File size: 15,061 Bytes
06d0fb4
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
449bc14
 
06d0fb4
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
3119b7f
 
 
 
 
06d0fb4
 
 
8e91a59
06d0fb4
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
79e1a03
6f4ab03
a01c15b
449bc14
48f38a9
bbc4ce7
06d0fb4
7fb97e6
f668242
7481b8b
06d0fb4
 
 
 
 
449bc14
 
06d0fb4
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
9331538
06d0fb4
 
 
3119b7f
06d0fb4
 
 
 
 
 
 
 
 
3119b7f
06d0fb4
f5f0556
06d0fb4
 
 
 
 
 
 
79e1a03
9331538
a01c15b
9331538
48f38a9
bbc4ce7
7fb97e6
f668242
7481b8b
3119b7f
06d0fb4
 
 
 
 
 
 
4e82511
06d0fb4
 
 
 
 
 
 
 
 
 
4e82511
 
 
06d0fb4
 
 
 
 
 
 
 
 
266f72f
 
 
 
 
 
 
 
 
 
 
 
06d0fb4
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
b5ac208
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
import os
import random
import shutil
import torch
import gradio as gr
from PIL import Image
from typing import List, Dict, Any

from .base_pipeline import BasePipeline
from core.settings import *
from utils.app_utils import sanitize_prompt
from core.workflow_assembler import WorkflowAssembler
from .workflow_executor import WorkflowExecutor
from .pipeline_input_processor import process_pipeline_inputs

class SdImagePipeline(BasePipeline):
    def get_required_models(self, model_display_name: str, **kwargs) -> List[str]:
        model_info = ALL_MODEL_MAP.get(model_display_name)
        if not model_info:
            return [model_display_name]

        path_or_components = model_info[1]
        if isinstance(path_or_components, dict):
            return [v for v in path_or_components.values() if v and v != "pixel_space"]
        else:
            return [model_display_name]

    def _gpu_logic(self, ui_inputs: Dict, loras_string: str, workflow: Dict[str, Any], assembler: WorkflowAssembler, progress=gr.Progress(track_tqdm=True)):
        model_display_name = ui_inputs['model_display_name']
        
        progress(0.4, desc="Executing workflow...")
        
        initial_objects = {}

        decoded_images_tensor = WorkflowExecutor.execute_workflow(workflow, initial_objects=initial_objects)
        
        output_images = []
        raw_seed = ui_inputs.get('seed')
        start_seed = int(raw_seed) if (raw_seed is not None and raw_seed != -1) else random.randint(0, 2**64 - 1)
        for i in range(decoded_images_tensor.shape[0]):
            img_tensor = decoded_images_tensor[i]
            pil_image = Image.fromarray((img_tensor.cpu().numpy() * 255.0).astype("uint8"))
            current_seed = start_seed + i
            
            width_for_meta = ui_inputs.get('width', 'N/A')
            height_for_meta = ui_inputs.get('height', 'N/A')

            params_string = f"{ui_inputs['positive_prompt']}\nNegative prompt: {ui_inputs['negative_prompt']}\n"
            params_string += f"Steps: {ui_inputs['num_inference_steps']}, Sampler: {ui_inputs['sampler']}, Scheduler: {ui_inputs['scheduler']}, CFG scale: {ui_inputs['guidance_scale']}, Seed: {current_seed}, Size: {width_for_meta}x{height_for_meta}, Base Model: {model_display_name}"
            if ui_inputs['task_type'] != 'txt2img': params_string += f", Denoise: {ui_inputs['denoise']}"
            if loras_string: params_string += f", {loras_string}"
            
            pil_image.info = {'parameters': params_string.strip()}
            output_images.append(pil_image)
            
        return output_images

    def run(self, ui_inputs: Dict, progress):
        progress(0, desc="Preparing models...")
        
        task_type = ui_inputs['task_type']
        model_display_name = ui_inputs['model_display_name']
        model_type = MODEL_TYPE_MAP.get(model_display_name, 'sdxl')
        
        architectures_dict = ARCHITECTURES_CONFIG.get('architectures', {})
        workflow_model_type = architectures_dict.get(model_type, {}).get("model_type", model_type.lower().replace(" ", "").replace(".", ""))
        
        ui_inputs['positive_prompt'] = sanitize_prompt(ui_inputs.get('positive_prompt', ''))
        ui_inputs['negative_prompt'] = sanitize_prompt(ui_inputs.get('negative_prompt', ''))

        required_models = self.get_required_models(model_display_name=model_display_name)

        arch_enabled_chains = FEATURES_CONFIG.get(workflow_model_type, {}).get('enabled_chains', [])
        task_enabled_chains = TASK_FEATURES_CONFIG.get(task_type, {}).get('enabled_chains', [])
        enabled_chains = set([c for c in arch_enabled_chains if c in task_enabled_chains])

