deviandice/DL-Art-School
1
0
Fork 0
forked from mrq/DL-Art-School
Code Issues Pull Requests Packages Projects Releases Wiki Activity

iq checkin

Browse Source
This commit is contained in:
James Betker 2022-07-18 18:40:14 -06:00
parent df27b98730
commit 0824708dc7
1 changed files with 15 additions and 9 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

24
codes/models/audio/music/instrument_quantizer.py
View File

@ -19,14 +19,22 @@ class SelfClassifyingHead(nn.Module):
self.temperature = init_temperature self.temperature = init_temperature
self.dec = Decoder(dim=dim, depth=head_depth, heads=4, ff_dropout=dropout, ff_mult=2, attn_dropout=dropout, self.dec = Decoder(dim=dim, depth=head_depth, heads=4, ff_dropout=dropout, ff_mult=2, attn_dropout=dropout,
use_rmsnorm=True, ff_glu=True, do_checkpointing=False) use_rmsnorm=True, ff_glu=True, do_checkpointing=False)
self.quantizer = VectorQuantize(dim, classes, codebook_dim=32, use_cosine_sim=True, threshold_ema_dead_code=2, self.quantizer = VectorQuantize(out_dim, classes, codebook_dim=16, use_cosine_sim=False, threshold_ema_dead_code=2,
sample_codebook_temp=init_temperature) sample_codebook_temp=init_temperature)
self.to_output = nn.Linear(dim, out_dim) self.to_output = nn.Linear(dim, out_dim)
self.to_decoder = nn.Linear(out_dim, dim)
self.scale = nn.Linear(dim, 1, bias=False)
self.scale.weight.data.zero_()
def do_ar_step(self, x, used_codes): def do_ar_step(self, x, used_codes):
MIN = -12
h = self.dec(x) h = self.dec(x)
h, c, _ = self.quantizer(h[:, -1], used_codes) o = self.to_output(h[:, -1])
return h, c scale = (self.scale(h[:, -1]) + 1)
q, c, _ = self.quantizer(o, used_codes)
q = F.relu(q * scale) + MIN
return q, c
def forward(self, x): def forward(self, x):
with torch.no_grad(): with torch.no_grad():
@ -39,12 +47,12 @@ class SelfClassifyingHead(nn.Module):
results = [] results = []
codes = [] codes = []
for i in range(self.seq_len): for i in range(self.seq_len):
h, c = checkpoint(functools.partial(self.do_ar_step, used_codes=codes), torch.stack(stack, dim=1)) q, c = checkpoint(functools.partial(self.do_ar_step, used_codes=codes), torch.stack(stack, dim=1))
c_mask = c c_mask = c
c_mask[c==0] = -1 # Mask this out because we want code=0 to be capable of being repeated. c_mask[c==0] = -1 # Mask this out because we want code=0 to be capable of being repeated.
codes.append(c) codes.append(c_mask)
stack.append(h.detach()) # Detach here to avoid piling up gradients from autoregression. We really just want the gradients to flow to the selected class embeddings and the selector for those classes. stack.append(self.to_decoder(q))
outputs.append(self.to_output(h)) outputs.append(q)
results.append(torch.stack(outputs, dim=1).sum(1)) results.append(torch.stack(outputs, dim=1).sum(1))
return results, torch.cat(codes, dim=0) return results, torch.cat(codes, dim=0)
@ -81,7 +89,6 @@ class InstrumentQuantizer(nn.Module):
self.op_dim = op_dim self.op_dim = op_dim
self.proj = nn.Linear(op_dim, dim) self.proj = nn.Linear(op_dim, dim)
self.encoder = nn.ModuleList([VectorResBlock(dim, dropout) for _ in range(enc_depth)]) self.encoder = nn.ModuleList([VectorResBlock(dim, dropout) for _ in range(enc_depth)])
self.final_bn = nn.BatchNorm1d(dim)
self.heads = SelfClassifyingHead(dim, num_classes, op_dim, head_depth, class_seq_len, dropout, max_temp) self.heads = SelfClassifyingHead(dim, num_classes, op_dim, head_depth, class_seq_len, dropout, max_temp)
self.min_gumbel_temperature = min_temp self.min_gumbel_temperature = min_temp
self.max_gumbel_temperature = max_temp self.max_gumbel_temperature = max_temp
@ -99,7 +106,6 @@ class InstrumentQuantizer(nn.Module):
h = self.proj(f) h = self.proj(f)
for lyr in self.encoder: for lyr in self.encoder:
h = lyr(h) h = lyr(h)
h = self.final_bn(h.unsqueeze(-1)).squeeze(-1)
reconstructions, codes = self.heads(h) reconstructions, codes = self.heads(h)
reconstruction_losses = torch.stack([F.mse_loss(r.reshape(b, s, c), px) for r in reconstructions]) reconstruction_losses = torch.stack([F.mse_loss(r.reshape(b, s, c), px) for r in reconstructions])