmonic if harmonics else None, |
|
|
"aperiodic": aperiodic if harmonics else None, |
|
|
"labels": labels, |
|
|
} |
|
|
|
|
|
def prepare_datasets(tokenizer, token, sanity_check=False, sample_rate=16000, streaming=False, **dataset_config): |
|
|
|
|
|
if sanity_check: |
|
|
test = load_dataset( |
|
|
"google/fleurs", "en_us", token=token, split="test", trust_remote_code=True |
|
|
).cast_column("audio", Audio(sampling_rate=sample_rate)).take(1) |
|
|
dataset = test.map( |
|
|
lambda x: extract_features(x, tokenizer, **dataset_config), |
|
|
remove_columns=test.column_names) |
|
|
train_dataset = dataset |
|
|
test_dataset = dataset |
|
|
return train_dataset, test_dataset |
|
|
else: |
|
|
|
|
|
cache_dir = "./processed_datasets" |
|
|
os.makedirs(cache_dir, exist_ok=True) |
|
|
cache_file_train = os.path.join(cache_dir, "train.arrow") |
|
|
cache_file_test = os.path.join(cache_dir, "test.arrow") |
|
|
|
|
|
if os.path.exists(cache_file_train) and os.path.exists(cache_file_test): |
|
|
from datasets import Dataset |
|
|
train_dataset = Dataset.load_from_disk(cache_file_train) |
|
|
test_dataset = Dataset.load_from_disk(cache_file_test) |
|
|
return train_dataset, test_dataset |
|
|
|
|
|
def filter_func(x): |
|
|
return (0 < len(x["transcription"]) < 2048 and |
|
|
len(x["audio"]["array"]) > 0 and |
|
|
len(x["audio"]["array"]) < 2048 * 160) |
|
|
|
|
|
raw_train = load_dataset("google/fleurs", "en_us", token=token, split="train", trust_remote_code=True, streaming=streaming) |
|
|
raw_test = load_dataset("google/fleurs", "en_us", token=token, split="test", trust_remote_code=True, streaming=streaming) |
|
|
|
|
|
raw_train = raw_train.filter(filter_func) |
|
|
raw_test = raw_test.filter(filter_func) |
|
|
|
|
|
raw_train = raw_train.cast_column("audio", Audio(sampling_rate=sample_rate)) |
|
|
raw_test = raw_test.cast_column("audio", Audio(sampling_rate=sample_rate)) |
|
|
|
|
|
train_dataset = raw_train.map( |
|
|
lambda x: extract_features(x, tokenizer, **dataset_config), |
|
|
remove_columns=raw_train.column_names) |
|
|
test_dataset = raw_test.map( |
|
|
lambda x: extract_features(x, tokenizer, **dataset_config), |
|
|
remove_columns=raw_test.column_names) |
|
|
|
|
|
train_dataset.save_to_disk(cache_file_train) if sanity_check or streaming is False else None |
|
|
test_dataset.save_to_disk(cache_file_test) if sanity_check or streaming is False else None |
|
|
return train_dataset, test_dataset |
|
|
|
|
|
@dataclass |
|
|
class DataCollator: |
|
|
tokenizer: Any |
|
|
|
|
|
def __call__(self, features: List[Dict[str, torch.Tensor]]) -> Dict[str, torch.Tensor]: |
|
|
all_keys = set() |
|
|
for f in features: |
|
|
all_keys.update(f.keys()) |
|
|
batch = {} |
|
|
pad_token_id = getattr(self.tokenizer, 'pad_token_id', 0) |
|
|
bos_token_id = getattr(self.tokenizer, 'bos_token_id', 1) |
|
|
eos_token_id = getattr(self.tokenizer, 'eos_token_id', 2) |
|
|
|
|
|
for key in all_keys: |
|
|
if key == "labels": |
|
|
labels_list = [f["labels"] for f in features] |
|
|
max_len = max(len(l) for l in labels_list) |
|
|
all_ids, all_labels = [], [] |
|
|
for label in labels_list: |
|
|
label_list = label.tolist() if isinstance(label, torch.Tensor) else label |
|
|
decoder_input = [bos_token_id] + label_list |
|
|
label_eos = label_list + [eos_token_id] |