        is_pid_enabled = (ui_inputs.get('pid_settings', 'OFF') == 'ON' and 'pid' in enabled_chains and task_type == 'txt2img')
        if is_pid_enabled:
            import yaml
            pid_config_path = os.path.join(os.path.dirname(os.path.dirname(os.path.dirname(os.path.abspath(__file__)))), 'yaml', 'pid.yaml')
            pid_unet_name = "pid_1.5_flux1_1024_to_4096_4step_int8_convrot.safetensors"
            try:
                with open(pid_config_path, 'r', encoding='utf-8') as f:
                    pid_config = yaml.safe_load(f) or {}
                pid_items = pid_config.get("PiD", [])
                for item in pid_items:
                    archs = item.get("architectures", [])
                    if workflow_model_type in archs:
                        pid_unet_name = item.get("filepath")
                        break
            except Exception as e:
                print(f"Error loading PiD config for download: {e}")

            if pid_unet_name not in required_models:
                required_models.append(pid_unet_name)
            if "gemma_2_2b_it_elm_fp8_scaled.safetensors" not in required_models:
                required_models.append("gemma_2_2b_it_elm_fp8_scaled.safetensors")

        self.model_manager.ensure_models_downloaded(required_models, progress=progress)
        
        temp_files_to_clean = []
        try:
            processed = process_pipeline_inputs(ui_inputs, progress, workflow_model_type)
            temp_files_to_clean.extend(processed["temp_files_to_clean"])
            
            active_loras_for_gpu = processed["active_loras_for_gpu"]
            active_loras_for_meta = processed["active_loras_for_meta"]
            active_controlnets = processed["active_controlnets"]
            active_ipadapters = processed["active_ipadapters"]
            active_flux1_ipadapters = processed["active_flux1_ipadapters"]
            active_sd3_ipadapters = processed["active_sd3_ipadapters"]
            active_styles = processed["active_styles"]
            active_reference_latents = processed["active_reference_latents"]
            active_hidream_o1_reference = processed["active_hidream_o1_reference"]
            active_sensenova_reference = processed.get("active_sensenova_reference", [])
            active_joyai_reference = processed.get("active_joyai_reference", [])
            active_krea2_identity_edit = processed.get("active_krea2_identity_edit", [])
            active_krea2_reference_edit = processed.get("active_krea2_reference_edit", [])
            active_qwen_image_edit = processed.get("active_qwen_image_edit", [])
            active_reference_images = processed.get("active_reference_images", [])
            active_conditioning = processed["active_conditioning"]
            active_prompt_enhancer = processed.get("active_prompt_enhancer", [])
            active_moss_image_control = processed.get("active_moss_image_control", [])
            active_moss_image_edit = processed.get("active_moss_image_edit", [])

            loras_string = f"LoRAs: [{', '.join(active_loras_for_meta)}]" if active_loras_for_meta else ""
            
            progress(0.8, desc="Assembling workflow...")
            
            seed_val = ui_inputs.get('seed')
            if seed_val is None or seed_val == -1:
                ui_inputs['seed'] = random.randint(0, 2**32 - 1)
            
            model_info = ALL_MODEL_MAP[model_display_name]
            path_or_components = model_info[1]
            latent_type = model_info[3] if len(model_info) > 3 and model_info[3] else 'latent'
            latent_generator_template = "EmptyLatentImage"
            if latent_type == 'sd3_latent':
                latent_generator_template = "EmptySD3LatentImage"
            elif latent_type == 'chroma_radiance_latent':
                latent_generator_template = "EmptyChromaRadianceLatentImage"
            elif latent_type == 'hunyuan_latent':
                latent_generator_template = "EmptyHunyuanImageLatent"

            dynamic_values = {
                'task_type': ui_inputs['task_type'], 
                'model_type': workflow_model_type,
                'latent_type': latent_type,
                'latent_generator_template': latent_generator_template
            }
            
            recipe_path = os.path.join(os.path.dirname(__file__), "workflow_recipes", "unified_recipe.yaml")
            assembler = WorkflowAssembler(recipe_path, dynamic_values=dynamic_values)
            
            hidream_o1_smoothing_data = []
            if 'hidream_o1_smoothing' in enabled_chains and model_display_name == "HiDream-O1-Image":
                hidream_o1_smoothing_data.append({})

