Merge 45a2363e6a into ffbecfffb9

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

AGENTS.md

@@ -127,6 +127,8 @@
 - Do not add unnecessary `try`/`except` blocks. Use them for optional dependency,
   platform, or backend capability detection only when the program has a useful
   fallback. Prefer specific exception types when changing new code.
+- If a library version is pinned in `requirements.txt`, do not add code to
+  ComfyUI to handle older versions of that library.
 - Remove any workarounds for PyTorch versions that ComfyUI no longer officially
   supports. Deprecated workarounds include catching an exception and rerunning
   the same op with the input cast to float. If a workaround does not have a

README.md

@@ -229,7 +229,7 @@ Python 3.14 works but some custom nodes may have issues. The free threaded varia
 
 Python 3.13 is very well supported. If you have trouble with some custom node dependencies on 3.13 you can try 3.12
 
-torch 2.4 and above is supported but some features and optimizations might only work on newer versions. We generally recommend using the latest major version of pytorch with the latest cuda version unless it is less than 2 weeks old.
+torch 2.5 is minimally supported but using a newer version is extremely recommended. Some features and optimizations might only work on newer versions. We generally recommend using the latest major version of pytorch with the latest cuda version unless it is less than 2 weeks old. If your pytorch is more than 6 months old, please update it.
 
 ### Instructions:
 

comfy/float.py

@@ -70,6 +70,11 @@ def stochastic_rounding(value, dtype, seed=0):
     if dtype == torch.bfloat16:
         return value.to(dtype=torch.bfloat16)
     if dtype == torch.float8_e4m3fn or dtype == torch.float8_e5m2:
+        # MPS does not support FP8 dtypes — perform rounding on CPU and return the result there.
+        on_mps = value.device.type == "mps"
+        if on_mps:
+            value = value.cpu()
+
         generator = torch.Generator(device=value.device)
         generator.manual_seed(seed)
         if _CK_STOCHASTIC_ROUNDING_AVAILABLE:
@@ -178,6 +183,12 @@ def stochastic_round_quantize_nvfp4(x, per_tensor_scale, pad_16x, seed=0):
         """Round up x to the nearest multiple."""
         return ((x + multiple - 1) // multiple) * multiple
 
+    # MPS does not support FP8 dtypes used for block scales — perform on CPU.
+    on_mps = x.device.type == "mps"
+    if on_mps:
+        x = x.cpu()
+        per_tensor_scale = per_tensor_scale.cpu() if isinstance(per_tensor_scale, torch.Tensor) else per_tensor_scale
+
     generator = torch.Generator(device=x.device)
     generator.manual_seed(seed)
 
@@ -198,6 +209,12 @@ def stochastic_round_quantize_nvfp4_by_block(x, per_tensor_scale, pad_16x, seed=
         """Round up x to the nearest multiple."""
         return ((x + multiple - 1) // multiple) * multiple
 
+    # MPS does not support FP8 dtypes used for block scales — perform on CPU.
+    on_mps = x.device.type == "mps"
+    if on_mps:
+        x = x.cpu()
+        per_tensor_scale = per_tensor_scale.cpu() if isinstance(per_tensor_scale, torch.Tensor) else per_tensor_scale
+
     orig_shape = x.shape
 
     # Handle padding

comfy/ldm/ace/ace_step15.py

@@ -217,10 +217,7 @@ class AceStepAttention(nn.Module):
                 cos, sin = position_embeddings
                 query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)
 
-        n_rep = self.num_heads // self.num_kv_heads
-        if n_rep > 1:
-            key_states = key_states.repeat_interleave(n_rep, dim=1)
-            value_states = value_states.repeat_interleave(n_rep, dim=1)
+        gqa_kwargs = {"enable_gqa": True} if self.num_heads != self.num_kv_heads else {}
 
         attn_bias = None
         if self.sliding_window is not None and not self.is_cross_attention:
@@ -244,7 +241,7 @@ class AceStepAttention(nn.Module):
             else:
                 attn_bias = window_bias
 
-        attn_output = optimized_attention(query_states, key_states, value_states, self.num_heads, attn_bias, skip_reshape=True, low_precision_attention=False)
+        attn_output = optimized_attention(query_states, key_states, value_states, self.num_heads, attn_bias, skip_reshape=True, low_precision_attention=False, **gqa_kwargs)
         attn_output = self.o_proj(attn_output)
 
         return attn_output

comfy/ldm/audio/dit.py

@@ -425,19 +425,16 @@ class Attention(nn.Module):
         if n == 1 and causal:
             causal = False
 
