import functools
import math
import random
import torch
import torch.nn as nn
import torch.nn.functional as F
try:
import deepspeed
from deepspeed.ops.transformer.inference import DeepSpeedTransformerInferenceKernel
except ImportError:
pass
import dlas.codes.torch_intermediary as ml
from dlas.codes.models.arch_util import AttentionBlock
from dlas.codes.trainer.networks import register_model
from dlas.codes.utils.transformers.stream_generator import init_stream_support
from dlas.codes.utils.util import opt_get
from transformers import GPT2Config, GPT2PreTrainedModel
from transformers.modeling_outputs import CausalLMOutputWithCrossAttentions
init_stream_support()
def null_position_embeddings(range, dim):
return torch.zeros((range.shape[0], range.shape[1], dim), device=range.device)
class ResBlock(nn.Module):
"""
Basic residual convolutional block that uses GroupNorm.
"""
def __init__(self, chan):
super().__init__()
self.net = nn.Sequential(
nn.Conv1d(chan, chan, kernel_size=3, padding=1),
nn.GroupNorm(chan // 8, chan),
nn.ReLU(),
nn.Conv1d(chan, chan, kernel_size=3, padding=1),
nn.GroupNorm(chan // 8, chan),
)
def forward(self, x):
return F.relu(self.net(x) + x)
class GPT2InferenceModel(GPT2PreTrainedModel):
"""Override GPT2LMHeadModel to allow for prefix conditioning."""
def __init__(self, config, gpt, pos_emb, embeddings, norm, linear, kv_cache):
super().__init__(config)
self.transformer = gpt
self.pos_embedding = pos_emb
self.embeddings = embeddings
self.final_norm = norm
self.lm_head = nn.Sequential(norm, linear)
self.kv_cache = kv_cache
def store_prefix_emb(self, prefix_emb):
self.cached_prefix_emb = prefix_emb
def prepare_inputs_for_generation(self, input_ids, past_key_values=None, **kwargs):
token_type_ids = kwargs.get("token_type_ids", None) # usually None
if not self.kv_cache:
past_key_values = None
# only last token for inputs_ids if past is defined in kwargs
if past_key_values is not None:
input_ids = input_ids[:, -1].unsqueeze(-1)
if token_type_ids is not None:
token_type_ids = token_type_ids[:, -1].unsqueeze(-1)
attention_mask = kwargs.get("attention_mask", None)
position_ids = kwargs.get("position_ids", None)
if attention_mask is not None and position_ids is None:
# create position_ids on the fly for batch generation
position_ids = attention_mask.long().cumsum(-1) - 1
position_ids.masked_fill_(attention_mask == 0, 1)
if past_key_values is not None:
position_ids = position_ids[:, -1].unsqueeze(-1)
else:
position_ids = None
return {
"input_ids": input_ids,
"past_key_values": past_key_values,
"use_cache": kwargs.get("use_cache"),
"position_ids": position_ids,
"attention_mask": attention_mask,
"token_type_ids": token_type_ids,
}
def forward(
self,
input_ids=None,
past_key_values=None,
attention_mask=None,
token_type_ids=None,
position_ids=None,
head_mask=None,
inputs_embeds=None,
encoder_hidden_states=None,
encoder_attention_mask=None,
labels=None,
use_cache=None,
output_attentions=None,
output_hidden_states=None,
return_dict=None,
):
assert self.cached_prefix_emb is not None
assert inputs_embeds is None # Not supported by this inference model.
assert labels is None # Training not supported by this inference model.
return_dict = return_dict if return_dict is not None else self.config.use_return_dict
# assert len(past_key_values) + len(input_ids) == attention_mask.shape[1]
# Create embedding
prefix_len = self.cached_prefix_emb.shape[1]
if input_ids.shape[1] != 1:
gen_inputs = input_ids[:, prefix_len:]
gen_emb = self.embeddings(gen_inputs)
gen_emb = gen_emb + self.pos_embedding(gen_emb)
if self.cached_prefix_emb.shape[0] != gen_emb.shape[0]:
prefix_emb = self.cached_prefix_emb.repeat_interleave(
gen_emb.shape[0] // self.cached_prefix_emb.shape[0], 0
)
else:
prefix_emb = self.cached_prefix_emb.to(gen_emb.dtype)
emb = torch.cat([prefix_emb, gen_emb], dim=1)
else:
emb = self.embeddings(input_ids)
emb = emb + self.pos_embedding.get_fixed_embedding(
attention_mask.shape[1] - (prefix_len + 1), attention_mask.device
)
transformer_outputs = self.transformer(
inputs_embeds=emb,
past_key_values=past_key_values,
attention_mask=attention_mask,
token_type_ids=token_type_ids,
position_ids=position_ids,
head_mask=head_mask,
encoder_hidden_states=encoder_hidden_states,
encoder_attention_mask=encoder_attention_mask,
use_cache=use_cache,
output_attentions=output_attentions,
output_hidden_states=output_hidden_states,
return_dict=return_dict,
)
hidden_states = transformer_outputs[0]
lm_logits = self.lm_head(hidden_states)
if not return_dict:
return (lm_logits,) + transformer_outputs[1:]
return CausalLMOutputWithCrossAttentions(
loss=None,
logits=lm_logits,
past_key_values=transformer_outputs.past_key_values,
hidden_states=transformer_outputs.hidden_states,
attentions=transformer_outputs.attentions,
cross_attentions=transformer_outputs.cross_attentions,
)
@staticmethod
def _reorder_cache(past, beam_idx):
"""
This function is used to re-order the :obj:`past_key_values` cache if
:meth:`~transformers.PreTrainedModel.beam_search` or :meth:`~transformers.PreTrainedModel.beam_sample` is
called. This is required to match :obj:`past_key_values` with the correct beam_idx at every generation step.
