libcom 0.1.8__py3-none-any.whl

libcom/reflection_generation/source/cldm/cldm.py

@@ -0,0 +1,509 @@
+import einops
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+import os
+import pickle
+import cv2
+import numpy as np
+from libcom.reflection_generation.source.ldm.modules.diffusionmodules.util import (
+    conv_nd,
+    linear,
+    zero_module,
+    timestep_embedding,
+)
+
+from einops import rearrange, repeat
+from torchvision.utils import make_grid
+from libcom.reflection_generation.source.ldm.modules.attention import SpatialTransformer
+from libcom.reflection_generation.source.ldm.modules.diffusionmodules.openaimodel import UNetModel, TimestepEmbedSequential, ResBlock, Downsample, AttentionBlock
+from libcom.reflection_generation.source.ldm.models.diffusion.ddpm import LatentDiffusion
+from libcom.reflection_generation.source.ldm.util import log_txt_as_img, exists, instantiate_from_config
+from libcom.reflection_generation.source.ldm.models.diffusion.ddim import DDIMSampler
+
+from .base_network import MaskCls, RegNetwork
+from ..cluster_feature import generate_shadow_clip_features
+    
+class ControlledUnetModel(UNetModel):
+
+    def forward(self, x, timesteps=None, context=None, control=None, only_mid_control=False, get_stable_features=False, remask=None, **kwargs):
+        if remask is not None:
+            return self.out(remask)
+        hs = []
+        with torch.no_grad():
+            t_emb = timestep_embedding(timesteps, self.model_channels, repeat_only=False)
+            emb = self.time_embed(t_emb)
+            h = x.type(self.dtype)
+            for module in self.input_blocks:
+                h = module(h, emb, context)
+                hs.append(h)
+            h = self.middle_block(h, emb, context)
+        encoder_features = hs.copy()
+        unet_features = h.clone()
+        decoder_features = []
+        if control is not None:
+            h += control.pop()
+
+        for i, module in enumerate(self.output_blocks):
+            if only_mid_control or control is None:
+                h = torch.cat([h, hs.pop()], dim=1)
+            else:
+                h = torch.cat([h, hs.pop() + control.pop()], dim=1)
+            h = module(h, emb, context)
+            decoder_features.append(h)
+        h = h.type(x.dtype)
+        if get_stable_features:
+            return self.out(h), [unet_features, encoder_features, decoder_features]
+        return self.out(h)
+
+
+class ControlNet(nn.Module):
+    def __init__(
+            self,
+            image_size,
+            in_channels,
+            model_channels,
+            hint_channels,
+            num_res_blocks,
+            attention_resolutions,
+            dropout=0,
+            channel_mult=(1, 2, 4, 8),
+            conv_resample=True,
+            dims=2,
+            use_checkpoint=False,
+            use_fp16=False,
+            num_heads=-1,
+            num_head_channels=-1,
+            num_heads_upsample=-1,
+            use_scale_shift_norm=False,
+            resblock_updown=False,
+            use_new_attention_order=False,
+            use_spatial_transformer=False,  # custom transformer support
+            transformer_depth=1,  # custom transformer support
+            context_dim=None,  # custom transformer support
+            n_embed=None,  # custom support for prediction of discrete ids into codebook of first stage vq model
+            legacy=True,
+            disable_self_attentions=None,
+            num_attention_blocks=None,
+            disable_middle_self_attn=False,
+            use_linear_in_transformer=False,
+    ):
+        super().__init__()
+        if use_spatial_transformer:
+            assert context_dim is not None, 'Fool!! You forgot to include the dimension of your cross-attention conditioning...'
+
+        if context_dim is not None:
+            assert use_spatial_transformer, 'Fool!! You forgot to use the spatial transformer for your cross-attention conditioning...'