|
|
input_len = max_len + 1 - len(decoder_input) |
|
|
label_len = max_len + 1 - len(label_eos) |
|
|
padded_input = decoder_input + [pad_token_id] * input_len |
|
|
padded_labels = label_eos + [pad_token_id] * label_len |
|
|
all_ids.append(padded_input) |
|
|
all_labels.append(padded_labels) |
|
|
batch["input_ids"] = torch.tensor(all_ids, dtype=torch.long) |
|
|
batch["labels"] = torch.tensor(all_labels, dtype=torch.long) |
|
|
|
|
|
elif key in ["spectrogram", "waveform", "pitch", "harmonic", "aperiodic", "f0t", "f0"]: |
|
|
items = [f[key] for f in features if key in f] |
|
|
items = [item for item in items if item is not None] |
|
|
if not items: |
|
|
continue |
|
|
items = [torch.tensor(item) if not isinstance(item, torch.Tensor) else item for item in items] |
|
|
max_len = max(item.shape[-1] for item in items) |
|
|
padded = [] |
|
|
for item in items: |
|
|
pad_width = max_len - item.shape[-1] |
|
|
if pad_width > 0: |
|
|
pad_item = F.pad(item, (0, pad_width), mode='constant', value=pad_token_id) |
|
|
else: |
|
|
pad_item = item |
|
|
padded.append(pad_item) |
|
|
batch[key] = torch.stack(padded) |
|
|
if key == "spectrogram": |
|
|
batch["spectrogram"] = batch[key] |
|
|
return batch |
|
|
|
|
|
def levenshtein(reference_words, hypothesis_words): |
|
|
m, n = len(reference_words), len(hypothesis_words) |
|
|
dist_matrix = [[0 for _ in range(n+1)] for _ in range(m+1)] |
|
|
for i in range(m+1): |
|
|
dist_matrix[i][0] = i |
|
|
for j in range(n+1): |
|
|
dist_matrix[0][j] = j |
|
|
for i in range(1, m+1): |
|
|
for j in range(1, n+1): |
|
|
if reference_words[i-1] == hypothesis_words[j-1]: |
|
|
dist_matrix[i][j] = dist_matrix[i-1][j-1] |
|
|
else: |
|
|
substitution = dist_matrix[i-1][j-1] + 1 |
|
|
insertion = dist_matrix[i][j-1] + 1 |
|
|
deletion = dist_matrix[i-1][j] + 1 |
|
|
dist_matrix[i][j] = min(substitution, insertion, deletion) |
|
|
return dist_matrix[m][n] |
|
|
|
|
|
def wer_batch(references, hypotheses): |
|
|
total_errors = 0 |
|
|
total_words = 0 |
|
|
for ref, hyp in zip(references, hypotheses): |
|
|
ref_words = ref.lower().split() |
|
|
errors = levenshtein(ref_words, hyp.lower().split()) |
|
|
total_errors += errors |
|
|
total_words += len(ref_words) |
|
|
return (total_errors / total_words) * 100 if total_words > 0 else 0.0 |
|
|
|
|
|
def clean_ids(ids, pad_token_id=0, bos_token_id=1, eos_token_id=2): |
|
|
if isinstance(ids, torch.Tensor): |
|
|
ids = ids.tolist() |
|
|
return [int(id) for id in ids if id != -100 and id != pad_token_id and id != bos_token_id and id != eos_token_id] |
|
|
|
|
|
def clean_batch(batch_ids, pad_token_id=0, bos_token_id=1, eos_token_id=2): |
|
|
return [clean_ids(seq, pad_token_id, bos_token_id, eos_token_id) for seq in batch_ids] |
|
|
|
|
|
def compute_metrics(pred, tokenizer=None, model=None, print_pred=False, num_samples=0, optimizer=None, scheduler=None): |
|
|
label_ids = pred.label_ids |
|
|
pred_ids = pred.predictions[0] |
|
|
label_ids = clean_batch(label_ids) |
|
|
pred_ids = clean_batch(pred_ids) |
|
|
label_str = tokenizer.batch_decode(label_ids, skip_special_tokens=True) |
|
|
pred_str = tokenizer.batch_decode(pred_ids, skip_special_tokens=True) |
|
|
|
|
|
if print_pred: |
|
|
for i in range(min(num_samples, len(pred_ids))): |
|
|