-        if h != kv_h:
-            # Repeat interleave kv_heads to match q_heads
-            heads_per_kv_head = h // kv_h
-            k, v = map(lambda t: t.repeat_interleave(heads_per_kv_head, dim = 1), (k, v))
+        gqa_kwargs = {"enable_gqa": True} if h != kv_h else {}
 
         if self.differential:
             q, q_diff = q.unbind(dim=1)
             k, k_diff = k.unbind(dim=1)
-            out      = optimized_attention(q,      k,      v, h, skip_reshape=True, low_precision_attention=False, transformer_options=transformer_options)
-            out_diff = optimized_attention(q_diff, k_diff, v, h, skip_reshape=True, low_precision_attention=False, transformer_options=transformer_options)
+            out      = optimized_attention(q,      k,      v, h, skip_reshape=True, low_precision_attention=False, transformer_options=transformer_options, **gqa_kwargs)
+            out_diff = optimized_attention(q_diff, k_diff, v, h, skip_reshape=True, low_precision_attention=False, transformer_options=transformer_options, **gqa_kwargs)
             out = out - out_diff
         else:
-            out = optimized_attention(q, k, v, h, skip_reshape=True, low_precision_attention=False, transformer_options=transformer_options)
+            out = optimized_attention(q, k, v, h, skip_reshape=True, low_precision_attention=False, transformer_options=transformer_options, **gqa_kwargs)
 
         out = self.to_out(out)
 

comfy/ldm/boogu/model.py

@@ -74,11 +74,8 @@ class BooguDoubleStreamProcessor(nn.Module):
         key = key.transpose(1, 2)
         value = value.transpose(1, 2)
 
-        if attn.kv_heads < attn.heads:
-            key = key.repeat_interleave(attn.heads // attn.kv_heads, dim=1)
-            value = value.repeat_interleave(attn.heads // attn.kv_heads, dim=1)
-
-        hidden_states = optimized_attention_masked(query, key, value, attn.heads, attention_mask, skip_reshape=True, transformer_options=transformer_options)
+        gqa_kwargs = {"enable_gqa": True} if attn.kv_heads < attn.heads else {}
+        hidden_states = optimized_attention_masked(query, key, value, attn.heads, attention_mask, skip_reshape=True, transformer_options=transformer_options, **gqa_kwargs)
 
         # Split back to instruction/image, apply per-stream output projections, recombine.
         instruct_hidden_states = self.instruct_out(hidden_states[:, :L_instruct])

comfy/ldm/modules/attention.py

@@ -1,5 +1,6 @@
 import math
 import sys
+import inspect
 
 import torch
 import torch.nn.functional as F
@@ -14,16 +15,16 @@ from .sub_quadratic_attention import efficient_dot_product_attention
 
 from comfy import model_management
 
-TORCH_HAS_GQA = model_management.torch_version_numeric >= (2, 5)
-
 if model_management.xformers_enabled():
     import xformers
     import xformers.ops
 
 SAGE_ATTENTION_IS_AVAILABLE = False
+SAGE_ATTENTION_SUPPORTS_MASK = False
 try:
     from sageattention import sageattn
     SAGE_ATTENTION_IS_AVAILABLE = True
+    SAGE_ATTENTION_SUPPORTS_MASK = "attn_mask" in inspect.signature(sageattn).parameters
 except ImportError as e:
     if model_management.sage_attention_enabled():
         if e.name == "sageattention":
@@ -89,6 +90,44 @@ def default(val, d):
         return val
     return d
 
+def _gqa_repeat_factor(query_heads, key_heads, value_heads):
+    if key_heads != value_heads:
+        raise ValueError(f"Key/value head count mismatch for GQA: {key_heads} != {value_heads}")
+    if query_heads == key_heads:
+        return 1
+    if query_heads % key_heads != 0:
+        raise ValueError(f"Query heads must be divisible by key/value heads for GQA: {query_heads} vs {key_heads}")
+    return query_heads // key_heads
+
+def _repeat_kv_for_gqa(k, v, query_heads, head_dim):
+    n_rep = _gqa_repeat_factor(query_heads, k.shape[head_dim], v.shape[head_dim])
+    if n_rep > 1:
+        k = k.repeat_interleave(n_rep, dim=head_dim)
+        v = v.repeat_interleave(n_rep, dim=head_dim)
+    return k, v
+
+def _heads_from_dim(tensor, dim_head, name):
+    inner_dim = tensor.shape[-1]
+    if inner_dim % dim_head != 0:
+        raise ValueError(f"{name} inner dimension {inner_dim} is not divisible by head dimension {dim_head}")
+    return inner_dim // dim_head
+
+def _reshape_qkv_to_heads(q, k, v, b, heads, dim_head, enable_gqa=False, expand_kv=True):
+    q = q.unsqueeze(3).reshape(b, -1, heads, dim_head)
+    if enable_gqa:
+        key_heads = _heads_from_dim(k, dim_head, "Key")
+        value_heads = _heads_from_dim(v, dim_head, "Value")
+    else:
+        key_heads = heads
+        value_heads = heads
+    k = k.unsqueeze(3).reshape(b, -1, key_heads, dim_head)
+    v = v.unsqueeze(3).reshape(b, -1, value_heads, dim_head)
+    if enable_gqa:
+        _gqa_repeat_factor(heads, key_heads, value_heads)
+        if expand_kv:
+            k, v = _repeat_kv_for_gqa(k, v, heads, -2)
+    return q, k, v
+
 
