File size: 17,649 Bytes
33d9042
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
# AUTOGENERATED! DO NOT EDIT! File to edit: ../nbs/5B. Text to semantic token modeling.ipynb.

# %% auto 0
__all__ = ['load_datasets', 'rand', 'Tunables', 'Encoder', 'Decoder', 'TSARTransformer', 'make_model']

# %% ../nbs/5B. Text to semantic token modeling.ipynb 1
import dataclasses
import random
import math
import torch
import torch.nn as nn
import torch.nn.functional as F
from torch.profiler import record_function

from huggingface_hub import hf_hub_download
from fastcore.basics import store_attr
from fastprogress import progress_bar

import webdataset as wds

# %% ../nbs/5B. Text to semantic token modeling.ipynb 2
from pathlib import Path
import pylab as plt
import pandas as pd
import numpy as np

# %% ../nbs/5B. Text to semantic token modeling.ipynb 3
import whisper
from whisperspeech.train import *
from whisperspeech.modules import *
from whisperspeech import vq_stoks

# %% ../nbs/5B. Text to semantic token modeling.ipynb 8
import re

class CharTokenizer:
    """Trivial tokenizer – just use UTF-8 bytes"""
    eot = 0
    
    def encode(self, txt):
        return list(bytes(txt.strip(), 'utf-8'))

    def decode(self, tokens):
        return bytes(tokens).decode('utf-8')
    
def tokenizer(ikey, okey, length):
    """Tokenizes a transcript"""
    tok = CharTokenizer()
    def _tokenizer(samples):
        for s in samples:
            toks = torch.tensor(tok.encode(s[ikey]))
            s[okey] = F.pad(toks, (0, length - toks.shape[-1]), value=tok.eot)
            yield s
    return _tokenizer

def ar_padder(ikey, okey, length, pad_token):
    """Pads the tokens for autoregresive training"""
    def _ar_padder(samples):
        for s in samples:
            toks = s[ikey]
            if isinstance(toks, (list, np.ndarray)): toks = torch.tensor(toks)
            toks = toks.to(torch.long)
            s['in_' +okey] = F.pad(toks, (1, length - toks.shape[-1] - 1), value=pad_token)
            s['out_'+okey] = F.pad(toks, (0, length - toks.shape[-1]), value=pad_token)
            yield s
    return _ar_padder

def char_per_seconder(txt_key, stoks_key, cps_key, stoks_per_second=25):
    """Adds the characters per second metric to the input data"""
    def _char_per_seconder(samples):
        for s in samples:
            secs = s[stoks_key].shape[-1] / stoks_per_second
            s[cps_key] = len(s[txt_key]) / secs
            yield s
    return _char_per_seconder

# %% ../nbs/5B. Text to semantic token modeling.ipynb 9
def build_speaker_map(shards):
    speakers = set()
    for shard in shards:
        with open(shard+'.speakers.txt') as f: speakers = speakers.union(set(x.strip() for x in f.readlines()))
    return {id:i for i,id in enumerate(speakers)}

def speaker_id_extractor(speaker_map):
    def _extractor(samples):
        for s in samples:
            s['speaker'] = torch.tensor(speaker_map[s['__key__'].split("/")[1]])
            yield s
    return _extractor

# %% ../nbs/5B. Text to semantic token modeling.ipynb 10
def load_datasets(
    input:str,             # webdataset folder or shard list
    samples:int,           # samples per epoch
    subsample:float=1,     # use a fraction of the files
    val_samples:int=512,
    vq_codes:int=4096,
):
    if isinstance(input, (Path, str)):
        path = Path(input)
        if path.is_dir():
            glob = '*-t2s-*.tar.gz'
        else:
            glob = path.name
            path = path.parent
        input = Path(path).glob(glob)
    elif isinstance(input, list):
        pass
    else:
        raise ArgumentError("input should be either a list of a path with an optional glob specifier")
    shards = [str(x) for x in input]

    speaker_map = build_speaker_map(shards)
    
    def ds(shards, length):
        ds = wds.WebDataset(wds.ResampledShards(shards)).compose(
            wds.decode(),
            speaker_id_extractor(speaker_map),
            wds.select(lambda s: s['stoks.npy'].shape[-1] > 12), # select samples > .5s
            tokenizer('txt', 'ttoks', length=550),
            ar_padder('stoks.npy', 'stoks', length=750, pad_token=vq_codes-1),
            char_per_seconder('txt', 'stoks.npy', 'cps', stoks_per_second=25),
            wds.to_tuple('ttoks', 'speaker', 'cps', 'in_stoks', 'out_stoks'),
            wds.batched(64)
        )
        ds.speakers = speaker_map
        ds.total_samples = length
        ds.stoks_len = 750
        ds.stoks_codes = vq_codes
        ds.ttoks_len = 550
        return ds.compose(wds.slice(length // 64)).with_epoch(length // 64).with_length(length // 64)
    
    return (
        ds(shards[1:], samples),
        ds(shards[:1], val_samples),
    )