            workflow_inputs = {
                **ui_inputs,
                "positive_prompt": ui_inputs['positive_prompt'], "negative_prompt": ui_inputs['negative_prompt'],
                "seed": ui_inputs['seed'], "steps": ui_inputs['num_inference_steps'], "cfg": ui_inputs['guidance_scale'],
                "sampler_name": ui_inputs['sampler'], "scheduler": ui_inputs['scheduler'],
                "batch_size": ui_inputs['batch_size'],
                "denoise": ui_inputs['denoise'],
                "vae_name": ui_inputs.get('vae_name') if 'vae' in enabled_chains else None,
                "lora_chain": active_loras_for_gpu,
                "controlnet_chain": active_controlnets,
                "ipadapter_chain": active_ipadapters,
                "flux1_ipadapter_chain": active_flux1_ipadapters,
                "sd3_ipadapter_chain": active_sd3_ipadapters,
                "style_chain": active_styles,
                "conditioning_chain": active_conditioning,
                "reference_latent_chain": active_reference_latents,
                "hidream_o1_reference_chain": active_hidream_o1_reference,
                "sensenova_reference_chain": active_sensenova_reference,
                "joyai_image_chain": active_joyai_reference,
                "krea2_identity_edit_chain": active_krea2_identity_edit,
                "krea2_style_reference_chain": active_krea2_reference_edit,
                "qwen_image_edit_chain": active_qwen_image_edit,
                "reference_image_chain": active_reference_images,
                "prompt_enhancer_chain": active_prompt_enhancer,
                "moss_image_control_chain": active_moss_image_control,
                "moss_image_edit_chain": active_moss_image_edit,
                "vae_chain": [ui_inputs.get('vae_name')] if (ui_inputs.get('vae_name') and 'vae' in enabled_chains) else [],
                "hidream_o1_smoothing_chain": hidream_o1_smoothing_data,
                "pid_chain": [ui_inputs.get('pid_settings', 'OFF')] if is_pid_enabled else [],
                "scheduler_width": ui_inputs.get('width', 1024),
                "scheduler_height": ui_inputs.get('height', 1024),
            }

            if isinstance(path_or_components, dict):
                workflow_inputs.update(path_or_components)
                workflow_inputs.update({
                    'unet_name': path_or_components.get('unet'),
                    'unet_uncond_name': path_or_components.get('unet_uncond'),
                    'vae_name': ui_inputs.get('vae_name') or path_or_components.get('vae'),
                    'clip_name': path_or_components.get('clip'),
                    'clip1_name': path_or_components.get('clip1'),
                    'clip2_name': path_or_components.get('clip2'),
                    'clip3_name': path_or_components.get('clip3'),
                    'clip4_name': path_or_components.get('clip4'),
                    'lora_name': path_or_components.get('lora'),
                    'qwen35_model': path_or_components.get('qwen35_model'),
                    'adapter_name': path_or_components.get('adapter_name'),
                    'adapter_strength': path_or_components.get('adapter_strength', 1.0),
                })
            else:
                workflow_inputs['model_name'] = path_or_components
            
            if task_type == 'txt2img':
                workflow_inputs['width'] = ui_inputs['width']
                workflow_inputs['height'] = ui_inputs['height']
            
            workflow = assembler.assemble(workflow_inputs)
            
            post_assembled_models = []
            for node_id, node in workflow.items():
                if not isinstance(node, dict):
                    continue
                inputs = node.get('inputs', {})
                for k, v in inputs.items():
                    if isinstance(v, str) and v in ALL_FILE_DOWNLOAD_MAP:
                        if v not in required_models and v not in post_assembled_models:
                            post_assembled_models.append(v)
            if post_assembled_models:
                self.model_manager.ensure_models_downloaded(post_assembled_models, progress=progress)

            progress(1.0, desc="All models ready. Requesting GPU for generation...")

            results = self._execute_gpu_logic(
                self._gpu_logic,
                duration=ui_inputs['zero_gpu_duration'],
                default_duration=60,
                task_name=f"ImageGen ({task_type})",
                ui_inputs=ui_inputs,
                loras_string=loras_string,
                workflow=workflow,
                assembler=assembler, 
                progress=progress
            )
            
            import json
            import glob
            from PIL import PngImagePlugin
            
            prompt_json = json.dumps(workflow)
            
            out_dir = os.path.abspath(OUTPUT_DIR)
            os.makedirs(out_dir, exist_ok=True)
            
            try:
                existing_files = glob.glob(os.path.join(out_dir, "gen_*.png"))
                existing_files.sort(key=os.path.getmtime)
                while len(existing_files) > 50:
                    os.remove(existing_files.pop(0))
            except Exception as e:
                print(f"Warning: Failed to cleanup output dir: {e}")

            final_results = []
            for img in results:
                if not isinstance(img, Image.Image):
                    final_results.append(img)
                    continue
                    
                metadata = PngImagePlugin.PngInfo()
                params_string = img.info.get("parameters", "")
                if params_string:
                    metadata.add_text("parameters", params_string)
                metadata.add_text("prompt", prompt_json)
                
                filename = f"gen_{random.randint(1000000, 9999999)}.png"
                filepath = os.path.join(out_dir, filename)
                img.save(filepath, "PNG", pnginfo=metadata)
                final_results.append(filepath)
            
            results = final_results

        finally:
            for temp_file in temp_files_to_clean:
                if temp_file and os.path.exists(temp_file):
                    os.remove(temp_file)
                    print(f"✅ Cleaned up temp file: {temp_file}")

        return results