"""
return tuple(
tuple(past_state.index_select(0, beam_idx.to(past_state.device)) for past_state in layer_past)
for layer_past in past
)
class ConditioningEncoder(nn.Module):
def __init__(
self,
spec_dim,
embedding_dim,
attn_blocks=6,
num_attn_heads=4,
do_checkpointing=False,
mean=False,
):
super().__init__()
attn = []
self.init = nn.Conv1d(spec_dim, embedding_dim, kernel_size=1)
for a in range(attn_blocks):
attn.append(AttentionBlock(embedding_dim, num_attn_heads, do_checkpoint=do_checkpointing))
self.attn = nn.Sequential(*attn)
self.dim = embedding_dim
self.do_checkpointing = do_checkpointing
self.mean = mean
def forward(self, x):
h = self.init(x)
h = self.attn(h)
if self.mean:
return h.mean(dim=2)
else:
return h[:, :, 0]
class LearnedPositionEmbeddings(nn.Module):
def __init__(self, seq_len, model_dim, init=0.02, relative=False):
super().__init__()
# nn.Embedding
self.emb = torch.nn.Embedding(seq_len, model_dim)
# Initializing this way is standard for GPT-2
self.emb.weight.data.normal_(mean=0.0, std=init)
self.relative = relative
self.seq_len = seq_len
def forward(self, x):
sl = x.shape[1]
if self.relative:
start = random.randint(sl, self.seq_len) - sl
return self.emb(torch.arange(start, start + sl, device=x.device))
else:
return self.emb(torch.arange(0, sl, device=x.device))
def get_fixed_embedding(self, ind, dev):
return self.emb(torch.tensor([ind], device=dev)).unsqueeze(0)
def build_hf_gpt_transformer(
layers,
model_dim,
heads,
max_mel_seq_len,
max_text_seq_len,
max_prompt_len,
checkpointing,
):
"""
GPT-2 implemented by the HuggingFace library.
"""
from transformers import GPT2Config, GPT2Model
gpt_config = GPT2Config(
vocab_size=256, # Unused.
n_positions=max_mel_seq_len + max_text_seq_len + max_prompt_len,
n_ctx=max_mel_seq_len + max_text_seq_len + max_prompt_len,
n_embd=model_dim,
n_layer=layers,
n_head=heads,
gradient_checkpointing=checkpointing,
use_cache=not checkpointing,
)
gpt = GPT2Model(gpt_config)
# Override the built in positional embeddings
del gpt.wpe
gpt.wpe = functools.partial(null_position_embeddings, dim=model_dim)
# Built-in token embeddings are unused.
del gpt.wte
# def _attn(self, query, key, value, attention_mask=None, head_mask=None):
# attn_output = torch.nn.functional.scaled_dot_product_attention(
# query, key, value, dropout_p=self.attn_dropout.p, is_causal=True
# )
# return attn_output, None
# for i in range(len(gpt.h)):
# gpt.h[i].attn._attn = types.MethodType(
# _attn, gpt.h[i].attn
# )
mel_pos_emb = (
LearnedPositionEmbeddings(max_mel_seq_len, model_dim)
if max_mel_seq_len != -1
else functools.partial(null_position_embeddings, dim=model_dim)
)
text_pos_emb = (
LearnedPositionEmbeddings(max_text_seq_len, model_dim)
if max_mel_seq_len != -1
else functools.partial(null_position_embeddings, dim=model_dim)
)
# gpt = torch.compile(gpt, mode="reduce-overhead", fullgraph=True)
return gpt, mel_pos_emb, text_pos_emb, None, None
class MelEncoder(nn.Module):
def __init__(self, channels, mel_channels=80, resblocks_per_reduction=2):
super().__init__()
self.channels = channels
self.encoder = nn.Sequential(
nn.Conv1d(mel_channels, channels // 4, kernel_size=3, padding=1),
nn.Sequential(*[ResBlock(channels // 4) for _ in range(resblocks_per_reduction)]),
nn.Conv1d(channels // 4, channels // 2, kernel_size=3, stride=2, padding=1),
nn.GroupNorm(channels // 16, channels // 2),
nn.ReLU(),
nn.Sequential(*[ResBlock(channels // 2) for _ in range(resblocks_per_reduction)]),
nn.Conv1d(channels // 2, channels, kernel_size=3, stride=2, padding=1),
nn.GroupNorm(channels // 8, channels),
nn.ReLU(),
nn.Sequential(*[ResBlock(channels) for _ in range(resblocks_per_reduction)]),
)
self.reduction = 4
def forward(self, x):
for e in self.encoder:
x = e(x)
return x.permute(0, 2, 1)
class UnifiedVoice(nn.Module):
def __init__(
self,
start_text_token=261,
stop_text_token=0,
layers=8,
model_dim=512,
heads=8,
max_text_tokens=120,
max_mel_tokens=250,
max_prompt_tokens=70,
max_conditioning_inputs=1,
mel_length_compression=1024,
number_text_tokens=256,
number_mel_codes=8194,
start_mel_token=8192,
stop_mel_token=8193,
train_solo_embeddings=False,
use_mel_codes_as_input=True,
checkpointing=True,
average_conditioning_embeddings=False,
freeze_everything_but_position_embeddings=False,
freeze_conditioning_encoder=False,
tortoise_compat=True,
label_smoothing=0.0,
):
"""
Args:
layers: Number of layers in transformer stack.
