import torch import json import os import time import random import itertools import folder_paths import nodes import comfy.utils import comfy.sd import comfy.samplers from PIL import Image import numpy as np from .html_generator import get_html_template class SamplerGridTester: @classmethod def INPUT_TYPES(s): return { "required": { "ckpt_name": (folder_paths.get_filename_list("checkpoints"), ), "positive_text": ("STRING", {"multiline": True, "default": "masterpiece, best quality, 1girl"}), "negative_text": ("STRING", {"multiline": True, "default": "bad quality, worst quality, lowres"}), "seed": ("INT", {"default": 0, "min": 0, "max": 0xffffffffffffffff}), "denoise": ("FLOAT", {"default": 1.0, "min": 0.0, "max": 1.0, "step": 0.01}), "vae_batch_size": ("INT", {"default": 4, "min": -1, "max": 64}), "configs_json": ("STRING", {"multiline": True, "default": '[{"sampler": "euler", "scheduler": "normal", "steps": 20, "cfg": 7.0}]'}), "resolutions_json": ("STRING", {"multiline": True, "default": '[[1024, 1024]]'}), "session_name": ("STRING", {"default": "my_session"}), }, "optional": { "optional_model": ("MODEL",), "optional_clip": ("CLIP",), "optional_vae": ("VAE",), "optional_positive": ("CONDITIONING",), "optional_negative": ("CONDITIONING",), }, "hidden": {"unique_id": "UNIQUE_ID"}, } RETURN_TYPES = ("STRING",) RETURN_NAMES = ("dashboard_html",) FUNCTION = "run_tests" CATEGORY = "sampling/testing" def run_tests(self, ckpt_name, positive_text, negative_text, seed, denoise, vae_batch_size, configs_json, resolutions_json, session_name, unique_id, optional_model=None, optional_clip=None, optional_vae=None, optional_positive=None, optional_negative=None): try: raw_configs = json.loads(configs_json.strip()) resolutions = json.loads(resolutions_json.strip()) except Exception as e: raise ValueError(f"JSON Parsing Error: {e}") # --- LOAD RESOURCES --- # 1. Model & Clip & VAE if optional_model is not None and optional_clip is not None and optional_vae is not None: model, clip, vae = optional_model, optional_clip, optional_vae model_name_for_meta = "External Model" else: ckpt_path = folder_paths.get_full_path("checkpoints", ckpt_name) out = comfy.sd.load_checkpoint_guess_config(ckpt_path, output_vae=True, output_clip=True, embedding_directory=folder_paths.get_folder_paths("embeddings")) model, clip, vae = out[:3] model_name_for_meta = ckpt_name # 2. Conditioning (Prompts) -- FIXED FORMATTING if optional_positive is not None: positive = optional_positive pos_text_for_meta = "External Conditioning" else: tokens = clip.tokenize(positive_text) cond, pooled = clip.encode_from_tokens(tokens, return_pooled=True) # Wrap in the list structure ComfyUI samplers expect: [[cond, {"pooled_output": pooled}]] positive = [[cond, {"pooled_output": pooled}]] pos_text_for_meta = positive_text if optional_negative is not None: negative = optional_negative neg_text_for_meta = "External Conditioning" else: tokens = clip.tokenize(negative_text) cond, pooled = clip.encode_from_tokens(tokens, return_pooled=True) # Wrap in the list structure ComfyUI samplers expect negative = [[cond, {"pooled_output": pooled}]] neg_text_for_meta = negative_text # --- SETUP FOLDERS & MANIFEST --- base_dir = os.path.join(folder_paths.get_output_directory(), "benchmarks", session_name) img_dir = os.path.join(base_dir, "images") os.makedirs(img_dir, exist_ok=True) manifest_path = os.path.join(base_dir, "manifest.json") existing_data = {"items": [], "meta": {}} if os.path.exists(manifest_path): try: with open(manifest_path, "r") as f: d = json.load(f) if isinstance(d, list): existing_data["items"] = d