 # feedforward
 class GEGLU(nn.Module):
@@ -152,28 +191,19 @@ def attention_basic(q, k, v, heads, mask=None, attn_precision=None, skip_reshape
         b, _, dim_head = q.shape
         dim_head //= heads
 
-    if kwargs.get("enable_gqa", False) and q.shape[-3] != k.shape[-3]:
-        n_rep = q.shape[-3] // k.shape[-3]
-        k = k.repeat_interleave(n_rep, dim=-3)
-        v = v.repeat_interleave(n_rep, dim=-3)
-
     scale = kwargs.get("scale", dim_head ** -0.5)
 
     h = heads
     if skip_reshape:
-         q, k, v = map(
+        if kwargs.get("enable_gqa", False):
+            k, v = _repeat_kv_for_gqa(k, v, q.shape[-3], -3)
+        q, k, v = map(
             lambda t: t.reshape(b * heads, -1, dim_head),
             (q, k, v),
         )
     else:
-        q, k, v = map(
-            lambda t: t.unsqueeze(3)
-            .reshape(b, -1, heads, dim_head)
-            .permute(0, 2, 1, 3)
-            .reshape(b * heads, -1, dim_head)
-            .contiguous(),
-            (q, k, v),
-        )
+        q, k, v = _reshape_qkv_to_heads(q, k, v, b, heads, dim_head, kwargs.get("enable_gqa", False))
+        q, k, v = map(lambda t: t.permute(0, 2, 1, 3).reshape(b * heads, -1, dim_head).contiguous(), (q, k, v))
 
     # force cast to fp32 to avoid overflowing
     if attn_precision == torch.float32:
@@ -231,13 +261,16 @@ def attention_sub_quad(query, key, value, heads, mask=None, attn_precision=None,
         query = query * (kwargs["scale"] * dim_head ** 0.5)
 
     if skip_reshape:
+        if kwargs.get("enable_gqa", False):
+            key, value = _repeat_kv_for_gqa(key, value, query.shape[-3], -3)
         query = query.reshape(b * heads, -1, dim_head)
         value = value.reshape(b * heads, -1, dim_head)
         key = key.reshape(b * heads, -1, dim_head).movedim(1, 2)
     else:
-        query = query.unsqueeze(3).reshape(b, -1, heads, dim_head).permute(0, 2, 1, 3).reshape(b * heads, -1, dim_head)
-        value = value.unsqueeze(3).reshape(b, -1, heads, dim_head).permute(0, 2, 1, 3).reshape(b * heads, -1, dim_head)
-        key = key.unsqueeze(3).reshape(b, -1, heads, dim_head).permute(0, 2, 3, 1).reshape(b * heads, dim_head, -1)
+        query, key, value = _reshape_qkv_to_heads(query, key, value, b, heads, dim_head, kwargs.get("enable_gqa", False))
+        query = query.permute(0, 2, 1, 3).reshape(b * heads, -1, dim_head)
+        value = value.permute(0, 2, 1, 3).reshape(b * heads, -1, dim_head)
+        key = key.permute(0, 2, 3, 1).reshape(b * heads, dim_head, -1)
 
 
     dtype = query.dtype
@@ -304,19 +337,15 @@ def attention_split(q, k, v, heads, mask=None, attn_precision=None, skip_reshape
     scale = kwargs.get("scale", dim_head ** -0.5)
 