model_dim: Operating dimensions of the transformer
heads: Number of transformer heads. Must be divisible by model_dim. Recommend model_dim//64
max_text_tokens: Maximum number of text tokens that will be encountered by model.
max_mel_tokens: Maximum number of MEL tokens that will be encountered by model.
max_conditioning_inputs: Maximum number of conditioning inputs provided to the model. If (1), conditioning input can be of format (b,80,s), otherwise (b,n,80,s).
mel_length_compression: The factor between <number_input_samples> and <mel_tokens>. Used to compute MEL code padding given wav input length.
number_text_tokens:
start_text_token:
stop_text_token:
number_mel_codes:
start_mel_token:
stop_mel_token:
train_solo_embeddings:
use_mel_codes_as_input:
checkpointing:
average_conditioning_embeddings: Whether or not conditioning embeddings should be averaged, instead of fed piecewise into the model.
"""
super().__init__()
self.label_smoothing = label_smoothing
self.number_text_tokens = number_text_tokens
self.start_text_token = start_text_token
self.stop_text_token = stop_text_token
self.number_mel_codes = number_mel_codes
self.start_mel_token = start_mel_token
self.stop_mel_token = stop_mel_token
self.start_prompt_token = start_mel_token
self.stop_prompt_token = stop_mel_token
self.layers = layers
self.heads = heads
self.model_dim = model_dim
self.max_conditioning_inputs = max_conditioning_inputs
self.max_mel_tokens = -1 if max_mel_tokens == -1 else max_mel_tokens + 2 + self.max_conditioning_inputs
self.max_text_tokens = -1 if max_text_tokens == -1 else max_text_tokens + 2
self.max_prompt_tokens = max_prompt_tokens
self.mel_length_compression = mel_length_compression
# self.conditioning_encoder = ConditioningEncoder(
# 80, model_dim, num_attn_heads=heads
# )
self.average_conditioning_embeddings = average_conditioning_embeddings
self.tortoise_compat = tortoise_compat # credit to https://github.com/152334H/DL-Art-School/commit/ae80992817059acf6eef38a680efa5124cee570b
# nn.Embedding
self.text_embedding = ml.Embedding(self.number_text_tokens, model_dim)
if use_mel_codes_as_input:
# nn.Embedding
self.mel_embedding = ml.Embedding(self.number_mel_codes, model_dim)
else:
self.mel_embedding = MelEncoder(model_dim, resblocks_per_reduction=1)
(
self.gpt,
self.mel_pos_embedding,
self.text_pos_embedding,
self.mel_layer_pos_embedding,
self.text_layer_pos_embedding,
) = build_hf_gpt_transformer(
layers,
model_dim,
heads,
self.max_mel_tokens,
self.max_text_tokens,
self.max_prompt_tokens,
checkpointing,
)
if train_solo_embeddings:
self.mel_solo_embedding = nn.Parameter(torch.randn(1, 1, model_dim) * 0.02, requires_grad=True)
self.text_solo_embedding = nn.Parameter(torch.randn(1, 1, model_dim) * 0.02, requires_grad=True)
else:
self.mel_solo_embedding = 0
self.text_solo_embedding = 0
self.final_norm = nn.LayerNorm(model_dim)
self.text_head = ml.Linear(model_dim, self.number_text_tokens)
self.mel_head = ml.Linear(model_dim, self.number_mel_codes)
# Initialize the embeddings per the GPT-2 scheme
embeddings = [self.text_embedding]
if use_mel_codes_as_input:
embeddings.append(self.mel_embedding)
for module in embeddings:
module.weight.data.normal_(mean=0.0, std=0.02)
if freeze_conditioning_encoder:
print(" > Freezing conditioning encoder.")
