Merge code

app_v2v.py

@@ -296,6 +296,285 @@ def set_mp4_comments_imageio_ffmpeg(input_file, comments):
         print(f"Error saving prompt to video metadata, ffmpeg may be required: "+str(e))
         return False
 
+@torch.no_grad()
+def worker(input_image, prompts, n_prompt, seed, total_second_length, latent_window_size, steps, cfg, gs, rs, gpu_memory_preservation, use_teacache, mp4_crf):
+    def encode_prompt(prompt, n_prompt):
+        llama_vec, clip_l_pooler = encode_prompt_conds(prompt, text_encoder, text_encoder_2, tokenizer, tokenizer_2)
+
+        if cfg == 1:
+            llama_vec_n, clip_l_pooler_n = torch.zeros_like(llama_vec), torch.zeros_like(clip_l_pooler)
+        else:
+            llama_vec_n, clip_l_pooler_n = encode_prompt_conds(n_prompt, text_encoder, text_encoder_2, tokenizer, tokenizer_2)
+
+        llama_vec, llama_attention_mask = crop_or_pad_yield_mask(llama_vec, length=512)
+        llama_vec_n, llama_attention_mask_n = crop_or_pad_yield_mask(llama_vec_n, length=512)
+
+        llama_vec = llama_vec.to(transformer.dtype)
+        llama_vec_n = llama_vec_n.to(transformer.dtype)
+        clip_l_pooler = clip_l_pooler.to(transformer.dtype)
+        clip_l_pooler_n = clip_l_pooler_n.to(transformer.dtype)
+        return [llama_vec, clip_l_pooler, llama_vec_n, clip_l_pooler_n, llama_attention_mask, llama_attention_mask_n]
+
+    total_latent_sections = (total_second_length * 30) / (latent_window_size * 4)
+    total_latent_sections = int(max(round(total_latent_sections), 1))
+
+    job_id = generate_timestamp()
+
+    stream.output_queue.push(('progress', (None, '', make_progress_bar_html(0, 'Starting ...'))))
+
+    try:
+        # Clean GPU
+        if not high_vram:
+            unload_complete_models(
+                text_encoder, text_encoder_2, image_encoder, vae, transformer
+            )
+
+        # Text encoding
+
+        stream.output_queue.push(('progress', (None, '', make_progress_bar_html(0, 'Text encoding ...'))))
+
+        if not high_vram:
+            fake_diffusers_current_device(text_encoder, gpu)  # since we only encode one text - that is one model move and one encode, offload is same time consumption since it is also one load and one encode.
+            load_model_as_complete(text_encoder_2, target_device=gpu)
+
+        prompt_parameters = []
+        
+        for prompt_part in prompts:
+            prompt_parameters.append(encode_prompt(prompt_part, n_prompt))
+
+        # Processing input image
+
+        stream.output_queue.push(('progress', (None, '', make_progress_bar_html(0, 'Image processing ...'))))
+
+        H, W, C = input_image.shape
+        height, width = find_nearest_bucket(H, W, resolution=640)
+        input_image_np = resize_and_center_crop(input_image, target_width=width, target_height=height)
+
+        Image.fromarray(input_image_np).save(os.path.join(outputs_folder, f'{job_id}.png'))
+
+        input_image_pt = torch.from_numpy(input_image_np).float() / 127.5 - 1
+        input_image_pt = input_image_pt.permute(2, 0, 1)[None, :, None]
+
+        # VAE encoding
+
+        stream.output_queue.push(('progress', (None, '', make_progress_bar_html(0, 'VAE encoding ...'))))
+
+        if not high_vram:
+            load_model_as_complete(vae, target_device=gpu)
+
+        start_latent = vae_encode(input_image_pt, vae)
+
+        # CLIP Vision
+
+        stream.output_queue.push(('progress', (None, '', make_progress_bar_html(0, 'CLIP Vision encoding ...'))))
+
+        if not high_vram:
+            load_model_as_complete(image_encoder, target_device=gpu)
+
+        image_encoder_output = hf_clip_vision_encode(input_image_np, feature_extractor, image_encoder)
+        image_encoder_last_hidden_state = image_encoder_output.last_hidden_state
+
+        # Dtype
+
+        image_encoder_last_hidden_state = image_encoder_last_hidden_state.to(transformer.dtype)
+
+        # Sampling
+
+        stream.output_queue.push(('progress', (None, '', make_progress_bar_html(0, 'Start sampling ...'))))