# %% ../nbs/5B. Text to semantic token modeling.ipynb 14
def rand(start, end):
    return random.random() * (end - start) + start

@dataclasses.dataclass
class Tunables:
    init_std :float = 1
    embeddings_std :float = .01
    embeddings_lr_scale: float = 5
    embedding_projector_lr_scale: float = 2.5
    output_mult :float = .35
    query_mult :float = 1
    encoder_depth_ratio :float = 0.25
    eot_dropout_p :float = .5
    cps_input: bool = True
    cps_bins: int = 32
        
    lr0 :float = 1.5e-3
    clip_gradient_norm :float = .2
    weight_decay :float = 1e-1
    warmup_steps :float = 4000

    random :bool = False

    def __post_init__(self):
        # randomize the hyperparams if requested
        if self.random:
            self.init_std = 10**rand(-1,1)
            self.embeddings_std = 10**rand(-3,-.7)
            self.embeddings_lr_scale = rand(2,6)
            self.output_mult = rand(0.25,0.65)
            self.query_mult = 2**rand(-2,3)
            self.encoder_depth_ratio = 0.25
            
            self.lr0 = rand(1,5)*1e-3
            self.clip_gradient_norm = 10**rand(-3,0)
            self.warmup_steps = 100*(10**rand(1,1.85))

# %% ../nbs/5B. Text to semantic token modeling.ipynb 15
class EmbeddingProjector(nn.Linear):
    pass

# %% ../nbs/5B. Text to semantic token modeling.ipynb 16
class Encoder(nn.Module):
    def __init__(self, depth=6, width=384, n_head=6, length=1500, codes=1024, emb_width=384, ffn_mult=4, pos_embs=None, tunables=Tunables()):
        super().__init__()
        self.emb_width = emb_width
        
        self.emb_factor = width != emb_width
    
        self.embedding = nn.Embedding(codes, emb_width)
        if self.emb_factor:
            self.emb_to_hidden = EmbeddingProjector(emb_width, width)

        if pos_embs is None: pos_embs = sinusoids(length, width)
        self.register_buffer("positional_embedding", pos_embs)

        self.layers = nn.Sequential(*[
            ResidualAttentionBlock(width, n_head,
                                   qk_scale=tunables.query_mult*8/math.sqrt(width/n_head), ffn_mult=ffn_mult) for _ in range(depth)
        ])

        self.ln_post = LayerNorm(width)
        
    def forward(self, Stoks):
        xin = self.embedding(Stoks)
        if self.emb_factor:
            xin = self.emb_to_hidden(xin)
        
        assert xin.shape[1:] == self.positional_embedding.shape, "incorrect semantic token shape"
        xin = (xin + self.positional_embedding).to(xin.dtype)

        return self.ln_post(self.layers(xin))

# %% ../nbs/5B. Text to semantic token modeling.ipynb 17
class Decoder(nn.Module):
    def __init__(self, depth=6, stoks_width=384, width=384, n_head=6, length=1500, codes=1024, ffn_mult=4, pos_embs=None, tunables=Tunables()):
        super().__init__()
        self.length = length
        self.codes = codes
        self.width = width
        self.stoks_width = stoks_width
        
        self.emb_factor = width != stoks_width
    
        # embed semantic tokens
        self.embedding = nn.Embedding(codes, stoks_width)
        if self.emb_factor:
            self.emb_to_hidden = EmbeddingProjector(stoks_width, width)
            self.hidden_to_emb = EmbeddingProjector(width, stoks_width)

        if pos_embs is None: pos_embs = sinusoids(length, width)
        self.register_buffer("positional_embedding", pos_embs)
        
        self.layers = nn.ModuleList([
            ResidualAttentionBlock(width, n_head, cross_attention=True,
                                   qk_scale=tunables.query_mult*8/math.sqrt(width/n_head), ffn_mult=ffn_mult) for _ in range(depth)
        ])
        self.ln_post = LayerNorm(width)
        
    def forward(self, Stoks, xenc, cps=None):
        Sembs = self.embedding(Stoks)
            
        if self.emb_factor:
            Sembs = self.emb_to_hidden(Sembs)
    
        xin = (Sembs + self.positional_embedding[:Sembs.shape[1]]).to(xenc.dtype)
        if cps is not None: xin = xin + cps
    
        x = xin
        for l in self.layers: x = l(x, xenc, causal=True)
        
        x = self.ln_post(x)
    
        if self.emb_factor:
            x = self.hidden_to_emb(x)
        