print(f"Pred tokens: {pred_ids[i]}") |
|
|
print(f"Label tokens: {label_ids[i]}") |
|
|
print(f"Pred: '{pred_str[i]}'") |
|
|
print(f"Label: '{label_str[i]}'") |
|
|
print("-" * 40) |
|
|
|
|
|
wer = wer_batch(label_str, pred_str) |
|
|
if model is not None: |
|
|
trainable_params = sum(p.numel() for p in model.parameters() if p.requires_grad) / 1000000 |
|
|
efficiency_score = (100 - wer) / trainable_params if trainable_params > 0 else 0.0 |
|
|
else: |
|
|
trainable_params = 0.0 |
|
|
efficiency_score = 0.0 |
|
|
return { "wer": float(wer), "efficiency_score": float(efficiency_score)} |
|
|
|
|
|
def preprocess_logits_for_metrics(logits, labels): |
|
|
pred_ids = torch.argmax(logits, dim=-1) |
|
|
labels = torch.where(labels == -100, 0, labels) |
|
|
pred_ids = torch.where(pred_ids == -100, 0, pred_ids) |
|
|
return pred_ids, labels |
|
|
|
|
|
def main(): |
|
|
token = "" |
|
|
log_dir = os.path.join('./output/logs', datetime.now().strftime('%m-%d_%H_%M_%S')) |
|
|
os.makedirs(log_dir, exist_ok=True) |
|
|
tokenizer = setup_tokenizer(token) |
|
|
train_dataset, test_dataset = prepare_datasets( |
|
|
tokenizer, |
|
|
token, |
|
|
sanity_check=False, |
|
|
|
|
|
) |
|
|
|
|
|
param = Dimensions( |
|
|
vocab=40000, |
|
|
mels=128, |
|
|
ctx=1500, |
|
|
dims=512, |
|
|
head=4, |
|
|
layer=4, |
|
|
act="swish", |
|
|
debug={"radius", "encoder"}, |
|
|
features = ["spectrogram"], |
|
|
) |
|
|
|
|
|
model = Echo(param).to('cuda') |
|
|
print(f"Trainable parameters: {sum(p.numel() for p in model.parameters() if p.requires_grad):,}") |
|
|
print(f"Total parameters: {sum(p.numel() for p in model.parameters()):,}") |
|
|
|
|
|
training_args = Seq2SeqTrainingArguments( |
|
|
output_dir=log_dir, |
|
|
per_device_train_batch_size=1, |
|
|
per_device_eval_batch_size=1, |
|
|
max_steps=1000, |
|
|
eval_steps=100, |
|
|
save_steps=1000, |
|
|
warmup_steps=100, |
|
|
logging_steps=10, |
|
|
logging_dir=log_dir, |
|
|
eval_strategy="steps", |
|
|
save_strategy="no", |
|
|
report_to=["tensorboard"], |
|
|
push_to_hub=False, |
|
|
disable_tqdm=False, |
|
|
save_total_limit=1, |
|
|
label_names=["labels"], |
|
|
save_safetensors=False, |
|
|
eval_on_start=True, |
|
|
batch_eval_metrics=False, |
|
|
) |
|
|
from functools import partial |
|
|
metrics_fn = partial(compute_metrics, |
|
|
print_pred=True, |
|
|
num_samples=2, |
|
|
tokenizer=tokenizer, model=model) |
|
|
|
|
|
optimizer = torch.optim.AdamW(model.parameters(), lr=0.00025, eps=1e-8, weight_decay=0.025, betas=(0.9, 0.999), |
|
|
amsgrad=False, foreach=False, fused=False, capturable=False, differentiable=False, maximize=False) |
|
|
|
|
|
scheduler = torch.optim.lr_scheduler.CosineAnnealingLR(optimizer, T_max=training_args.max_steps, eta_min=1e-9, last_epoch=-1) |
|
|
|
|
|
trainer = Seq2SeqTrainer( |
|
|
args=training_args, |
|
|
model=model, |
|
|
train_dataset=train_dataset, |
|
|
eval_dataset=test_dataset, |
|
|
data_collator=DataCollator(tokenizer=tokenizer), |
|
|
preprocess_logits_for_metrics=preprocess_logits_for_metrics, |
|
|
compute_metrics=metrics_fn, |
|
|
optimizers=(optimizer, scheduler) |
|
|
) |
|
|
print(tokenizer.pad_token_id, tokenizer.bos_token_id, tokenizer.eos_token_id) |
|
|
model.init_weights() |
|
|
trainer.train() |
|
|
|
|
|
if __name__ == "__main__": |
|
|
main() |
|
|
|
|
|
|