for p in self.conditioning_encoder.parameters():
p.requires_grad = False
p.DO_NOT_TRAIN = True
if freeze_everything_but_position_embeddings:
for p in self.parameters():
p.requires_grad = False
p.DO_NOT_TRAIN = True
for m in [self.mel_pos_embedding, self.text_pos_embedding]:
for p in m.parameters():
del p.DO_NOT_TRAIN
p.requires_grad = True
def get_grad_norm_parameter_groups(self):
return {
"conditioning_encoder": list(self.conditioning_encoder.parameters()),
"gpt": list(self.gpt.parameters()),
"heads": list(self.text_head.parameters()) + list(self.mel_head.parameters()),
}
def post_init_gpt2_config(self, kv_cache=True, use_deepspeed=False, use_deepspeed_f16=False):
seq_length = self.max_prompt_tokens + self.max_mel_tokens + self.max_text_tokens + 1
gpt_config = GPT2Config(
vocab_size=self.max_mel_tokens,
n_positions=seq_length,
n_ctx=seq_length,
n_embd=self.model_dim,
n_layer=self.layers,
n_head=self.heads,
gradient_checkpointing=False,
use_cache=True,
)
self.inference_model = GPT2InferenceModel(
gpt_config,
self.gpt,
self.mel_pos_embedding,
self.mel_embedding,
self.final_norm,
self.mel_head,
kv_cache=kv_cache,
)
# self.inference_model = PrunedGPT2InferenceModel(gpt_config, self.gpt, self.mel_pos_embedding, self.mel_embedding, self.final_norm, self.mel_head)
self.gpt.wte = self.mel_embedding
if use_deepspeed:
# init deepspeed inference engine
if use_deepspeed_f16:
self.gpt.wte = self.mel_embedding.half()
self.gpt.wpe = self.mel_pos_embedding.half()
self.ds_engine = deepspeed.init_inference(
model=self.inference_model.half(), # Transformers models
mp_size=1, # Number of GPU
dtype=torch.float16 if use_deepspeed_f16 else torch.float32, # desired data type of output
replace_method="auto", # Lets DS autmatically identify the layer to replace
replace_with_kernel_inject=True, # replace the model with the kernel injector
)
self.inference_model = self.ds_engine.module.eval()
def build_aligned_inputs_and_targets(self, input, start_token, stop_token):
inp = F.pad(input, (1, 0), value=start_token)
tar = F.pad(input, (0, 1), value=stop_token)
return inp, tar
def set_mel_padding(self, mel_input_tokens, mel_lengths):
"""
Given mel tokens that are derived from a padded audio clip and the actual lengths of each batch element in
that audio clip, reformats the tokens with STOP_MEL_TOKEN in place of the zero padding. This is required
preformatting to create a working TTS model.
"""
# Set padding areas within MEL (currently it is coded with the MEL code for <zero>).
for b in range(len(mel_lengths)):
actual_end = mel_lengths[b]
if actual_end < mel_input_tokens.shape[-1]:
mel_input_tokens[b, actual_end:] = self.stop_mel_token
return mel_input_tokens
def get_logits(
self,
speech_conditioning_inputs,
first_inputs,
first_head,
second_inputs=None,
second_head=None,
prompt=None,
get_attns=False,
return_latent=False,
attn_mask_text=None,
attn_mask_mel=None,
):
if prompt is not None and speech_conditioning_inputs is not None:
offset = speech_conditioning_inputs.shape[1] + prompt.shape[1]
if second_inputs is not None:
emb = torch.cat(
[speech_conditioning_inputs, prompt, first_inputs, second_inputs],
dim=1,
)
else:
emb = torch.cat([speech_conditioning_inputs, prompt, first_inputs], dim=1)
elif speech_conditioning_inputs is not None:
offset = speech_conditioning_inputs.shape[1]
if second_inputs is not None:
emb = torch.cat([speech_conditioning_inputs, first_inputs, second_inputs], dim=1)
else:
emb = torch.cat([speech_conditioning_inputs, first_inputs], dim=1)
elif prompt is not None:
offset = prompt.shape[1]
if second_inputs is not None:
emb = torch.cat([prompt, first_inputs, second_inputs], dim=1)
else:
emb = torch.cat([prompt, first_inputs], dim=1)
# with torch.backends.cuda.sdp_kernel(enable_flash=True, enable_math=False, enable_mem_efficient=False):
attn_mask = None
if attn_mask_text is not None:
attn_mask = torch.cat([attn_mask_text, attn_mask_mel], dim=1)
if prompt is not None:
attn_mask_prompt = torch.ones(prompt.shape[0], offset, dtype=torch.bool, device=emb.device)
attn_mask = torch.cat([attn_mask_prompt, attn_mask], dim=1)
gpt_out = self.gpt(
inputs_embeds=emb,
return_dict=True,
output_attentions=get_attns,
attention_mask=attn_mask,
)
if get_attns:
return gpt_out.attentions
enc = gpt_out.last_hidden_state[:, offset:]
enc = self.final_norm(enc)
if return_latent:
return enc[:, : first_inputs.shape[1]], enc[:, -second_inputs.shape[1] :]
first_logits = enc[:, : first_inputs.shape[1]]
first_logits = first_head(first_logits)
first_logits = first_logits.permute(0, 2, 1)
if second_inputs is not None:
second_logits = enc[:, -second_inputs.shape[1] :]
second_logits = second_head(second_logits)
second_logits = second_logits.permute(0, 2, 1)
return first_logits, second_logits
else:
return first_logits
def get_conditioning(self, speech_conditioning_input):
speech_conditioning_input = (
speech_conditioning_input.unsqueeze(1)
if len(speech_conditioning_input.shape) == 3
else speech_conditioning_input
)
conds = []
for j in range(speech_conditioning_input.shape[1]):
conds.append(self.conditioning_encoder(speech_conditioning_input[:, j]))
conds = torch.stack(conds, dim=1)
conds = conds.mean(dim=1)
return conds
def get_prompts(self, prompt_codes):
"""
Create a prompt from the mel codes. This is used to condition the model on the mel codes.