+            from omegaconf.listconfig import ListConfig
+            if type(context_dim) == ListConfig:
+                context_dim = list(context_dim)
+
+        if num_heads_upsample == -1:
+            num_heads_upsample = num_heads
+
+        if num_heads == -1:
+            assert num_head_channels != -1, 'Either num_heads or num_head_channels has to be set'
+
+        if num_head_channels == -1:
+            assert num_heads != -1, 'Either num_heads or num_head_channels has to be set'
+
+        self.dims = dims
+        self.image_size = image_size
+        self.in_channels = in_channels
+        self.model_channels = model_channels
+        if isinstance(num_res_blocks, int):
+            self.num_res_blocks = len(channel_mult) * [num_res_blocks]
+        else:
+            if len(num_res_blocks) != len(channel_mult):
+                raise ValueError("provide num_res_blocks either as an int (globally constant) or "
+                                 "as a list/tuple (per-level) with the same length as channel_mult")
+            self.num_res_blocks = num_res_blocks
+        if disable_self_attentions is not None:
+            # should be a list of booleans, indicating whether to disable self-attention in TransformerBlocks or not
+            assert len(disable_self_attentions) == len(channel_mult)
+        if num_attention_blocks is not None:
+            assert len(num_attention_blocks) == len(self.num_res_blocks)
+            assert all(map(lambda i: self.num_res_blocks[i] >= num_attention_blocks[i], range(len(num_attention_blocks))))
+            print(f"Constructor of UNetModel received num_attention_blocks={num_attention_blocks}. "
+                  f"This option has LESS priority than attention_resolutions {attention_resolutions}, "
+                  f"i.e., in cases where num_attention_blocks[i] > 0 but 2**i not in attention_resolutions, "
+                  f"attention will still not be set.")
+
+        self.attention_resolutions = attention_resolutions
+        self.dropout = dropout
+        self.channel_mult = channel_mult
+        self.conv_resample = conv_resample
+        self.use_checkpoint = use_checkpoint
+        self.dtype = torch.float16 if use_fp16 else torch.float32
+        self.num_heads = num_heads
+        self.num_head_channels = num_head_channels
+        self.num_heads_upsample = num_heads_upsample
+        self.predict_codebook_ids = n_embed is not None
+        self.add_color_channel = 0
+
+        time_embed_dim = model_channels * 4
+        self.time_embed = nn.Sequential(
+            linear(model_channels, time_embed_dim),
+            nn.SiLU(),
+            linear(time_embed_dim, time_embed_dim),
+        )
+
+        self.input_blocks = nn.ModuleList(
+            [
+                TimestepEmbedSequential(
+                    conv_nd(dims, in_channels, model_channels, 3, padding=1)
+                )
+            ]
+        )
+        self.zero_convs = nn.ModuleList([self.make_zero_conv(model_channels, self.add_color_channel)])
+
+        self.input_hint_block = TimestepEmbedSequential(
+            conv_nd(dims, 5, 16, 3, padding=1), #TODO
+            nn.SiLU(),
+            conv_nd(dims, 16, 16, 3, padding=1),
+            nn.SiLU(),
+            conv_nd(dims, 16, 32, 3, padding=1, stride=2),
+            nn.SiLU(),
+            conv_nd(dims, 32, 32, 3, padding=1),
+            nn.SiLU(),
+            conv_nd(dims, 32, 96, 3, padding=1, stride=2),
+            nn.SiLU(),
+            conv_nd(dims, 96, 96, 3, padding=1),
+            nn.SiLU(),
+            conv_nd(dims, 96, 256, 3, padding=1, stride=2),
+            nn.SiLU(),
+            zero_module(conv_nd(dims, 256, model_channels, 3, padding=1))
+        )
+
+        self._feature_size = model_channels
+        input_block_chans = [model_channels]
+        ch = model_channels
+        ds = 1
+        for level, mult in enumerate(channel_mult):
+            for nr in range(self.num_res_blocks[level]):
+                layers = [
+                    ResBlock(
+                        ch,
+                        time_embed_dim,
+                        dropout,
+                        out_channels=mult * model_channels,
+                        dims=dims,
+                        use_checkpoint=use_checkpoint,
+                        use_scale_shift_norm=use_scale_shift_norm,
+                    )
+                ]
+                ch = mult * model_channels
+                if ds in attention_resolutions:
+                    if num_head_channels == -1:
+                        dim_head = ch // num_heads
+                    else:
+                        num_heads = ch // num_head_channels
+                        dim_head = num_head_channels
+                    if legacy:
+                        # num_heads = 1
+                        dim_head = ch // num_heads if use_spatial_transformer else num_head_channels