     if skip_reshape:
-         q, k, v = map(
+        if kwargs.get("enable_gqa", False):
+            k, v = _repeat_kv_for_gqa(k, v, q.shape[-3], -3)
+        q, k, v = map(
             lambda t: t.reshape(b * heads, -1, dim_head),
             (q, k, v),
         )
     else:
-        q, k, v = map(
-            lambda t: t.unsqueeze(3)
-            .reshape(b, -1, heads, dim_head)
-            .permute(0, 2, 1, 3)
-            .reshape(b * heads, -1, dim_head)
-            .contiguous(),
-            (q, k, v),
-        )
+        q, k, v = _reshape_qkv_to_heads(q, k, v, b, heads, dim_head, kwargs.get("enable_gqa", False))
+        q, k, v = map(lambda t: t.permute(0, 2, 1, 3).reshape(b * heads, -1, dim_head).contiguous(), (q, k, v))
 
     r1 = torch.zeros(q.shape[0], q.shape[1], v.shape[2], device=q.device, dtype=q.dtype)
 
@@ -438,7 +467,7 @@ def attention_xformers(q, k, v, heads, mask=None, attn_precision=None, skip_resh
             disabled_xformers = True
 
     if disabled_xformers:
-        return attention_pytorch(q, k, v, heads, mask, skip_reshape=skip_reshape, **kwargs)
+        return attention_pytorch(q, k, v, heads, mask, skip_reshape=skip_reshape, skip_output_reshape=skip_output_reshape, **kwargs)
 
     if skip_reshape:
         # b h k d -> b k h d
@@ -446,13 +475,12 @@ def attention_xformers(q, k, v, heads, mask=None, attn_precision=None, skip_resh
             lambda t: t.permute(0, 2, 1, 3),
             (q, k, v),
         )
+        if kwargs.get("enable_gqa", False):
+            k, v = _repeat_kv_for_gqa(k, v, q.shape[-2], -2)
     # actually do the reshaping
     else:
         dim_head //= heads
-        q, k, v = map(
-            lambda t: t.reshape(b, -1, heads, dim_head),
-            (q, k, v),
-        )
+        q, k, v = _reshape_qkv_to_heads(q, k, v, b, heads, dim_head, kwargs.get("enable_gqa", False))
 
     if mask is not None:
         # add a singleton batch dimension
@@ -474,7 +502,7 @@ def attention_xformers(q, k, v, heads, mask=None, attn_precision=None, skip_resh
         mask = mask_out[..., :mask.shape[-1]]
         mask = mask.expand(b, heads, -1, -1)
 
-    out = xformers.ops.memory_efficient_attention(q, k, v, attn_bias=mask)
+    out = xformers.ops.memory_efficient_attention(q, k, v, attn_bias=mask, scale=kwargs.get("scale", None))
 
     if skip_output_reshape:
         out = out.permute(0, 2, 1, 3)
@@ -498,10 +526,8 @@ def attention_pytorch(q, k, v, heads, mask=None, attn_precision=None, skip_resha
     else:
         b, _, dim_head = q.shape
         dim_head //= heads
-        q, k, v = map(
-            lambda t: t.view(b, -1, heads, dim_head).transpose(1, 2),
-            (q, k, v),
-        )
+        q, k, v = _reshape_qkv_to_heads(q, k, v, b, heads, dim_head, kwargs.get("enable_gqa", False), expand_kv=False)
+        q, k, v = map(lambda t: t.transpose(1, 2), (q, k, v))
 
     if mask is not None:
         # add a batch dimension if there isn't already one
@@ -511,9 +537,7 @@ def attention_pytorch(q, k, v, heads, mask=None, attn_precision=None, skip_resha
         if mask.ndim == 3:
             mask = mask.unsqueeze(1)
 
-    # Pass through extra SDPA kwargs (scale, enable_gqa) if provided
-    # enable_gqa requires PyTorch 2.5+; older versions use manual KV expansion above
-    sdpa_keys = ("scale", "enable_gqa") if TORCH_HAS_GQA else ("scale",)
+    sdpa_keys = ("scale", "enable_gqa")
     sdpa_extra = {k: v for k, v in kwargs.items() if k in sdpa_keys}
 
     if SDP_BATCH_LIMIT >= b:
@@ -541,20 +565,19 @@ def attention_pytorch(q, k, v, heads, mask=None, attn_precision=None, skip_resha
 