Pad the prompt with start and stop mel tokens.
"""
prompt = prompt_codes
if self.training:
prompt_len = random.randint(1, 9) # in secs
prompt_len = prompt_len * 24 # in frames
if prompt_codes.shape[1] < prompt_len:
prompt_len = prompt_codes.shape[-1]
start = 0
else:
start = random.randint(0, prompt_codes.shape[-1] - prompt_len)
prompt = prompt_codes[:, start : start + prompt_len]
# add start and stop tokens
prompt = F.pad(prompt, (1, 0), value=self.start_prompt_token)
prompt = F.pad(prompt, (0, 1), value=self.stop_prompt_token)
return prompt
# def get_prompts(self, prompt_codes):
# """
# Create a prompt from the mel codes. This is used to condition the model on the mel codes.
# Pad the prompt with start and stop mel tokens.
# """
# prompt = prompt_codes
# if self.training:
# max_prompt_len = 9 * 24
# if prompt_codes.shape[1] < max_prompt_len:
# prompt = prompt_codes
# else:
# start = random.randint(0, prompt_codes.shape[1] - max_prompt_len)
# prompt = prompt_codes[:, start : start + max_prompt_len]
# # add start and stop tokens
# prompt = F.pad(prompt, (1, 0), value=self.start_prompt_token)
# prompt = F.pad(prompt, (0, 1), value=self.stop_prompt_token)
# return prompt
def forward(
self,
speech_conditioning_input,
text_inputs,
text_lengths,
mel_codes,
wav_lengths,
prompt_codes,
loss_weights=None,
text_first=True,
return_attentions=False,
return_latent=False,
):
"""
Forward pass that uses both text and voice in either text conditioning mode or voice conditioning mode
(actuated by `text_first`).
speech_conditioning_input: MEL float tensor, (b,80,s)
text_inputs: long tensor, (b,t)
text_lengths: long tensor, (b,)
mel_inputs: long tensor, (b,m)
wav_lengths: long tensor, (b,)
If return_attentions is specified, only logits are returned.
If return_latent is specified, loss & logits are not computed or returned. Only the predicted latents are returned.
"""
# ❗ FIXIT
speech_conditioning_input = None
if self.max_conditioning_inputs == 0:
assert (
speech_conditioning_input is None
), " ❗ speech_conditioning_input is not None, but max_conditioning_inputs == 0"
max_text_len = text_lengths.max()
# Due to the convolution in DVAE, codes do not end with silence at the right place. Rather it predicts some intermediate values
# Like [..., 186, 45, 45, 83] where actually it should end with 186.
# We take last 3 codes to prevent abrupt ending of the audio.
# TODO: This is might need some testing.
mel_lengths = torch.ceil(wav_lengths / self.mel_length_compression).long() + 3
# If len(codes) + 3 is larger than maxiumum allowed length, we truncate the codes.
max_mel_len = mel_lengths.max()
if max_mel_len > mel_codes.shape[-1]:
mel_codes = F.pad(mel_codes, (0, max_mel_len - mel_codes.shape[-1]))
