From 38d8b17d1864a0b5c9d2fe33d43688d4292cc856 Mon Sep 17 00:00:00 2001 From: James Betker Date: Sat, 4 Jun 2022 23:09:54 -0600 Subject: [PATCH] tfd8 gets real verbose grad norm metrics --- .../audio/music/transformer_diffusion8.py | 51 ++++++++----------- 1 file changed, 22 insertions(+), 29 deletions(-) diff --git a/codes/models/audio/music/transformer_diffusion8.py b/codes/models/audio/music/transformer_diffusion8.py index e6189b60..9ffd8440 100644 --- a/codes/models/audio/music/transformer_diffusion8.py +++ b/codes/models/audio/music/transformer_diffusion8.py @@ -89,23 +89,9 @@ class TransformerDiffusion(nn.Module): self.time_embed = nn.Sequential( linear(prenet_channels, prenet_channels), nn.SiLU(), - linear(prenet_channels, prenet_channels), + linear(prenet_channels, model_channels), ) prenet_heads = prenet_channels//64 - self.conditioning_embedder = nn.Sequential(nn.Conv1d(in_channels, prenet_channels // 2, 3, padding=1, stride=2), - nn.Conv1d(prenet_channels//2, prenet_channels,3,padding=1,stride=2)) - self.conditioning_encoder = Encoder( - dim=prenet_channels, - depth=4, - heads=prenet_heads, - ff_dropout=dropout, - attn_dropout=dropout, - use_rmsnorm=True, - ff_glu=True, - rotary_pos_emb=True, - zero_init_branch_output=True, - ff_mult=1, - ) self.input_converter = nn.Linear(input_vec_dim, prenet_channels) self.code_converter = Encoder( @@ -123,7 +109,6 @@ class TransformerDiffusion(nn.Module): self.unconditioned_embedding = nn.Parameter(torch.randn(1,1,prenet_channels)) self.rotary_embeddings = RotaryEmbedding(rotary_emb_dim) - self.cond_intg = nn.Linear(prenet_channels*2, model_channels) self.intg = nn.Linear(prenet_channels*2, model_channels) self.layers = TimestepRotaryEmbedSequential(*[DietAttentionBlock(model_channels, block_channels, block_channels // 64, dropout) for _ in range(num_layers)]) @@ -137,16 +122,13 @@ class TransformerDiffusion(nn.Module): def get_grad_norm_parameter_groups(self): groups = { - 'contextual_embedder': list(self.conditioning_embedder.parameters()), 'layers': list(self.layers.parameters()) + list(self.inp_block.parameters()), 'code_converters': list(self.input_converter.parameters()) + list(self.code_converter.parameters()), 'time_embed': list(self.time_embed.parameters()), } return groups - def timestep_independent(self, codes, conditioning_input, expected_seq_len): - cond_emb = self.conditioning_embedder(conditioning_input).permute(0,2,1) - cond_emb = self.conditioning_encoder(cond_emb)[:, 0] + def timestep_independent(self, codes, expected_seq_len): code_emb = self.input_converter(codes) # Mask out the conditioning branch for whole batch elements, implementing something similar to classifier-free guidance. @@ -158,29 +140,24 @@ class TransformerDiffusion(nn.Module): code_emb = self.code_converter(code_emb) expanded_code_emb = F.interpolate(code_emb.permute(0,2,1), size=expected_seq_len, mode='nearest').permute(0,2,1) - return expanded_code_emb, cond_emb + return expanded_code_emb - def forward(self, x, timesteps, codes=None, conditioning_input=None, precomputed_code_embeddings=None, - precomputed_cond_embeddings=None, conditioning_free=False): + def forward(self, x, timesteps, codes=None, conditioning_input=None, precomputed_code_embeddings=None, conditioning_free=False): if precomputed_code_embeddings is not None: assert codes is None and conditioning_input is None, "Do not provide precomputed embeddings and the other parameters. It is unclear what you want me to do here." unused_params = [] if conditioning_free: code_emb = self.unconditioned_embedding.repeat(x.shape[0], x.shape[-1], 1) - cond_emb = self.conditioning_embedder(conditioning_input).permute(0,2,1) - cond_emb = self.conditioning_encoder(cond_emb)[:, 0] unused_params.extend(list(self.code_converter.parameters())) else: if precomputed_code_embeddings is not None: code_emb = precomputed_code_embeddings - cond_emb = precomputed_cond_embeddings else: - code_emb, cond_emb = self.timestep_independent(codes, conditioning_input, x.shape[-1]) + code_emb = self.timestep_independent(codes, x.shape[-1]) unused_params.append(self.unconditioned_embedding) - blk_emb = torch.cat([self.time_embed(timestep_embedding(timesteps, self.prenet_channels)), cond_emb], dim=-1) - blk_emb = self.cond_intg(blk_emb) + blk_emb = self.time_embed(timestep_embedding(timesteps, self.prenet_channels)) x = self.inp_block(x).permute(0,2,1) rotary_pos_emb = self.rotary_embeddings(x.shape[1], x.device) @@ -245,6 +222,21 @@ class TransformerDiffusionWithQuantizer(nn.Module): else: return {} + def get_grad_norm_parameter_groups(self): + groups = { + 'attention_layers': [lyr.attn for lyr in self.diff.layers], + 'ff_layers': [lyr.ff for lyr in self.diff.layers], + 'quantizer_encoder': list(self.quantizer.encoder.parameters()), + 'quant_codebook': [self.quantizer.quantizer.codevectors], + 'rotary_embeddings': list(self.diff.rotary_embeddings.parameters()), + 'out': list(self.diff.out.parameters()), + 'x_proj': list(self.diff.inp_block.parameters()), + 'layers': list(self.diff.layers.parameters()), + 'code_converters': list(self.diff.input_converter.parameters()) + list(self.diff.code_converter.parameters()), + 'time_embed': list(self.diff.time_embed.parameters()), + } + return groups + @register_model def register_transformer_diffusion8(opt_net, opt): @@ -274,6 +266,7 @@ if __name__ == '__main__': cond = torch.randn(2, 256, 400) ts = torch.LongTensor([600, 600]) model = TransformerDiffusionWithQuantizer(model_channels=2048, block_channels=1024, prenet_channels=1024, input_vec_dim=1024, num_layers=16, prenet_layers=6) + model.get_grad_norm_parameter_groups() #quant_weights = torch.load('D:\\dlas\\experiments\\train_music_quant\\models\\18000_generator_ema.pth') #diff_weights = torch.load('X:\\dlas\\experiments\\train_music_diffusion_tfd5\\models\\48000_generator_ema.pth')