        logits = (x @ self.embedding.weight.to(x.dtype).T).float()
        return logits

# %% ../nbs/5B. Text to semantic token modeling.ipynb 18
class TSARTransformer(nn.Module):
    def __init__(self, depth=6, n_head=6, head_width=64, ffn_mult=4, language='en',
                 ttoks_len=200, ttoks_codes=50364, ttoks_width=None,
                 stoks_len=1500, stoks_codes=1024, stoks_width=None,
                 tunables=Tunables()):
        assert language == 'en', "only english is supported right now"
        super().__init__()
        store_attr("depth,n_head,head_width,ffn_mult,stoks_width,ttoks_width,ttoks_len,stoks_len,ttoks_codes,stoks_codes,language")

        width = n_head * head_width
        self.width = width
        self.base_width = 3 * head_width
        self.tunables = tunables
        if self.stoks_width is None: self.stoks_width = self.width
        if self.ttoks_width is None: self.ttoks_width = self.width
        
        if tunables.cps_input:
            self.cps_embeddings = nn.Embedding(tunables.cps_bins, self.width)
        else:
            self.cps_embeddings = None
        
        encoder_depth = int(depth * 2 * tunables.encoder_depth_ratio)
        decoder_depth = depth * 2 - encoder_depth
        tformer_args = dict(width=width, n_head=n_head, ffn_mult=ffn_mult, tunables=tunables)
        self.encoder = Encoder(length=ttoks_len, codes=ttoks_codes, emb_width=self.ttoks_width, depth=encoder_depth, **tformer_args)
        self.decoder = Decoder(length=stoks_len, codes=stoks_codes, stoks_width=self.stoks_width, depth=decoder_depth, **tformer_args)
        
        self.tokenizer = None
        
        self.apply(self.init_transformer)

    def load_frozen_semantic_embeddings(self, vqmodel):
        with torch.no_grad():
            self.decoder.embedding.weight[:] = vqmodel.rq.layers[0]._codebook.embed[0]
            self.decoder.embedding.lr_scale = 0

    def setup(self, device):
        pass

    def init_transformer(self, m):
        if isinstance(m, LinearHead):
            m.no_weight_decay = True
            torch.nn.init.constant_(m.weight, 0)
        elif isinstance(m, QueryHead):
            m.lr_scale = 1/(m.weight.shape[1] / self.base_width)
            torch.nn.init.constant_(m.weight, 0)
        elif isinstance(m, nn.Embedding):
            m.no_weight_decay = True
            m.lr_scale = self.tunables.embeddings_lr_scale
            std = self.tunables.embeddings_std
            torch.nn.init.trunc_normal_(m.weight, std=std, a=-3*std, b=3*std)
        elif isinstance(m, EmbeddingProjector):
            m.lr_scale = self.tunables.embedding_projector_lr_scale
            std = self.tunables.init_std
            torch.nn.init.trunc_normal_(m.weight, std=std, a=-3*std, b=3*std)
        elif isinstance(m, nn.Linear):
            m.lr_scale = 1/(m.weight.shape[1] / self.base_width)
            std = self.tunables.init_std / m.weight.shape[1]
            torch.nn.init.trunc_normal_(m.weight, std=std, a=-3*std, b=3*std)
            if m.bias is not None:
                torch.nn.init.trunc_normal_(m.bias, std=std, a=-3*std, b=3*std)
        elif isinstance(m, nn.LayerNorm):
            m.no_weight_decay = True
            torch.nn.init.constant_(m.bias, 0)
            torch.nn.init.constant_(m.weight, 1)
            
    def forward(self, Ttoks, speakers, cpss, in_stoks, out_stoks=None, loss=True):
        with record_function("encoder"):
            xenc = self.encoder(Ttoks.to(torch.long))
        with record_function("decoder"):
            if self.cps_embeddings:
                cps_bin = (cpss / 20 * self.tunables.cps_bins).to(torch.long)
                cps_bin[cps_bin >= self.tunables.cps_bins] = self.tunables.cps_bins-1
                cps_embs = self.cps_embeddings(cps_bin).unsqueeze(1)
            else:
                cps_embs = None
            logits = self.decoder(in_stoks, xenc, cps=cps_embs) * self.tunables.output_mult / (self.width / self.base_width)
        if loss is not None:
            with record_function("loss"):
                loss = F.cross_entropy(logits.transpose(-1,-2), out_stoks)#, reduction='none')
        return logits, loss