# mel_lengths[mel_lengths >= max_mel_len] = max_mel_len
# silence aware lengths, skip the silence tokens at the end of the mel codes.
silence = True
for idx, l in enumerate(mel_lengths):
length = l.item()
while silence:
if mel_codes[idx, length - 1] != 83:
break
length -= 1
mel_lengths[idx] = length
# Lovely assertions
assert (
max_mel_len <= mel_codes.shape[-1]
), f" ❗ max_mel_len ({max_mel_len}) > mel_codes.shape[-1] ({mel_codes.shape[-1]})"
assert (
max_text_len <= text_inputs.shape[-1]
), f" ❗ max_text_len ({max_text_len}) > text_inputs.shape[-1] ({text_inputs.shape[-1]})"
# Append stop token to text inputs
text_inputs = F.pad(text_inputs[:, :max_text_len], (0, 1), value=self.stop_text_token)
# Append silence token to mel codes
mel_codes = F.pad(mel_codes[:, :max_mel_len], (0, 1), value=self.stop_mel_token)
# Pad mel codes with STOP_MEL_TOKEN
mel_codes = self.set_mel_padding(mel_codes, mel_lengths)
# Compute speech conditioning input
conds = None
if speech_conditioning_input is not None:
if not return_latent:
# Compute speech conditioning input
speech_conditioning_input = (
speech_conditioning_input.unsqueeze(1)
if len(speech_conditioning_input.shape) == 3
else speech_conditioning_input
)
conds = []
for j in range(speech_conditioning_input.shape[1]):
conds.append(self.conditioning_encoder(speech_conditioning_input[:, j]))
conds = torch.stack(conds, dim=1)
if self.average_conditioning_embeddings:
conds = conds.mean(dim=1).unsqueeze(1)
else:
# already computed
conds = speech_conditioning_input.unsqueeze(1)
# Build input and target tensors
# Prepend start token to inputs and append stop token to targets
text_inputs, text_targets = self.build_aligned_inputs_and_targets(
text_inputs, self.start_text_token, self.stop_text_token
)
mel_codes, mel_targets = self.build_aligned_inputs_and_targets(
mel_codes, self.start_mel_token, self.stop_mel_token
)
# Set attn_mask
attn_mask_text = None
attn_mask_mel = None
if not return_latent:
attn_mask_text = torch.ones(
text_inputs.shape[0],
text_inputs.shape[1],
dtype=torch.bool,
device=text_inputs.device,
)
attn_mask_mel = torch.ones(
mel_codes.shape[0],
mel_codes.shape[1],
dtype=torch.bool,
device=mel_codes.device,
)
for idx, l in enumerate(text_lengths):
attn_mask_text[idx, l + 1 :] = 0.0
for idx, l in enumerate(mel_lengths):
attn_mask_mel[idx, l + 1 :] = 0.0
# Compute text embeddings + positional embeddings
# print(" > text input latent:", text_inputs)
text_emb = self.text_embedding(text_inputs) + self.text_pos_embedding(text_inputs)
# Compute mel embeddings + positional embeddings
mel_emb = self.mel_embedding(mel_codes) + self.mel_pos_embedding(mel_codes)
# Compute prompt embeddings + positional embeddings
prompt = self.get_prompts(prompt_codes)
prompt_emb = self.mel_embedding(prompt).detach() + self.mel_pos_embedding(prompt).detach()
# Get logits
sub = -4 # don't ask me why 😄
if self.training:
sub = -1
text_logits, mel_logits = self.get_logits(
conds,
text_emb,
self.text_head,
mel_emb,
self.mel_head,
prompt=prompt_emb,
get_attns=return_attentions,
return_latent=return_latent,
attn_mask_text=attn_mask_text,
attn_mask_mel=attn_mask_mel,
)
if return_latent:
return mel_logits[:, :sub] # sub to prevent bla.
if return_attentions:
return mel_logits
# Set paddings to -1 to ignore them in loss
for idx, l in enumerate(text_lengths):
text_targets[idx, l + 1 :] = -1
for idx, l in enumerate(mel_lengths):
mel_targets[idx, l + 1 :] = -1
# check if stoptoken is in every row of mel_targets
assert (mel_targets == self.stop_mel_token).sum() >= mel_targets.shape[
0
], f" ❗ mel_targets does not contain stop token ({self.stop_mel_token}) in every row."
# Compute losses
loss_text = F.cross_entropy(
text_logits, text_targets.long(), ignore_index=-1, label_smoothing=self.label_smoothing
)
loss_mel = F.cross_entropy(
mel_logits, mel_targets.long(), ignore_index=-1, label_smoothing=self.label_smoothing
)
# if loss_weights is not None:
# loss_text = loss_text * loss_weights[:, None]
# loss_mel = loss_mel * loss_weights[:, None]
return loss_text.mean(), loss_mel.mean(), mel_logits
def text_forward(self, speech_conditioning_input, text_inputs, text_lengths):
"""
Performs autoregressive modeling on only text. Still requires a speech_conditioning_input due to the way the
model inputs are formatted. Just provide any audio clip (arguably, zeros could be provided).