+
+        rnd = torch.Generator("cpu").manual_seed(seed)
+
+        history_latents = torch.zeros(size=(1, 16, 16 + 2 + 1, height // 8, width // 8), dtype=torch.float32).cpu()
+        history_pixels = None
+
+        history_latents = torch.cat([history_latents, start_latent.to(history_latents)], dim=2)
+        total_generated_latent_frames = 1
+
+        for section_index in range(total_latent_sections):
+            if stream.input_queue.top() == 'end':
+                stream.output_queue.push(('end', None))
+                return
+
+            print(f'section_index = {section_index}, total_latent_sections = {total_latent_sections}')
+
+            if len(prompt_parameters) > 0:
+                [llama_vec, clip_l_pooler, llama_vec_n, clip_l_pooler_n, llama_attention_mask, llama_attention_mask_n] = prompt_parameters.pop(0)
+
+            if not high_vram:
+                unload_complete_models()
+                move_model_to_device_with_memory_preservation(transformer, target_device=gpu, preserved_memory_gb=gpu_memory_preservation)
+
+            if use_teacache:
+                transformer.initialize_teacache(enable_teacache=True, num_steps=steps)
+            else:
+                transformer.initialize_teacache(enable_teacache=False)
+
+            def callback(d):
+                preview = d['denoised']
+                preview = vae_decode_fake(preview)
+
+                preview = (preview * 255.0).detach().cpu().numpy().clip(0, 255).astype(np.uint8)
+                preview = einops.rearrange(preview, 'b c t h w -> (b h) (t w) c')
+
+                if stream.input_queue.top() == 'end':
+                    stream.output_queue.push(('end', None))
+                    raise KeyboardInterrupt('User ends the task.')
+
+                current_step = d['i'] + 1
+                percentage = int(100.0 * current_step / steps)
+                hint = f'Sampling {current_step}/{steps}'
+                desc = f'Total generated frames: {int(max(0, total_generated_latent_frames * 4 - 3))}, Video length: {max(0, (total_generated_latent_frames * 4 - 3) / 30) :.2f} seconds (FPS-30). The video is being extended now ...'
+                stream.output_queue.push(('progress', (preview, desc, make_progress_bar_html(percentage, hint))))
+                return
+
+            indices = torch.arange(0, sum([1, 16, 2, 1, latent_window_size])).unsqueeze(0)
+            clean_latent_indices_start, clean_latent_4x_indices, clean_latent_2x_indices, clean_latent_1x_indices, latent_indices = indices.split([1, 16, 2, 1, latent_window_size], dim=1)
+            clean_latent_indices = torch.cat([clean_latent_indices_start, clean_latent_1x_indices], dim=1)
+
+            clean_latents_4x, clean_latents_2x, clean_latents_1x = history_latents[:, :, -sum([16, 2, 1]):, :, :].split([16, 2, 1], dim=2)
+            clean_latents = torch.cat([start_latent.to(history_latents), clean_latents_1x], dim=2)
+
+            generated_latents = sample_hunyuan(
+                transformer=transformer,
+                sampler='unipc',
+                width=width,
+                height=height,
+                frames=latent_window_size * 4 - 3,
+                real_guidance_scale=cfg,
+                distilled_guidance_scale=gs,
+                guidance_rescale=rs,
+                # shift=3.0,
+                num_inference_steps=steps,
+                generator=rnd,
+                prompt_embeds=llama_vec,
+                prompt_embeds_mask=llama_attention_mask,
+                prompt_poolers=clip_l_pooler,
+                negative_prompt_embeds=llama_vec_n,
+                negative_prompt_embeds_mask=llama_attention_mask_n,
+                negative_prompt_poolers=clip_l_pooler_n,
+                device=gpu,
+                dtype=torch.bfloat16,
+                image_embeddings=image_encoder_last_hidden_state,
+                latent_indices=latent_indices,
+                clean_latents=clean_latents,
+                clean_latent_indices=clean_latent_indices,
+                clean_latents_2x=clean_latents_2x,
+                clean_latent_2x_indices=clean_latent_2x_indices,
+                clean_latents_4x=clean_latents_4x,
+                clean_latent_4x_indices=clean_latent_4x_indices,
+                callback=callback,
+            )
+
+            total_generated_latent_frames += int(generated_latents.shape[2])
+            history_latents = torch.cat([history_latents, generated_latents.to(history_latents)], dim=2)
+
+            if not high_vram:
+                offload_model_from_device_for_memory_preservation(transformer, target_device=gpu, preserved_memory_gb=8)
+                load_model_as_complete(vae, target_device=gpu)
+
+            real_history_latents = history_latents[:, :, -total_generated_latent_frames:, :, :]
+
+            if history_pixels is None:
+                history_pixels = vae_decode(real_history_latents, vae).cpu()
+            else:
+                section_latent_frames = latent_window_size * 2
+                overlapped_frames = latent_window_size * 4 - 3
+
+                current_pixels = vae_decode(real_history_latents[:, :, -section_latent_frames:], vae).cpu()
+                history_pixels = soft_append_bcthw(history_pixels, current_pixels, overlapped_frames)
+
+            if not high_vram:
+                unload_complete_models()
+
+            output_filename = os.path.join(outputs_folder, f'{job_id}_{total_generated_latent_frames}.mp4')
+
+            save_bcthw_as_mp4(history_pixels, output_filename, fps=30, crf=mp4_crf)
+
+            print(f'Decoded. Current latent shape {real_history_latents.shape}; pixel shape {history_pixels.shape}')
+
+            stream.output_queue.push(('file', output_filename))
+    except:
+        traceback.print_exc()
+
+        if not high_vram:
+            unload_complete_models(
+                text_encoder, text_encoder_2, image_encoder, vae, transformer
+            )
+
+    stream.output_queue.push(('end', None))
+    return
+
+def get_duration(input_image, prompt, t2v, n_prompt, randomize_seed, seed, total_second_length, latent_window_size, steps, cfg, gs, rs, gpu_memory_preservation, use_teacache, mp4_crf):
+    global total_second_length_debug_value
+
+    if total_second_length_debug_value is not None:
+        return min(total_second_length_debug_value * 60, 600)
+    return total_second_length * 60
+
+
+@spaces.GPU(duration=get_duration)
+def process(input_image, prompt, 
+            t2v=False, 
+            n_prompt="",
+            randomize_seed=True,
+            seed=31337, 
+            total_second_length=5, 
+            latent_window_size=9, 
+            steps=25, 
+            cfg=1.0, 
+            gs=10.0, 
+            rs=0.0, 
+            gpu_memory_preservation=6, 
+            use_teacache=True,
+            mp4_crf=16
+           ):
+    global stream, input_image_debug_value, prompt_debug_value, total_second_length_debug_value
+
+    if torch.cuda.device_count() == 0:
+        gr.Warning('Set this space to GPU config to make it work.')
+        return None, None, None, None, None, None
+
+    if input_image_debug_value is not None or prompt_debug_value is not None or total_second_length_debug_value is not None:
+        print("Debug mode")
+        input_image = input_image_debug_value
+        prompt = prompt_debug_value
+        total_second_length = total_second_length_debug_value
+        input_image_debug_value = prompt_debug_value = total_second_length_debug_value = None
+
+    if randomize_seed:
+        seed = random.randint(0, np.iinfo(np.int32).max)
+
+    prompts = prompt.split(";")
+    
+    # assert input_image is not None, 'No input image!'
+    if t2v:
+        default_height, default_width = 640, 640
+        input_image = np.ones((default_height, default_width, 3), dtype=np.uint8) * 255
+        print("No input image provided. Using a blank white image.")
+    
+    yield None, None, '', '', gr.update(interactive=False), gr.update(interactive=True)
+
+    stream = AsyncStream()
+
+    async_run(worker, input_image, prompts, n_prompt, seed, total_second_length, latent_window_size, steps, cfg, gs, rs, gpu_memory_preservation, use_teacache, mp4_crf)
+
+    output_filename = None
+
+    while True:
+        flag, data = stream.output_queue.next()
+
+        if flag == 'file':
+            output_filename = data
+            yield output_filename, gr.update(), gr.update(), gr.update(), gr.update(interactive=False), gr.update(interactive=True)
+
+        if flag == 'progress':
+            preview, desc, html = data
+            yield gr.update(), gr.update(visible=True, value=preview), desc, html, gr.update(interactive=False), gr.update(interactive=True)
+
+        if flag == 'end':
+            yield output_filename, gr.update(visible=False), gr.update(), '', gr.update(interactive=True), gr.update(interactive=False)
+            break
+
 # 20250506 pftq: Modified worker to accept video input and clean frame count
 @spaces.GPU()
 @torch.no_grad()