    #
    # inference
    #
    @classmethod
    def load_model(cls, repo_id="collabora/whisperspeech", filename="t2s_up_wds.model", local_filename=None):
        if not local_filename:
            local_filename = hf_hub_download(repo_id=repo_id, filename=filename)
        spec = torch.load(local_filename)
        model = cls(**spec['config'], tunables=Tunables(**spec['tunables']))
        model.load_state_dict(spec['state_dict'])
        model.eval()
        return model

    def load_checkpoint(self, local_filename):
        spec = torch.load(local_filename, map_location='cpu')
        assert 'pytorch-lightning_version' in spec, 'not a valid PyTorch Lightning checkpoint'
        state_dict = {k.replace('model.', ''):v
                      for k,v in spec['state_dict'].items()}
        self.load_state_dict(state_dict)
        return self

    def save_model(self, fname):
        torch.save(dict(config = self.__stored_args__,
                        tunables = dataclasses.asdict(self.tunables),
                        state_dict = self.state_dict()), fname)

    def ensure_tokenizer(self):
        assert not self.training
        if self.tokenizer is None: self.tokenizer = CharTokenizer()
        #whisper.tokenizer.get_tokenizer(multilingual=True)  

    @property
    def device(self):
        return next(self.parameters()).device
    
    @torch.no_grad()
    def generate(self, txt, cps=15, N=None, T=0.7, top_k=None, show_progress_bar=True):
        self.ensure_tokenizer()
        N = N or self.stoks_len
        dev = self.device
        ttoks = torch.tensor(self.tokenizer.encode(txt), device=dev)
        ttoks = F.pad(ttoks, (0, self.ttoks_len - len(ttoks)), value=self.tokenizer.eot).unsqueeze(0)
        cpss = torch.tensor([cps], device=dev)
        toks = torch.zeros((1,N), dtype=torch.long, device=dev)
        toks[0,0] = self.stoks_codes-1
        it = range(1,N)
        if show_progress_bar: it = progress_bar(it)
        for i in it:
            p, _ = self(ttoks, None, cpss, toks[:,:i], loss=None)
            last_p = p[0,-1]
            if top_k:
                last_p[last_p < torch.topk(last_p, top_k).values[-1,None]] = -torch.inf
            tok = torch.multinomial((last_p / float(T)).softmax(-1), 1)
            toks[0,i] = tok
            if toks[0,i] == self.stoks_codes-1: return toks[0,1:i]
        return toks[0,1:]
    
    @torch.no_grad()
    def generate_batch(self, txts, N=None, T=1.1, top_k=7, show_progress_bar=True):
        self.ensure_tokenizer()
        N = self.stoks_len
        dev = self.device
        ttoks = []
        for txt in txts:
            ttoks_ = torch.tensor(self.tokenizer.encode(txt), device=dev)
            ttoks_ = F.pad(ttoks_, (0, self.ttoks_len - len(ttoks_)), value=self.tokenizer.eot).unsqueeze(0)
            ttoks.append(ttoks_)
        ttoks = torch.cat(ttoks, dim=0)
        toks = torch.zeros((len(ttoks),N), dtype=torch.long, device=dev)
        it = range(N)
        if show_progress_bar: it = progress_bar(it)
        for i in it:
            p, _ = self(ttoks, toks[:,:i], loss=None)
            last_p = p[:,-1]
            if top_k:
                last_p[last_p < torch.topk(last_p, top_k).values[:,-1,None]] = -torch.inf
            tok = torch.multinomial((last_p / float(T)).softmax(-1), 1)
            toks[:,i] = tok[:,0]
            if (toks[:,i] == self.stoks_codes-1).all(): return toks[:,:i]
        return toks

# %% ../nbs/5B. Text to semantic token modeling.ipynb 19
def _make_model(size:str, tunables:Tunables=Tunables(), dataset=None, **kwargs):
    kwargs = dict(stoks_len = dataset.stoks_len, ttoks_len = dataset.ttoks_len, tunables=tunables, **kwargs)
    if 'stoks_codes' not in kwargs: kwargs['stoks_codes'] = dataset.stoks_codes
    if size == 'micro':
        return TSARTransformer(depth=2, n_head=3, ffn_mult=1, **kwargs)
    if size == 'tiny':
        return TSARTransformer(depth=4, n_head=6, **kwargs)
    if size == 'base':
        return TSARTransformer(depth=6, n_head=8, **kwargs)
    if size == 'small':
        return TSARTransformer(depth=12, n_head=16, **kwargs)

def make_model(size:str, frozen_embeddings_model:str=None, tunables:Tunables=Tunables(), dataset:torch.utils.data.Dataset=None):
    if frozen_embeddings_model:
        vqmodel = vq_stoks.RQBottleneckTransformer.load_model(frozen_embeddings_model)
        model = _make_model(size, tunables, dataset, stoks_codes=vqmodel.vq_codes+1, stoks_width=vqmodel.rq.layers[0]._codebook.embed[0].shape[-1])
        model.load_frozen_semantic_embeddings(vqmodel)
    else:
        model = _make_model(size, quantizers, tunables, dataset)
    return model