else: existing_data = d except: pass # Update Session Meta existing_data["meta"] = { "model": model_name_for_meta, "positive": pos_text_for_meta, "negative": neg_text_for_meta, "updated": int(time.time()) } # --- EXPAND CONFIGS --- ALL_SCHEDULERS = comfy.samplers.KSampler.SCHEDULERS ALL_SAMPLERS = comfy.samplers.KSampler.SAMPLERS expanded = [] for entry in raw_configs: def to_list(x): return x if isinstance(x, list) else [x] samplers = ALL_SAMPLERS if entry.get("sampler") == "*" else to_list(entry.get("sampler", "euler")) schedulers = ALL_SCHEDULERS if entry.get("scheduler") == "*" else to_list(entry.get("scheduler", "normal")) steps_l = to_list(entry.get("steps", 20)) cfgs = to_list(entry.get("cfg", 7.0)) loras = to_list(entry.get("lora", "None")) str_m = to_list(entry.get("str_model", 1.0)) str_c = to_list(entry.get("str_clip", 1.0)) for combo in itertools.product(samplers, schedulers, steps_l, cfgs, loras, str_m, str_c): expanded.append({ "sampler": combo[0], "scheduler": combo[1], "steps": combo[2], "cfg": combo[3], "lora": combo[4], "str_model": combo[5], "str_clip": combo[6] }) print(f"[GridTester] Processing {len(expanded) * len(resolutions)} items...") # --- CACHING & BATCHING --- cached_lora_key = None cached_model, cached_clip = None, None pending_batch = [] def flush_batch(batch_list): if not batch_list: return latents_to_decode = torch.cat([x[0] for x in batch_list], dim=0) decoded = vae.decode(latents_to_decode) for i, img_tensor in enumerate(decoded): img_np = 255. * img_tensor.cpu().numpy() img = Image.fromarray(np.clip(img_np, 0, 255).astype(np.uint8)) meta = batch_list[i][1] ts = int(time.time() * 100000) + random.randint(0,1000) filename = f"img_{ts}.webp" img.save(os.path.join(img_dir, filename), quality=80) meta.update({ "id": ts, "file": f"/view?filename={filename}&type=output&subfolder=benchmarks/{session_name}/images", "rejected": False }) existing_data["items"].insert(0, meta) for res in resolutions: w, h = res[0], res[1] flush_batch(pending_batch) pending_batch = [] for conf in expanded: # --- MODEL LOADING LOGIC --- current_lora_key = (conf["lora"], conf["str_model"], conf["str_clip"]) # Check cache if current_lora_key == cached_lora_key and cached_model is not None: curr_model, curr_clip = cached_model, cached_clip else: # Reset to base curr_model, curr_clip = model, clip # Apply LoRA if needed if conf["lora"] != "None": path = folder_paths.get_full_path("loras", conf["lora"]) if path: lora_data = comfy.utils.load_torch_file(path) curr_model, curr_clip = comfy.sd.load_lora_for_models(curr_model, curr_clip, lora_data, conf["str_model"], conf["str_clip"]) cached_lora_key = current_lora_key cached_model, cached_clip = curr_model, curr_clip # Generate latent_in = {"samples": torch.zeros([1, 4, h // 8, w // 8])} t0 = time.time() try: result = nodes.common_ksampler(model=curr_model, seed=seed, steps=conf["steps"], cfg=conf["cfg"], sampler_name=conf["sampler"], scheduler=conf["scheduler"], positive=positive, negative=negative, latent=latent_in, denoise=denoise) except Exception as e: print(f"Generation Failed: {e}") continue duration = round(time.time() - t0, 3) meta = conf.copy() meta.update({"width": w, "height": h, "duration": duration, "seed": seed}) pending_batch.append((result[0]["samples"], meta)) if vae_batch_size != -1 and len(pending_batch) >= vae_batch_size: flush_batch(pending_batch) pending_batch = [] flush_batch(pending_batch) with open(manifest_path, "w") as f: json.dump(existing_data, f, indent=4) html = get_html_template(session_name, existing_data, unique_id) return (html,)