+                    if exists(disable_self_attentions):
+                        disabled_sa = disable_self_attentions[level]
+                    else:
+                        disabled_sa = False
+
+                    if not exists(num_attention_blocks) or nr < num_attention_blocks[level]:
+                        layers.append(
+                            AttentionBlock(
+                                ch,
+                                use_checkpoint=use_checkpoint,
+                                num_heads=num_heads,
+                                num_head_channels=dim_head,
+                                use_new_attention_order=use_new_attention_order,
+                            ) if not use_spatial_transformer else SpatialTransformer(
+                                ch, num_heads, dim_head, depth=transformer_depth, context_dim=context_dim,
+                                disable_self_attn=disabled_sa, use_linear=use_linear_in_transformer,
+                                use_checkpoint=use_checkpoint
+                            )
+                        )
+                self.input_blocks.append(TimestepEmbedSequential(*layers))
+                self.zero_convs.append(self.make_zero_conv(ch, self.add_color_channel))
+                self._feature_size += ch
+                input_block_chans.append(ch)
+            if level != len(channel_mult) - 1:
+                out_ch = ch
+                self.input_blocks.append(
+                    TimestepEmbedSequential(
+                        ResBlock(
+                            ch,
+                            time_embed_dim,
+                            dropout,
+                            out_channels=out_ch,
+                            dims=dims,
+                            use_checkpoint=use_checkpoint,
+                            use_scale_shift_norm=use_scale_shift_norm,
+                            down=True,
+                        )
+                        if resblock_updown
+                        else Downsample(
+                            ch, conv_resample, dims=dims, out_channels=out_ch
+                        )
+                    )
+                )
+                ch = out_ch
+                input_block_chans.append(ch)
+                self.zero_convs.append(self.make_zero_conv(ch, self.add_color_channel))
+                ds *= 2
+                self._feature_size += ch
+
+        if num_head_channels == -1:
+            dim_head = ch // num_heads
+        else:
+            num_heads = ch // num_head_channels
+            dim_head = num_head_channels
+        if legacy:
+            # num_heads = 1
+            dim_head = ch // num_heads if use_spatial_transformer else num_head_channels
+        self.middle_block = TimestepEmbedSequential(
+            ResBlock(
+                ch,
+                time_embed_dim,
+                dropout,
+                dims=dims,
+                use_checkpoint=use_checkpoint,
+                use_scale_shift_norm=use_scale_shift_norm,
+            ),
+            AttentionBlock(
+                ch,
+                use_checkpoint=use_checkpoint,
+                num_heads=num_heads,
+                num_head_channels=dim_head,
+                use_new_attention_order=use_new_attention_order,
+            ) if not use_spatial_transformer else SpatialTransformer(  # always uses a self-attn
+                ch, num_heads, dim_head, depth=transformer_depth, context_dim=context_dim,
+                disable_self_attn=disable_middle_self_attn, use_linear=use_linear_in_transformer,
+                use_checkpoint=use_checkpoint
+            ),
+            ResBlock(
+                ch,
+                time_embed_dim,
+                dropout,
+                dims=dims,
+                use_checkpoint=use_checkpoint,
+                use_scale_shift_norm=use_scale_shift_norm,
+            ),
+        )
+        self.middle_block_out = self.make_zero_conv(ch)
+        self._feature_size += ch
+
+    def make_zero_conv(self, channels, color_hist_channels=0):
+        return TimestepEmbedSequential(zero_module(conv_nd(self.dims, channels + color_hist_channels, channels, 1, padding=0)))
+
+    def forward(self, x, hint, timesteps, context, **kwargs): # TODO
+        t_emb = timestep_embedding(timesteps, self.model_channels, repeat_only=False)
+        emb = self.time_embed(t_emb)
+
+        guided_hint = self.input_hint_block(hint, emb, context)
+
+        outs = []
+
+        h = x.type(self.dtype)
+        for module, zero_conv in zip(self.input_blocks, self.zero_convs):
+            if guided_hint is not None:
+                h = module(h, emb, context)
+                h += guided_hint
+                guided_hint = None
+            else:
+                h = module(h, emb, context)
+            outs.append(zero_conv(h, emb, context))
+
