Free all old model patchers, gc.collect(), then free tensors.

Signed-off-by: bigcat88 <bigcat88@icloud.com>

AGENTS.md

@@ -171,6 +171,9 @@
 - Reuse existing model classes, blocks, ops, and helper modules when appropriate.
   Before implementing a new version of a model component, search the existing
   model code for a class or helper that already provides the behavior.
+- Model detection code that inspects linear weight shapes should only use the
+  first dimension. The second dimension may be half the original size for
+  NVFP4 or other 4-bit quantized models.
 - Avoid adding `einops` usage in core inference code. Use native torch tensor
   ops such as `reshape`, `view`, `permute`, `transpose`, `flatten`, `unflatten`,
   `unsqueeze`, and `squeeze` instead.

comfy_api/latest/_input_impl/video_types.py

@@ -281,18 +281,11 @@ class VideoFromFile(VideoInput):
         video_done = False
         audio_done = True
 
-        # Use the last decodable audio stream. Streams FFmpeg has no decoder for have no codec context,
-        # and decoding their packets crashes the process. (e.g. APAC spatial-audio track in iPhone)
-        audio_stream = next(
-            (s for s in reversed(container.streams.audio) if s.codec_context is not None),
-            None,
-        )
-        if audio_stream is not None:
+        if len(container.streams.audio):
+            audio_stream = container.streams.audio[-1]
             streams += [audio_stream]
             resampler = av.audio.resampler.AudioResampler(format='fltp')
             audio_done = False
-        elif len(container.streams.audio):
-            logging.warning("No decodable audio stream found in video; ignoring audio.")
 
         for packet in container.demux(*streams):
             if video_done and audio_done:
@@ -464,13 +457,10 @@ class VideoFromFile(VideoInput):
                         else:
                             output_container.metadata[key] = json.dumps(value)
 
-                # Add streams to the new container. Streams with no codec context cannot be used as an output template.
+                # Add streams to the new container
                 stream_map = {}
                 for stream in streams:
                     if isinstance(stream, (av.VideoStream, av.AudioStream, SubtitleStream)):
-                        if stream.codec_context is None:
-                            logging.warning("Skipping %s stream %d with unsupported codec", stream.type, stream.index)
-                            continue
                         out_stream = output_container.add_stream_from_template(template=stream, opaque=True)
                         stream_map[stream] = out_stream
 

comfy_api_nodes/nodes_bytedance.py

@@ -2611,7 +2611,7 @@ class ByteDanceSeedAudioNode(IO.ComfyNode):
         return IO.Schema(
             node_id="ByteDanceSeedAudio",
             display_name="ByteDance Seed Audio 1.0",
-            category="api node/audio/ByteDance",
+            category="partner/audio/ByteDance",
             description=(
                 "Generate speech, music, sound effects and multi-speaker dialogue from a single prompt "
                 "with ByteDance Seed Audio 1.0. Describe the voice(s), emotion, ambience, background music "

comfy_api_nodes/util/upload_helpers.py

@@ -158,14 +158,7 @@ async def upload_video_to_comfyapi(
 
     # Convert VideoInput to BytesIO using specified container/codec
     video_bytes_io = BytesIO()
-    try:
-        video.save_to(video_bytes_io, format=container, codec=codec)
-    except Exception as e:
-        raise ValueError(
-            f"Could not convert the input video to {container.value.upper()} for upload; "
-            f"the file may be corrupt or use an unsupported codec. "
-            f"Try re-exporting it as MP4 (H.264). Original error: {e}"
-        ) from e
+    video.save_to(video_bytes_io, format=container, codec=codec)
     video_bytes_io.seek(0)
 
     return await upload_file_to_comfyapi(cls, video_bytes_io, filename, upload_mime_type, wait_label)

comfy_execution/caching.py

@@ -1,5 +1,6 @@
 import asyncio
 import bisect
+import gc
 import itertools
 import psutil
 import time
@@ -528,6 +529,38 @@ class RAMPressureCache(LRUCache):
         if psutil.virtual_memory().available >= target:
             return
 
+        def remove_cache_key(key):
+            del self.cache[key]
+            self.used_generation.pop(key, None)
+            self.timestamps.pop(key, None)
+            self.children.pop(key, None)
+
+        def has_old_model_patcher(outputs):
+            if outputs is None:
+                return False
+            for output in outputs:
+                if isinstance(output, (list, tuple)):
+                    if has_old_model_patcher(output):
+                        return True
+                elif isinstance(output, ModelPatcher):
+                    return True
+            return False
+
+        old_modelpatcher_keys = []
+        for key, cache_entry in self.cache.items():
+            if self.used_generation[key] == self.generation:
+                continue
+            if has_old_model_patcher(cache_entry.outputs):
+                old_modelpatcher_keys.append(key)
+
+        for key in old_modelpatcher_keys:
+            remove_cache_key(key)
+
+        if old_modelpatcher_keys:
+            gc.collect()
+            if psutil.virtual_memory().available >= target:
+                return
+
         clean_list = []
 
         for key, cache_entry in self.cache.items():
@@ -545,19 +578,17 @@ class RAMPressureCache(LRUCache):
                         scan_list_for_ram_usage(output)
                     elif isinstance(output, torch.Tensor) and output.device.type == 'cpu':
                         ram_usage += output.numel() * output.element_size()
-                    elif isinstance(output, ModelPatcher) and self.used_generation[key] != self.generation:
-                        #old ModelPatchers are the first to go
-                        ram_usage = 1e30
             scan_list_for_ram_usage(cache_entry.outputs)
 
             oom_score *= ram_usage
             #In the case where we have no information on the node ram usage at all,
             #break OOM score ties on the last touch timestamp (pure LRU)
-            bisect.insort(clean_list, (oom_score, self.timestamps[key], key))
+            bisect.insort(clean_list, (oom_score, self.timestamps[key], ram_usage, key))
 
-        while psutil.virtual_memory().available < target and clean_list:
-            _, _, key = clean_list.pop()
-            del self.cache[key]
-            self.used_generation.pop(key, None)
-            self.timestamps.pop(key, None)
-            self.children.pop(key, None)
+        to_free = target - psutil.virtual_memory().available
+        while to_free > 0 and clean_list:
+            _, _, ram_usage, key = clean_list.pop()
+            remove_cache_key(key)
+            to_free -= ram_usage
+
+        gc.collect()

main.py

@@ -314,7 +314,7 @@ def prompt_worker(q, server_instance):
     cache_ram = 0
     cache_ram_inactive = 0
     if not args.cache_classic and not args.cache_none and args.cache_lru <= 0:
-        cache_ram = min(10.0, max(2.0, comfy.model_management.total_ram * 0.10 / 1024.0))
+        cache_ram = min(10.0, max(1.5, comfy.model_management.total_ram * 0.05 / 1024.0))
         cache_ram_inactive = min(96.0, comfy.model_management.total_ram / 1024.0)
         if len(args.cache_ram) > 0:
             cache_ram = args.cache_ram[0]