 @wrap_attn
 def attention_sage(q, k, v, heads, mask=None, attn_precision=None, skip_reshape=False, skip_output_reshape=False, **kwargs):
-    if kwargs.get("low_precision_attention", True) is False:
+    if kwargs.get("low_precision_attention", True) is False or (mask is not None and not SAGE_ATTENTION_SUPPORTS_MASK):
         return attention_pytorch(q, k, v, heads, mask=mask, skip_reshape=skip_reshape, skip_output_reshape=skip_output_reshape, **kwargs)
 
     exception_fallback = False
     if skip_reshape:
         b, _, _, dim_head = q.shape
         tensor_layout = "HND"
+        if kwargs.get("enable_gqa", False):
+            k, v = _repeat_kv_for_gqa(k, v, q.shape[-3], -3)
     else:
         b, _, dim_head = q.shape
         dim_head //= heads
-        q, k, v = map(
-            lambda t: t.view(b, -1, heads, dim_head),
-            (q, k, v),
-        )
+        q, k, v = _reshape_qkv_to_heads(q, k, v, b, heads, dim_head, kwargs.get("enable_gqa", False))
         tensor_layout = "NHD"
 
     if mask is not None:
@@ -565,8 +588,12 @@ def attention_sage(q, k, v, heads, mask=None, attn_precision=None, skip_reshape=
         if mask.ndim == 3:
             mask = mask.unsqueeze(1)
 
+    sage_kwargs = {"is_causal": False, "tensor_layout": tensor_layout, "sm_scale": kwargs.get("scale", None), "smooth_k": False}
+    if mask is not None:
+        sage_kwargs["attn_mask"] = mask
+
     try:
-        out = sageattn(q, k, v, attn_mask=mask, is_causal=False, tensor_layout=tensor_layout)
+        out = sageattn(q, k, v, **sage_kwargs)
     except Exception as e:
         logging.error("Error running sage attention: {}, using pytorch attention instead.".format(e))
         exception_fallback = True
@@ -616,7 +643,6 @@ def attention3_sage(q, k, v, heads, mask=None, attn_precision=None, skip_reshape
                 skip_output_reshape=skip_output_reshape,
                 **kwargs
             )
-        q_s, k_s, v_s = q, k, v
         N = q.shape[2]
         dim_head = D
     else:
@@ -642,11 +668,15 @@ def attention3_sage(q, k, v, heads, mask=None, attn_precision=None, skip_reshape
                 **kwargs
             )
 
-    if not skip_reshape:
-        q_s, k_s, v_s = map(
-            lambda t: t.view(B, -1, heads, dim_head).permute(0, 2, 1, 3).contiguous(),
-            (q, k, v),
-        )
+    if skip_reshape:
+        q_s = q
+        if kwargs.get("enable_gqa", False):
+            k_s, v_s = _repeat_kv_for_gqa(k, v, H, -3)
+        else:
+            k_s, v_s = k, v
+    else:
+        q_s, k_s, v_s = _reshape_qkv_to_heads(q, k, v, B, heads, dim_head, kwargs.get("enable_gqa", False))
+        q_s, k_s, v_s = map(lambda t: t.permute(0, 2, 1, 3).contiguous(), (q_s, k_s, v_s))
         B, H, L, D = q_s.shape
 
     try:
@@ -662,7 +692,7 @@ def attention3_sage(q, k, v, heads, mask=None, attn_precision=None, skip_reshape
                 q, k, v, heads,
                 mask=mask,
                 attn_precision=attn_precision,
-                skip_reshape=False,
+                skip_reshape=skip_reshape,
                 skip_output_reshape=skip_output_reshape,
                 **kwargs
             )
@@ -681,19 +711,20 @@ def attention3_sage(q, k, v, heads, mask=None, attn_precision=None, skip_reshape
 try:
     @torch.library.custom_op("flash_attention::flash_attn", mutates_args=())
     def flash_attn_wrapper(q: torch.Tensor, k: torch.Tensor, v: torch.Tensor,
-                    dropout_p: float = 0.0, causal: bool = False) -> torch.Tensor:
-        return flash_attn_func(q, k, v, dropout_p=dropout_p, causal=causal)
+                    dropout_p: float = 0.0, causal: bool = False, softmax_scale: float = -1.0) -> torch.Tensor:
+        softmax_scale_arg = None if softmax_scale == -1.0 else softmax_scale
+        return flash_attn_func(q, k, v, dropout_p=dropout_p, causal=causal, softmax_scale=softmax_scale_arg)
 
 
     @flash_attn_wrapper.register_fake
-    def flash_attn_fake(q, k, v, dropout_p=0.0, causal=False):
+    def flash_attn_fake(q, k, v, dropout_p=0.0, causal=False, softmax_scale=-1.0):
         # Output shape is the same as q
         return q.new_empty(q.shape)
 except AttributeError as error:
     FLASH_ATTN_ERROR = error
 
     def flash_attn_wrapper(q: torch.Tensor, k: torch.Tensor, v: torch.Tensor,
-                    dropout_p: float = 0.0, causal: bool = False) -> torch.Tensor:
+                    dropout_p: float = 0.0, causal: bool = False, softmax_scale: float = -1.0) -> torch.Tensor:
         assert False, f"Could not define flash_attn_wrapper: {FLASH_ATTN_ERROR}"
 