+        h = self.middle_block(h, emb, context)
+        outs.append(self.middle_block_out(h, emb, context))
+
+        return outs
+
+class ControlLDM(LatentDiffusion):
+
+    def __init__(self, control_stage_config, control_key, only_mid_control, reg_net_path, *args, **kwargs):
+
+        super().__init__(*args, **kwargs)
+        self.control_model = instantiate_from_config(control_stage_config)
+        self.control_key = control_key
+        self.only_mid_control = only_mid_control
+        self.control_scales = [1.0] * 13
+        self.bbx_reg = RegNetwork()
+        if os.path.exists(reg_net_path):
+            state_dict = torch.load(reg_net_path, weights_only=False)['net']
+            state_dict = {k.replace('module.', ''): v for k, v in state_dict.items()}
+            self.bbx_reg.load_state_dict(state_dict)
+        for param in self.bbx_reg.parameters():
+            param.requires_grad = False
+        self.mask_threshold = 0.5
+
+    @torch.no_grad()
+    def get_input(self, batch, k, bs=None, *args, **kwargs):
+        x, c = super().get_input(batch, self.first_stage_key, *args, **kwargs)
+        control = batch[self.control_key]
+        if bs is not None:
+            control = control[:bs]
+        control = control.to(self.device)
+        control = einops.rearrange(control, 'b h w c -> b c h w')
+        control = control.to(memory_format=torch.contiguous_format).float()
+        return x, dict(c_crossattn=[c], c_concat=[control])
+
+    def gen_bbx(self, pred_t, fg_instance_bbx, mask):
+        img_height, img_width = 512, 512
+
+        pred_bbx = pred_t.new(pred_t.shape)
+        pred_bbx[:,0] = pred_t[:,0] * fg_instance_bbx[:,2] + fg_instance_bbx[:,0]
+        pred_bbx[:,1] = pred_t[:,1] * fg_instance_bbx[:,3] + fg_instance_bbx[:,1]
+        pred_bbx[:,2] = fg_instance_bbx[:,2] * torch.exp(pred_t[:,2])
+        pred_bbx[:,3] = fg_instance_bbx[:,3] * torch.exp(pred_t[:,3])
+        pred_bbx[:,4] = pred_t[:,4] * 180 / np.pi +fg_instance_bbx[:,4]
+        mask = mask.cpu().numpy()
+        bs = mask.shape[0]
+        for i in range(bs):
+            if pred_bbx[i,4] > 0:
+                temp = pred_bbx[i,2]
+                pred_bbx[i,2] = pred_bbx[i,3]
+                pred_bbx[i,3] = temp
+                pred_bbx[i,4] = pred_bbx[i,4] - 90
+            x, y, w, h, theta = pred_bbx[i,0], pred_bbx[i,1], pred_bbx[i,2], pred_bbx[i,3], pred_bbx[i,4]
+            box = ((float(x), float(y)), (float(w), float(h)), float(theta))
+            box_points = cv2.boxPoints(box)
+            box_points = np.clip(box_points, 0, [img_width - 1, img_height - 1])
+            box_points_int = np.int32(box_points)
+            cv2.fillPoly(mask[i], [box_points_int], 1)
+        mask = torch.tensor(mask).unsqueeze(1).to(dtype=torch.float32).to(self.device)
+        return mask
+
+    def apply_model(self, x_noisy, t, cond, get_stable_features=True, batch=None, *args, **kwargs): # TODO
+        assert isinstance(cond, dict)
+        diffusion_model = self.model.ip_adapter
+        cond_txt = torch.cat(cond['c_crossattn'], 1)
+
+        if cond['c_concat'] is None:
+            eps = diffusion_model(sample=x_noisy, timestep=t, encoder_hidden_states=cond_txt)
+            return eps
+        else:
+            input = batch['cls']
+            input = input.to(self.device)
+            input = einops.rearrange(input, 'b h w c -> b c h w')
+            pred_t = self.bbx_reg(input)
+
+            bbx_mask = batch['bbx']
+            fg_instance_bbx = batch['fg']
+            fg_instance_bbx = fg_instance_bbx.to(self.device)
+            bbx_region = self.gen_bbx(pred_t, fg_instance_bbx, bbx_mask)
+            hint = torch.cat(cond['c_concat'], 1)
+
+            control = self.control_model(x=x_noisy, hint=torch.cat((hint, bbx_region), 1), timesteps=t, context=cond_txt)
+            control = [c * scale for c, scale in zip(control, self.control_scales)]
+            mid_control = control.pop()
+            down_control = tuple(control)
+            reflection_features = generate_shadow_clip_features(batch['cls'], batch['objectmask'])
+
+            eps = diffusion_model(noisy_latents=x_noisy, timesteps=t, encoder_hidden_states=cond_txt, down_block_additional_residuals=down_control, mid_block_additional_residual=mid_control, mask_embeddings=reflection_features)
+            
+            return eps
+
+    @torch.no_grad()
+    def get_unconditional_conditioning(self, N):
+        return self.get_learned_conditioning([""] * N)
+
+    @torch.no_grad()
+    def log_images(self, batch, N=1, n_row=2, sample=False, ddim_steps=50, ddim_eta=0.0, return_keys=None,
+                   quantize_denoised=True, inpaint=True, plot_denoise_rows=False, plot_progressive_rows=True,
+                   plot_diffusion_rows=False, unconditional_guidance_scale=9.0, unconditional_guidance_label=None,
+                   use_ema_scope=True, use_x_T=False,
