[pull] master from Comfy-Org:master

app/assets/scanner.py

@@ -33,6 +33,7 @@
     verify_file_unchanged,
 )
 from app.assets.services.hashing import HashCheckpoint, compute_blake3_hash
+from app.assets.services.image_dimensions import extract_image_dimensions
 from app.assets.services.metadata_extract import extract_file_metadata
 from app.assets.services.path_utils import (
     compute_relative_filename,
@@ -506,6 +507,10 @@ def enrich_asset(
 
     if extract_metadata and metadata:
         system_metadata = metadata.to_user_metadata()
+        if mime_type and mime_type.startswith("image/"):
+            dims = extract_image_dimensions(file_path, mime_type=mime_type)
+            if dims:
+                system_metadata.update(dims)
         set_reference_system_metadata(session, reference_id, system_metadata)
 
     if full_hash:

app/assets/services/image_dimensions.py

@@ -0,0 +1,63 @@
+"""Image dimension extraction for asset ingest.
+
+Reads only the image header via Pillow to capture width/height cheaply,
+without a full pixel decode. Returns a metadata dict suitable for merging
+into ``AssetReference.system_metadata``.
+"""
+from __future__ import annotations
+
+import logging
+from typing import Any
+
+logger = logging.getLogger(__name__)
+
+
+def extract_image_dimensions(
+    file_path: str, mime_type: str | None = None
+) -> dict[str, Any] | None:
+    """Extract image dimensions for the file at ``file_path``.
+
+    Args:
+        file_path: Absolute path to a file on disk.
+        mime_type: Optional MIME type hint. When provided and not prefixed
+            with ``image/``, extraction is skipped without touching the file.
+
+    Returns:
+        ``{"kind": "image", "width": W, "height": H}`` when the file is a
+        recognizable image with positive dimensions, otherwise ``None``.
+
+    The dict shape is intended to be merged into ``system_metadata`` so the
+    asset response surfaces ``metadata.kind`` plus dimension fields for image
+    assets. Forward-compatible: future media kinds (e.g. ``"video"`` with
+    duration/fps) can extend this shape without schema changes.
+    """
+    if mime_type is not None and not mime_type.startswith("image/"):
+        return None
+
+    try:
+        from PIL import Image, UnidentifiedImageError
+    except ImportError:
+        logger.debug(
+            "Pillow not available; skipping image dimension extraction for %s",
+            file_path,
+        )
+        return None
+
+    try:
+        with Image.open(file_path) as img:
+            width, height = img.size
+    except (OSError, UnidentifiedImageError, ValueError) as exc:
+        logger.debug(
+            "Failed to read image dimensions from %s: %s", file_path, exc
+        )
+        return None
+
+    if (
+        not isinstance(width, int)
+        or not isinstance(height, int)
+        or width <= 0
+        or height <= 0
+    ):
+        return None
+
+    return {"kind": "image", "width": width, "height": height}

app/assets/services/ingest.py

@@ -17,9 +17,11 @@
     get_reference_by_file_path,
     get_reference_tags,
     get_or_create_reference,
+    list_references_by_asset_id,
     reference_exists,
     remove_missing_tag_for_asset_id,
     set_reference_metadata,
+    set_reference_system_metadata,
     set_reference_tags,
     update_asset_hash_and_mime,
     upsert_asset,
@@ -29,6 +31,7 @@
 from app.assets.helpers import get_utc_now, normalize_tags
 from app.assets.services.bulk_ingest import batch_insert_seed_assets
 from app.assets.services.file_utils import get_size_and_mtime_ns
+from app.assets.services.image_dimensions import extract_image_dimensions
 from app.assets.services.path_utils import (
     compute_relative_filename,
     get_name_and_tags_from_asset_path,
@@ -118,6 +121,14 @@ def _ingest_file_from_path(
                 user_metadata=user_metadata,
             )
 
+            _maybe_store_image_dimensions(
+                session,
+                reference_id=reference_id,
+                file_path=locator,
+                mime_type=mime_type,
+                current_system_metadata=ref.system_metadata,
+            )
+
         try:
             remove_missing_tag_for_asset_id(session, asset_id=asset.id)
         except Exception:
@@ -288,6 +299,13 @@ def _register_existing_asset(
                 user_metadata=new_meta,
             )
 
+        _backfill_image_dimensions_from_siblings(
+            session,
+            asset_id=asset.id,
+            new_reference_id=ref.id,
+            current_system_metadata=ref.system_metadata,
+        )
+
         if tags is not None:
             set_reference_tags(
                 session,
@@ -334,6 +352,87 @@ def _update_metadata_with_filename(
         )
 