 @wrap_attn
@@ -703,10 +734,8 @@ def attention_flash(q, k, v, heads, mask=None, attn_precision=None, skip_reshape
     else:
         b, _, dim_head = q.shape
         dim_head //= heads
-        q, k, v = map(
-            lambda t: t.view(b, -1, heads, dim_head).transpose(1, 2),
-            (q, k, v),
-        )
+        q, k, v = _reshape_qkv_to_heads(q, k, v, b, heads, dim_head, kwargs.get("enable_gqa", False), expand_kv=False)
+        q, k, v = map(lambda t: t.transpose(1, 2), (q, k, v))
 
     if mask is not None:
         # add a batch dimension if there isn't already one
@@ -725,10 +754,16 @@ def attention_flash(q, k, v, heads, mask=None, attn_precision=None, skip_reshape
             v.transpose(1, 2),
             dropout_p=0.0,
             causal=False,
+            softmax_scale=kwargs.get("scale", -1.0),
         ).transpose(1, 2)
     except Exception as e:
         logging.warning(f"Flash Attention failed, using default SDPA: {e}")
-        out = torch.nn.functional.scaled_dot_product_attention(q, k, v, attn_mask=mask, dropout_p=0.0, is_causal=False)
+        sdpa_extra = {}
+        if kwargs.get("enable_gqa", False):
+            sdpa_extra["enable_gqa"] = True
+        if "scale" in kwargs:
+            sdpa_extra["scale"] = kwargs["scale"]
+        out = torch.nn.functional.scaled_dot_product_attention(q, k, v, attn_mask=mask, dropout_p=0.0, is_causal=False, **sdpa_extra)
     if not skip_output_reshape:
         out = (
             out.transpose(1, 2).reshape(b, -1, heads * dim_head)
@@ -1209,5 +1244,3 @@ class SpatialVideoTransformer(SpatialTransformer):
             x = self.proj_out(x)
         out = x + x_in
         return out
-
-

comfy/ldm/omnigen/omnigen2.py

@@ -141,11 +141,8 @@ class Attention(nn.Module):
         key = key.transpose(1, 2)
         value = value.transpose(1, 2)
 
-        if self.kv_heads < self.heads:
-            key = key.repeat_interleave(self.heads // self.kv_heads, dim=1)
-            value = value.repeat_interleave(self.heads // self.kv_heads, dim=1)
-
-        hidden_states = optimized_attention_masked(query, key, value, self.heads, attention_mask, skip_reshape=True, transformer_options=transformer_options)
+        gqa_kwargs = {"enable_gqa": True} if self.kv_heads < self.heads else {}
+        hidden_states = optimized_attention_masked(query, key, value, self.heads, attention_mask, skip_reshape=True, transformer_options=transformer_options, **gqa_kwargs)
         hidden_states = self.to_out[0](hidden_states)
         return hidden_states
 

comfy/ops.py

@@ -174,6 +174,8 @@ def cast_modules_with_vbar(comfy_modules, dtype, device, bias_dtype, non_blockin
                     elif xfer_dest2 is not None:
                         xfer_source.prepare(xfer_dest2, stream, copy=True, commit=False)
                         return
+                    else:
+                        return
                 comfy.model_management.cast_to_gathered(xfer_source, xfer_dest, non_blocking=non_blocking, stream=stream, r2=xfer_dest2)
 
         def handle_pin(m, pin, source, dest, subset="weights", size=None):

comfy/quant_ops.py

@@ -97,6 +97,12 @@ class _TensorCoreFP8LayoutBase(_CKFp8Layout):
         if not isinstance(scale, torch.Tensor):
             scale = torch.tensor(scale, device=tensor.device, dtype=torch.float32)
 
+        # MPS does not support FP8 dtypes — move to CPU for quantization.
+        on_mps = tensor.device.type == "mps"
+        if on_mps:
+            tensor = tensor.cpu()
+            scale = scale.cpu()
+
         if stochastic_rounding > 0:
             if inplace_ops:
                 tensor *= (1.0 / scale).to(tensor.dtype)

comfy/text_encoders/gpt_oss.py

@@ -12,7 +12,7 @@ import torch.nn.functional as F
 
 import comfy.ops
 from comfy import sd1_clip
-from comfy.ldm.modules.attention import TORCH_HAS_GQA, optimized_attention_for_device
+from comfy.ldm.modules.attention import optimized_attention_for_device
 from comfy.text_encoders.llama import RMSNorm, apply_rope
 