"""
# This model will receive micro-batches with a ton of padding for both the text and MELs. Ameliorate this by
# chopping the inputs by the maximum actual length.
max_text_len = text_lengths.max()
text_inputs = F.pad(text_inputs[:, :max_text_len], (0, 1), value=self.stop_text_token)
speech_conditioning_input = (
speech_conditioning_input.unsqueeze(1)
if len(speech_conditioning_input.shape) == 3
else speech_conditioning_input
)
conds = []
for j in range(speech_conditioning_input.shape[1]):
conds.append(self.conditioning_encoder(speech_conditioning_input[:, j]))
conds = torch.stack(conds, dim=1)
if self.average_conditioning_embeddings:
conds = conds.mean(dim=1).unsqueeze(1)
text_inputs, text_targets = self.build_aligned_inputs_and_targets(
text_inputs, self.start_text_token, self.stop_text_token
)
text_emb = self.text_embedding(text_inputs) + self.text_pos_embedding(text_inputs) + self.text_solo_embedding
text_logits = self.get_logits(conds, text_emb, self.text_head)
loss_text = F.cross_entropy(text_logits, text_targets.long())
return loss_text.mean()
def speech_forward(self, speech_conditioning_input, mel_codes, wav_lengths, raw_mels=None):
"""
Performs autoregressive modeling on only speech data.
"""
assert self.max_mel_tokens >= mel_codes.shape[1], f"{mel_codes.shape[1]}"
# This model will receive micro-batches with a ton of padding for both the text and MELs. Ameliorate this by
# chopping the inputs by the maximum actual length.
max_mel_len = wav_lengths.max() // self.mel_length_compression
mel_codes = F.pad(mel_codes[:, :max_mel_len], (0, 1), value=self.stop_mel_token)
mel_codes = self.set_mel_padding(mel_codes, wav_lengths)
if raw_mels is not None:
raw_mels = raw_mels[:, :, : max_mel_len * 4]
speech_conditioning_input = (
speech_conditioning_input.unsqueeze(1)
if len(speech_conditioning_input.shape) == 3
else speech_conditioning_input
)
conds = []
for j in range(speech_conditioning_input.shape[1]):
conds.append(self.conditioning_encoder(speech_conditioning_input[:, j]))
conds = torch.stack(conds, dim=1)
if self.average_conditioning_embeddings:
conds = conds.mean(dim=1).unsqueeze(1)
mel_codes, mel_targets = self.build_aligned_inputs_and_targets(
mel_codes, self.start_mel_token, self.stop_mel_token
)
if raw_mels is not None:
mel_inp = F.pad(raw_mels, (0, 4))
else:
mel_inp = mel_codes
mel_emb = self.mel_embedding(mel_inp)
mel_emb = mel_emb + self.mel_pos_embedding(mel_codes) + self.mel_solo_embedding
mel_logits = self.get_logits(conds, mel_emb, self.mel_head)
loss_mel = F.cross_entropy(mel_logits, mel_targets.long())
return loss_mel.mean()
def get_generator(self, fake_inputs, **hf_generate_kwargs):
return self.inference_model.generate_stream(
fake_inputs,
bos_token_id=self.start_mel_token,
pad_token_id=self.stop_mel_token,
eos_token_id=self.stop_mel_token,
max_length=self.max_mel_tokens * 2 + self.max_prompt_tokens + self.max_text_tokens,
do_stream=True,
**hf_generate_kwargs,
)
def compute_embeddings(
self,
speech_conditioning_latent,
text_inputs,
input_tokens=None,
prompt_codes=None,
pad_input_text=False,
):
if pad_input_text and text_inputs.shape[1] < 250:
text_inputs = F.pad(text_inputs, (0, 250 - text_inputs.shape[1]), value=self.stop_text_token)
else:
text_inputs = F.pad(text_inputs, (0, 1), value=self.stop_text_token)
text_inputs = F.pad(text_inputs, (1, 0), value=self.start_text_token)
emb = self.text_embedding(text_inputs) + self.text_pos_embedding(text_inputs)
print(" > Text inputs:", text_inputs)
if prompt_codes is not None:
prompt_codes = self.get_prompts(prompt_codes)
prompt_emb = self.mel_embedding(prompt_codes) + self.mel_pos_embedding(prompt_codes)
print(" > Prompt inputs:", prompt_codes)
print(" > Prompt inputs shape:", prompt_codes.shape)
emb = torch.cat([prompt_emb, emb], dim=1)
if speech_conditioning_latent is not None:
conds = speech_conditioning_latent.unsqueeze(1)
emb = torch.cat([conds, emb], dim=1)
self.inference_model.store_prefix_emb(emb)
fake_inputs = torch.full(
(
emb.shape[0],
emb.shape[1] + 1, # +1 for the start_mel_token
),
fill_value=1,
dtype=torch.long,
device=text_inputs.device,
)
fake_inputs[:, -1] = self.start_mel_token
if input_tokens is not None:
fake_inputs = torch.cat([fake_inputs, input_tokens], dim=1)
return fake_inputs
def inference_speech(
self,