 
+_IMAGE_DIMENSION_KEYS = ("kind", "width", "height")
+
+
+def _maybe_store_image_dimensions(
+    session: Session,
+    reference_id: str,
+    file_path: str,
+    mime_type: str | None,
+    current_system_metadata: dict | None,
+) -> None:
+    """Populate ``kind``/``width``/``height`` on system_metadata for image refs.
+
+    Non-image MIME types are a no-op. Pre-existing keys (e.g. enricher-written
+    safetensors metadata, download provenance) are preserved by merge.
+    """
+    if not mime_type or not mime_type.startswith("image/"):
+        return
+
+    dims = extract_image_dimensions(file_path, mime_type=mime_type)
+    if not dims:
+        return
+
+    current = current_system_metadata or {}
+    merged = dict(current)
+    merged.update(dims)
+    if merged != current:
+        set_reference_system_metadata(
+            session,
+            reference_id=reference_id,
+            system_metadata=merged,
+        )
+
+
+def _backfill_image_dimensions_from_siblings(
+    session: Session,
+    asset_id: str,
+    new_reference_id: str,
+    current_system_metadata: dict | None,
+) -> None:
+    """Copy image dimension keys from any sibling reference of the same asset.
+
+    The from-hash path doesn't read the file bytes, so dimensions can't be
+    extracted there directly. When another reference of the same asset already
+    carries image dimensions, copy them onto the new reference so consumers
+    see consistent metadata regardless of how the asset was registered.
+
+    Best-effort: missing siblings, non-image siblings, or absent dimension
+    keys leave the target reference unchanged.
+    """
+    current = current_system_metadata or {}
+    if current.get("kind") == "image" and "width" in current and "height" in current:
+        return
+
+    for sibling in list_references_by_asset_id(session, asset_id):
+        if sibling.id == new_reference_id:
+            continue
+        meta = sibling.system_metadata or {}
+        if meta.get("kind") != "image":
+            continue
+        width = meta.get("width")
+        height = meta.get("height")
+        if (
+            type(width) is not int
+            or type(height) is not int
+            or width <= 0
+            or height <= 0
+        ):
+            continue
+        merged = dict(current)
+        merged["kind"] = "image"
+        merged["width"] = width
+        merged["height"] = height
+        if merged != current:
+            set_reference_system_metadata(
+                session,
+                reference_id=new_reference_id,
+                system_metadata=merged,
+            )
+        return
+
+
 def _sanitize_filename(name: str | None, fallback: str) -> str:
     n = os.path.basename((name or "").strip() or fallback)
     return n if n else fallback

comfy/latent_formats.py

@@ -4,7 +4,6 @@ class LatentFormat:
     scale_factor = 1.0
     latent_channels = 4
     latent_dimensions = 2
-    preserve_empty_channel_multiples = False
     latent_rgb_factors = None
     latent_rgb_factors_bias = None
     latent_rgb_factors_reshape = None
@@ -780,10 +779,6 @@ class ACEAudio(LatentFormat):
     latent_channels = 8
     latent_dimensions = 2
 
-class SeedVR2(LatentFormat):
-    latent_channels = 16
-    preserve_empty_channel_multiples = True
-
 class ACEAudio15(LatentFormat):
     latent_channels = 64
     latent_dimensions = 1

comfy/ldm/modules/attention.py

@@ -735,86 +735,7 @@ def attention_flash(q, k, v, heads, mask=None, attn_precision=None, skip_reshape
         )
     return out
 
-def _var_attention_qkv(q, k, v, heads, skip_reshape):
-    if skip_reshape:
-        return q, k, v, q.shape[-1]
-    total_tokens, embed_dim = q.shape
-    head_dim = embed_dim // heads
-    return (
-        q.view(total_tokens, heads, head_dim),
-        k.view(k.shape[0], heads, head_dim),
-        v.view(v.shape[0], heads, head_dim),
-        head_dim,
-    )
 