 
@@ -110,10 +110,6 @@ def _attention_with_sinks(q: torch.Tensor, k: torch.Tensor, v: torch.Tensor, sin
     putting the sink logit in the mask at that column.
     """
 
-    if num_kv_groups > 1 and not TORCH_HAS_GQA:
-        k = k.repeat_interleave(num_kv_groups, dim=1)
-        v = v.repeat_interleave(num_kv_groups, dim=1)
-
     B, _, S_q, D = q.shape
     H_kv = k.shape[1]
     S_kv = k.shape[-2]

comfy/text_encoders/llama.py

@@ -550,10 +550,8 @@ class Attention(nn.Module):
                 xv = xv[:, :, -sliding_window:]
                 attention_mask = attention_mask[..., -sliding_window:] if attention_mask is not None else None
 
-        xk = xk.repeat_interleave(self.num_heads // self.num_kv_heads, dim=1)
-        xv = xv.repeat_interleave(self.num_heads // self.num_kv_heads, dim=1)
-
-        output = optimized_attention(xq, xk, xv, self.num_heads, mask=attention_mask, skip_reshape=True)
+        gqa_kwargs = {"enable_gqa": True} if self.num_heads != self.num_kv_heads else {}
+        output = optimized_attention(xq, xk, xv, self.num_heads, mask=attention_mask, skip_reshape=True, **gqa_kwargs)
         return self.o_proj(output), present_key_value
 
 class MLP(nn.Module):

comfy/text_encoders/qwen35.py

@@ -366,12 +366,8 @@ class GatedAttention(nn.Module):
                     xv = torch.cat((past_value[:, :, :index], xv), dim=2)
                 present_key_value = (xk, xv, index + num_tokens)
 
-        # Expand KV heads for GQA
-        if self.num_heads != self.num_kv_heads:
-            xk = xk.repeat_interleave(self.num_heads // self.num_kv_heads, dim=1)
-            xv = xv.repeat_interleave(self.num_heads // self.num_kv_heads, dim=1)
-
-        output = optimized_attention(xq, xk, xv, self.num_heads, mask=attention_mask, skip_reshape=True)
+        gqa_kwargs = {"enable_gqa": True} if self.num_heads != self.num_kv_heads else {}
+        output = optimized_attention(xq, xk, xv, self.num_heads, mask=attention_mask, skip_reshape=True, **gqa_kwargs)
         output = output * gate.sigmoid()
 