+                   **kwargs):
+        use_ddim = ddim_steps is not None
+
+        log = dict()
+        z, c = self.get_input(batch, self.first_stage_key, bs=N)
+        c_cat, c = c["c_concat"][0][:N], c["c_crossattn"][0][:N]
+        N = min(z.shape[0], N)
+        n_row = min(z.shape[0], n_row)
+        log["reconstruction"] = self.decode_first_stage(z)
+        log_control = c_cat * 2.0 - 1.0
+        log["control"] = log_control
+        log["conditioning"] = log_txt_as_img((512, 512), batch[self.cond_stage_key], size=16)
+        
+        if use_x_T == True:
+            x_T = super().q_sample(x_start=z, t=torch.tensor([999], device=self.device, dtype=torch.long), noise=torch.randn_like(z))
+        else:
+            x_T = None
+
+        if plot_diffusion_rows:
+            # get diffusion row
+            diffusion_row = list()
+            z_start = z[:n_row]
+            for t in range(self.num_timesteps):
+                if t % self.log_every_t == 0 or t == self.num_timesteps - 1:
+                    t = repeat(torch.tensor([t]), '1 -> b', b=n_row)
+                    t = t.to(self.device).long()
+                    noise = torch.randn_like(z_start)
+                    z_noisy = self.q_sample(x_start=z_start, t=t, noise=noise)
+                    diffusion_row.append(self.decode_first_stage(z_noisy))
+
+            diffusion_row = torch.stack(diffusion_row)  # n_log_step, n_row, C, H, W
+            diffusion_grid = rearrange(diffusion_row, 'n b c h w -> b n c h w')
+            diffusion_grid = rearrange(diffusion_grid, 'b n c h w -> (b n) c h w')
+            diffusion_grid = make_grid(diffusion_grid, nrow=diffusion_row.shape[0])
+            log["diffusion_row"] = diffusion_grid
+
+        if sample:
+            # get denoise row
+            samples, z_denoise_row = self.sample_log(cond={"c_concat": [c_cat], "c_crossattn": [c]},
+                                                     batch_size=N, ddim=use_ddim,
+                                                     ddim_steps=ddim_steps, eta=ddim_eta,
+                                                     batch=batch)
+            x_samples = self.decode_first_stage(samples)
+            log["samples"] = x_samples
+            if plot_denoise_rows:
+                denoise_grid = self._get_denoise_row_from_list(z_denoise_row)
+                log["denoise_row"] = denoise_grid
+
+        if unconditional_guidance_scale >= 1.0:
+            uc_cross = self.get_unconditional_conditioning(N)
+            uc_cat = c_cat  # torch.zeros_like(c_cat)
+            uc_full = {"c_concat": [uc_cat], "c_crossattn": [uc_cross]}
+            samples_cfg, _ = self.sample_log(cond={"c_concat": [c_cat], "c_crossattn": [c]},
+                                             batch_size=N, ddim=use_ddim,
+                                             ddim_steps=ddim_steps, eta=ddim_eta,
+                                             unconditional_guidance_scale=unconditional_guidance_scale,
+                                             unconditional_conditioning=uc_full, x_T=x_T,
+                                             batch=batch)
+            x_samples_cfg = self.decode_first_stage(samples_cfg)
+            log[f"samples_cfg_scale_{unconditional_guidance_scale:.2f}"] = x_samples_cfg
+            # log['predicted_mask'] = pred_mask
+
+        return log
+
+    @torch.no_grad()
+    def sample_log(self, cond, batch_size, ddim, ddim_steps, x_T=None, **kwargs):
+        ddim_sampler = DDIMSampler(self)
+        b, c, h, w = cond["c_concat"][0].shape
+        shape = (self.channels, h // 8, w // 8)
+
+        samples, intermediates = ddim_sampler.sample(ddim_steps, batch_size, shape, cond, verbose=False, x_T=x_T, **kwargs) #add x_T
+        return samples, intermediates
+
+    def configure_optimizers(self):
+        lr = self.learning_rate
+        params = []
+        params += list(self.control_model.parameters())
+        params += list(self.model.adapter_modules.parameters())
+        params += list(self.model.image_proj_model.parameters())
+        if not self.sd_locked:
+            params += list(self.model.diffusion_model.output_blocks.parameters())
+            params += list(self.model.diffusion_model.out.parameters())
+        opt = torch.optim.AdamW(params, lr=lr)
+        return opt
+
+    def low_vram_shift(self, is_diffusing):
+        if is_diffusing:
+            self.model = self.model.cuda()
+            self.control_model = self.control_model.cuda()
+            self.first_stage_model = self.first_stage_model.cpu()
+            self.cond_stage_model = self.cond_stage_model.cpu()
+        else:
+            self.model = self.model.cpu()
+            self.control_model = self.control_model.cpu()
+            self.first_stage_model = self.first_stage_model.cuda()
+            self.cond_stage_model = self.cond_stage_model.cuda()