bench: --ref-downscale option — pre-shrink refs before the node (24GB OOM mitigation)

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

bench/minimax_bench.py

@@ -85,7 +85,8 @@ def snap_length(seconds: float) -> int:
 
 def build_graph(task, model_file, prompt, width, height, length, steps, seed,
                 sampler="res_multistep", scheduler="simple", image=BENCH_IMAGE,
-                ref_image_size="match", ref_count=1, filename_prefix="bench/run"):
+                ref_image_size="match", ref_count=1, ref_downscale=0.0,
+                filename_prefix="bench/run"):
     g = {
         "1": {"class_type": "UNETLoader",
               "inputs": {"unet_name": model_file, "weight_dtype": "default"}},
@@ -122,15 +123,25 @@ def build_graph(task, model_file, prompt, width, height, length, steps, seed,
                              **common}}
     elif task == "ref2v":
         ref_inputs = {}
+
+        def ref_source(nid_load, fname):
+            g[nid_load] = {"class_type": "LoadImage", "inputs": {"image": fname}}
+            if not ref_downscale:
+                return [nid_load, 0]
+            nid_scale = str(int(nid_load) + 40)
+            g[nid_scale] = {"class_type": "ImageScaleToTotalPixels",
+                            "inputs": {"image": [nid_load, 0],
+                                       "upscale_method": "lanczos",
+                                       "megapixels": ref_downscale,
+                                       "resolution_steps": 1}}
+            return [nid_scale, 0]
+
         if ref_count <= 1:
-            g["10"] = {"class_type": "LoadImage", "inputs": {"image": image}}
+            ref_inputs["ref_images.ref_image_0"] = ref_source("10", image)
-            ref_inputs["ref_images.ref_image_0"] = ["10", 0]
         else:
             for i in range(ref_count):
-                nid = str(20 + i)
+                ref_inputs[f"ref_images.ref_image_{i}"] = ref_source(
-                g[nid] = {"class_type": "LoadImage",
+                    str(20 + i), BENCH_REF_SET[i % len(BENCH_REF_SET)])
-                          "inputs": {"image": BENCH_REF_SET[i % len(BENCH_REF_SET)]}}
-                ref_inputs[f"ref_images.ref_image_{i}"] = [nid, 0]
         g["5"] = {"class_type": "MiniMaxH3ReferenceToVideo",
                   "inputs": {"clip": ["2", 0], "vae": ["3", 0], "audio_vae": ["4", 0],
                              "ref_image_size": ref_image_size,
@@ -173,7 +184,7 @@ def run_one(api, cfg, poll=2.0, timeout=3600):
     graph = build_graph(**{k: v for k, v in cfg.items() if k in (
         "task", "model_file", "prompt", "width", "height", "length", "steps", "seed",
         "sampler", "scheduler", "image", "ref_image_size", "ref_count",
-        "filename_prefix")})
+        "ref_downscale", "filename_prefix")})
     peak = {"vram": 0}
     stop = threading.Event()
 
@@ -263,7 +274,7 @@ def default_prompt(task):
 
 def make_cfg(task, model_file, width, height, seconds, steps, seed, label, warm,
              sampler="res_multistep", scheduler="simple", ref_image_size="match",
-             ref_count=1):
+             ref_count=1, ref_downscale=0.0):
     prompt = ref2v_prompt(ref_count) if task == "ref2v" else default_prompt(task)
     return {
         "label": label, "task": task, "model_file": model_file,
@@ -271,7 +282,7 @@ def make_cfg(task, model_file, width, height, seconds, steps, seed, label, warm,
         "seconds": seconds, "length": snap_length(seconds), "steps": steps,
         "seed": seed, "sampler": sampler, "scheduler": scheduler,
         "image": BENCH_IMAGE, "ref_image_size": ref_image_size,
-        "ref_count": ref_count, "warm": warm,
+        "ref_count": ref_count, "ref_downscale": ref_downscale, "warm": warm,
         "filename_prefix": f"bench/{label}",
     }
 
@@ -395,6 +406,8 @@ def main():
     ap.add_argument("--ref-image-size", default="match", choices=["match", "max"])
     ap.add_argument("--ref-count", type=int, default=1,
                     help="number of reference images for ref2v (1-8)")
+    ap.add_argument("--ref-downscale", type=float, default=0.0,
+                    help="pre-downscale refs to this many megapixels before the node (0=off)")
     ap.add_argument("--label", default=None)
     ap.add_argument("--prefix", default="", help="label prefix for suites, e.g. prunedint8_")
     ap.add_argument("--warm", action="store_true", help="mark run as warm (model preloaded)")
@@ -423,7 +436,7 @@ def main():
         cfgs = [make_cfg(args.task, model, args.width, args.height, args.seconds,
                          args.steps, args.seed, label, args.warm,
                          args.sampler, args.scheduler, args.ref_image_size,
-                         args.ref_count)]
+                         args.ref_count, args.ref_downscale)]
     else:
         ap.error("need --suite or --task")