speech_conditioning_latent,
text_inputs,
input_tokens=None,
prompt_codes=None,
pad_input_text=False,
**hf_generate_kwargs,
):
if pad_input_text and text_inputs.shape[1] < 250:
text_inputs = F.pad(text_inputs, (0, 250 - text_inputs.shape[1]), value=self.stop_text_token)
else:
text_inputs = F.pad(text_inputs, (0, 1), value=self.stop_text_token)
text_inputs = F.pad(text_inputs, (1, 0), value=self.start_text_token)
emb = self.text_embedding(text_inputs) + self.text_pos_embedding(text_inputs)
print(" > Text inputs:", text_inputs)
if prompt_codes is not None:
prompt_codes = self.get_prompts(prompt_codes)
prompt_emb = self.mel_embedding(prompt_codes) + self.mel_pos_embedding(prompt_codes)
print(" > Prompt inputs:", prompt_codes)
print(" > Prompt inputs shape:", prompt_codes.shape)
emb = torch.cat([prompt_emb, emb], dim=1)
if speech_conditioning_latent is not None:
conds = speech_conditioning_latent.unsqueeze(1)
emb = torch.cat([conds, emb], dim=1)
self.inference_model.store_prefix_emb(emb)
fake_inputs = torch.full(
(
emb.shape[0],
emb.shape[1] + 1, # +1 for the start_mel_token
),
fill_value=1,
dtype=torch.long,
device=text_inputs.device,
)
fake_inputs[:, -1] = self.start_mel_token
if input_tokens is not None:
fake_inputs = torch.cat([fake_inputs, input_tokens], dim=1)
gen = self.inference_model.generate(
fake_inputs,
bos_token_id=self.start_mel_token,
pad_token_id=self.stop_mel_token,
eos_token_id=self.stop_mel_token,
max_length=self.max_mel_tokens * 2 + self.max_prompt_tokens + self.max_text_tokens,
**hf_generate_kwargs,
)
if "return_dict_in_generate" in hf_generate_kwargs:
return gen.sequences[:, fake_inputs.shape[1] :], gen
return gen[:, fake_inputs.shape[1] :]
# Turns the (utterly insane) output of HF.generate() into a far more sane output:
# [tensors(B,H,S,S)]. Outer=layers, B=batch,H=head,S=sequence
def make_hf_generate_attentions_sane(self, attentions):
layers = [[] for _ in range(len(attentions[0]))]
full_attention_size = attentions[-1][0].shape[-1]
for i, gen in enumerate(attentions):
for j, lyr in enumerate(gen):
layers[j].append(F.pad(lyr, (0, full_attention_size - lyr.shape[-1])))
catted = []
for lyr in layers:
catted.append(torch.cat(lyr, dim=2))
return catted
def convert_attentions_to_aligned_codes(self, text, attentions, codes, num_conds):
"""
This was an attempt to make some sense out of the attention matrix retrieved from the unified_voice model. Unfortunately, I can't use it for aligning text & voice.
"""
text_padding = num_conds + 2
num_text = text.shape[-1]
num_context = num_text + text_padding
assert num_context + 1 == attentions[0][0].shape[-1]
attentions = self.make_hf_generate_attentions_sane(attentions)
results = [torch.empty_like(codes) for _ in range(len(attentions))]
for l, layer in enumerate(attentions):
dec_context = layer[:, :, num_context:, :]
# Mask out everything that isn't text (including the start token, which gets a LOT of attention)
dec_context[:, :, :, : text_padding + 1] = 0
dec_context[:, :, :, num_context:] = 0
for h in range(dec_context.shape[1]):
dec_context_indices = torch.argmax(dec_context[0, h], dim=-1)
print(f"layer_{l};head_{h}: " + str(dec_context_indices))
for t, att_tok in enumerate(attentions):
combined_attention_weights = torch.zeros((codes.shape[0], num_text), device=codes.device)
for lyr in att_tok:
token_to_text_attentions = lyr[:, :, -1, text_padding : (text_padding + num_text)].sum(dim=1)
combined_attention_weights = combined_attention_weights + token_to_text_attentions
break
most_attended_text_token = combined_attention_weights.argmax(dim=-1)
results[:, t] = most_attended_text_token
eos_token_mask = codes != self.stop_mel_token
return results * eos_token_mask
@register_model
def register_unified_voice_prompt(opt_net, opt):
return UnifiedVoice(**opt_get(opt_net, ["kwargs"], {}))
if __name__ == "__main__":
gpt = UnifiedVoice(
model_dim=256,
heads=4,
train_solo_embeddings=True,
use_mel_codes_as_input=True,
max_conditioning_inputs=4,
freeze_everything_but_position_embeddings=True,
)
l = gpt(
torch.randn(2, 3, 80, 800),
torch.randint(high=256, size=(2, 120)),
torch.tensor([32, 120]),
torch.randint(high=8192, size=(2, 250)),
torch.tensor([250 * 256, 195 * 256]),
)
# gpt.text_forward(torch.randn(2,80,800), torch.randint(high=50, size=(2,80)), torch.tensor([32, 80]))