-
-def _var_attention_output(out, heads, head_dim, skip_output_reshape):
-    if skip_output_reshape:
-        return out
-    return out.reshape(-1, heads * head_dim)
-
-
-def _use_blackwell_attention():
-    device = model_management.get_torch_device()
-    if device.type != "cuda":
-        return False
-    major, minor = torch.cuda.get_device_capability(device)
-    return (major, minor) >= (12, 0)
-
-
-def _validate_split_cu_seqlens(name, cu_seqlens, token_count):
-    if cu_seqlens.dtype not in (torch.int32, torch.int64):
-        raise ValueError(f"{name} must use an integer dtype")
-    if cu_seqlens.ndim != 1 or cu_seqlens.numel() < 2:
-        raise ValueError(f"{name} must be a 1D tensor with at least two offsets")
-    if cu_seqlens[0].item() != 0:
-        raise ValueError(f"{name} must start at 0")
-    if (cu_seqlens[1:] <= cu_seqlens[:-1]).any().item():
-        raise ValueError(f"{name} must be strictly increasing")
-    if cu_seqlens[-1].item() != token_count:
-        raise ValueError(f"{name} does not match token count")
-
-
-def _split_indices(cu_seqlens):
-    return cu_seqlens[1:-1].to(device="cpu", dtype=torch.long)
-
-
-def var_attention_optimized_split(q, k, v, heads, cu_seqlens_q, cu_seqlens_k, *args, skip_reshape=False, skip_output_reshape=False, **kwargs):
-    q, k, v, head_dim = _var_attention_qkv(q, k, v, heads, skip_reshape)
-
-    _validate_split_cu_seqlens("cu_seqlens_q", cu_seqlens_q, q.shape[0])
-    _validate_split_cu_seqlens("cu_seqlens_k", cu_seqlens_k, k.shape[0])
-    if cu_seqlens_k[-1].item() != v.shape[0]:
-        raise ValueError("cu_seqlens_k does not match v token count")
-
-    q_split_indices = _split_indices(cu_seqlens_q)
-    k_split_indices = _split_indices(cu_seqlens_k)
-    q_splits = torch.tensor_split(q, q_split_indices, dim=0)
-    k_splits = torch.tensor_split(k, k_split_indices, dim=0)
-    v_splits = torch.tensor_split(v, k_split_indices, dim=0)
-    if len(q_splits) != len(k_splits) or len(q_splits) != len(v_splits):
-        raise ValueError("cu_seqlens_q and cu_seqlens_k must describe the same sequence count")
-
-    out = []
-    for q_i, k_i, v_i in zip(q_splits, k_splits, v_splits):
-        q_i = q_i.permute(1, 0, 2).unsqueeze(0)
-        k_i = k_i.permute(1, 0, 2).unsqueeze(0)
-        v_i = v_i.permute(1, 0, 2).unsqueeze(0)
-        out_dtype = q_i.dtype
-        if optimized_attention is attention_sage and q_i.dtype not in (torch.float16, torch.bfloat16):
-            q_i = q_i.to(torch.bfloat16)
-            k_i = k_i.to(torch.bfloat16)
-            v_i = v_i.to(torch.bfloat16)
-        out_i = optimized_attention(q_i, k_i, v_i, heads, skip_reshape=True, skip_output_reshape=True)
-        if out_i.dtype != out_dtype:
-            out_i = out_i.to(out_dtype)
-        out.append(out_i.squeeze(0).permute(1, 0, 2))
-
-    out = torch.cat(out, dim=0)
-    return _var_attention_output(out, heads, head_dim, skip_output_reshape)
-
-
-optimized_var_attention = var_attention_optimized_split
 optimized_attention = attention_basic
 
 if model_management.sage_attention_enabled():
@@ -837,8 +758,6 @@ def var_attention_optimized_split(q, k, v, heads, cu_seqlens_q, cu_seqlens_k, *a
         logging.info("Using sub quadratic optimization for attention, if you have memory or speed issues try using: --use-split-cross-attention")
         optimized_attention = attention_sub_quad
 
-logging.info("Using optimized_attention split-loop for variable-length attention")
-
 optimized_attention_masked = optimized_attention
 
 
@@ -854,7 +773,6 @@ def var_attention_optimized_split(q, k, v, heads, cu_seqlens_q, cu_seqlens_k, *a
 register_attention_function("pytorch", attention_pytorch)
 register_attention_function("sub_quad", attention_sub_quad)
 register_attention_function("split", attention_split)
-register_attention_function("var_attention_optimized_split", var_attention_optimized_split)
 
 
 def optimized_attention_for_device(device, mask=False, small_input=False):
@@ -1291,3 +1209,5 @@ def forward(
             x = self.proj_out(x)
         out = x + x_in
         return out
+
+

comfy/ldm/modules/diffusionmodules/model.py

@@ -13,7 +13,6 @@
     import xformers
     import xformers.ops
 
-
 def torch_cat_if_needed(xl, dim):
     xl = [x for x in xl if x is not None and x.shape[dim] > 0]
     if len(xl) > 1:
@@ -23,8 +22,7 @@ def torch_cat_if_needed(xl, dim):
     else:
         return None
 
-
-def get_timestep_embedding(timesteps, embedding_dim, flip_sin_to_cos=False, downscale_freq_shift=1):
+def get_timestep_embedding(timesteps, embedding_dim):
     """
     This matches the implementation in Denoising Diffusion Probabilistic Models:
     From Fairseq.
@@ -35,13 +33,11 @@ def get_timestep_embedding(timesteps, embedding_dim, flip_sin_to_cos=False, down
     assert len(timesteps.shape) == 1
 
     half_dim = embedding_dim // 2
-    emb = math.log(10000) / (half_dim - downscale_freq_shift)
+    emb = math.log(10000) / (half_dim - 1)
     emb = torch.exp(torch.arange(half_dim, dtype=torch.float32) * -emb)
     emb = emb.to(device=timesteps.device)
     emb = timesteps.float()[:, None] * emb[None, :]
     emb = torch.cat([torch.sin(emb), torch.cos(emb)], dim=1)
-    if flip_sin_to_cos:
-        emb = torch.cat([emb[:, half_dim:], emb[:, :half_dim]], dim=-1)
     if embedding_dim % 2 == 1:  # zero pad
         emb = torch.nn.functional.pad(emb, (0,1,0,0))
     return emb