         return self.o_proj(output), present_key_value

tests/test_fp8_mps.py

@@ -0,0 +1,147 @@
+"""
+Tests for FP8 quantization on MPS (Apple Silicon) devices.
+
+MPS does not natively support float8_e4m3fn or float8_e5m2 dtypes.
+These tests verify that:
+  1. FP8 operations correctly fall back to CPU when on MPS.
+  2. The round-trip (quantize on CPU -> result on original device) is numerically sound.
+  3. No "Placeholder storage has not been allocated on MPS device!" errors occur.
+"""
+import pytest
+import torch
+
+import comfy.float
+from comfy.quant_ops import TensorCoreFP8E4M3Layout, TensorCoreFP8E5M2Layout
+
+# Skip the entire module if MPS is not available
+pytestmark = pytest.mark.skipif(
+    not torch.backends.mps.is_available(),
+    reason="MPS backend not available"
+)
+
+# ── helpers ──────────────────────────────────────────────────────────────────
+
+def _make_mps_tensor(shape=(256, 256), dtype=torch.float32):
+    return torch.randn(shape, device="mps", dtype=dtype)
+
+
+# ── Tests for comfy.float ────────────────────────────────────────────────────
+
+class TestStochasticRoundingMPS:
+    """Tests for comfy.float.stochastic_rounding on MPS device."""
+
+    def test_stochastic_rounding_fp8_e4m3fn_on_mps(self):
+        """stochastic_rounding must not crash when input is on MPS and target dtype is float8_e4m3fn."""
+        x = _make_mps_tensor((64, 64), dtype=torch.float32)
+        result = comfy.float.stochastic_rounding(x, dtype=torch.float8_e4m3fn, seed=42)
+
+        assert result.dtype == torch.float8_e4m3fn
+        assert result.shape == x.shape
+
+    def test_stochastic_rounding_fp8_e5m2_on_mps(self):
+        """stochastic_rounding must not crash when input is on MPS and target dtype is float8_e5m2."""
+        x = _make_mps_tensor((64, 64), dtype=torch.float32)
+        result = comfy.float.stochastic_rounding(x, dtype=torch.float8_e5m2, seed=42)
+
+        assert result.dtype == torch.float8_e5m2
+        assert result.shape == x.shape
+
+    def test_stochastic_rounding_fp8_result_on_cpu(self):
+        """Result of FP8 rounding from MPS input should be on CPU (since MPS can't hold FP8)."""
+        x = _make_mps_tensor((32, 32), dtype=torch.float32)
+        result = comfy.float.stochastic_rounding(x, dtype=torch.float8_e4m3fn, seed=42)
+
+        # FP8 tensors cannot live on MPS, so result must be on CPU
+        assert result.device.type == "cpu"
+
+    def test_stochastic_rounding_non_fp8_still_works(self):
+        """Non-FP8 dtypes on MPS must still work as before (no regression)."""
+        x = _make_mps_tensor((32, 32), dtype=torch.float32)
+
+        r16 = comfy.float.stochastic_rounding(x, dtype=torch.float16, seed=0)
+        assert r16.dtype == torch.float16
+        assert r16.device.type == "mps"
+
+        rbf16 = comfy.float.stochastic_rounding(x, dtype=torch.bfloat16, seed=0)
+        assert rbf16.dtype == torch.bfloat16
+        assert rbf16.device.type == "mps"
+
+    def test_stochastic_rounding_fp8_numerical_sanity(self):
+        """FP8 round-trip (float32 -> fp8 -> float32) should have bounded error."""
+        x = torch.randn(128, 128, device="mps", dtype=torch.float32)
+        x_clamped = torch.clamp(x, min=-448, max=448)  # FP8 e4m3fn range
+
+        fp8 = comfy.float.stochastic_rounding(x_clamped, dtype=torch.float8_e4m3fn, seed=123)
+        # Convert back to float32 for comparison
+        reconstructed = fp8.to(torch.float32)
+
+        # Max relative error should be bounded (FP8 e4m3fn has ~0.125 relative precision)
+        x_cpu = x_clamped.cpu()
+        max_abs_err = (reconstructed - x_cpu).abs().max().item()
+        # FP8 e4m3fn max value is 448, min subnormal ~0.001953
+        # For random normal data, error should be well under 1.0
+        assert max_abs_err < 2.0, f"FP8 round-trip error too large: {max_abs_err}"
+
+
+class TestManualStochasticRoundMPS:
+    """Tests for comfy.float.manual_stochastic_round_to_float8 on MPS device."""
+
+    def test_manual_round_fp8_on_mps_tensor(self):
+        """stochastic_rounding handles MPS generator internally without 'Placeholder storage' error."""
+        x = _make_mps_tensor((16, 16), dtype=torch.float32)
+        result = comfy.float.stochastic_rounding(x, dtype=torch.float8_e4m3fn, seed=42)
+        assert result.dtype == torch.float8_e4m3fn
+
+
+class TestNVFP4StochasticRoundMPS:
+    """Tests for NVFP4 stochastic rounding on MPS - also creates FP8 tensors internally."""
+
+    def test_nvfp4_stochastic_round_on_mps(self):
+        """stochastic_round_quantize_nvfp4 creates FP8 block scales internally."""
+        # NVFP4 requires 2D input with dimensions divisible by 16
+        x = torch.randn(32, 32, device="mps", dtype=torch.float32)
+        scale = torch.tensor(1.0, device="mps", dtype=torch.float32)
+
+        # This should not crash - internally creates float8_e4m3fn block scales
+        qdata, block_scale = comfy.float.stochastic_round_quantize_nvfp4(
+            x, scale, pad_16x=False, seed=42
+        )
+        assert qdata.dtype == torch.uint8
+
+
+# ── Tests for comfy.quant_ops (integration) ──────────────────────────────────
+
+class TestQuantOpsMPS:
+    """Tests for the quantization ops layer that calls into comfy.float."""
+
+    def test_fp8_layout_quantize_on_mps(self):
+        """TensorCoreFP8E4M3Layout.quantize must work with MPS tensors."""
+        x = _make_mps_tensor((64, 64), dtype=torch.bfloat16)
+        qdata, params = TensorCoreFP8E4M3Layout.quantize(
+            x, scale="recalculate", stochastic_rounding=42
+        )
+
+        assert qdata.dtype == torch.float8_e4m3fn
+        assert params.orig_dtype == torch.bfloat16
+
+    def test_fp8_layout_quantize_without_stochastic_on_mps(self):
+        """TensorCoreFP8E4M3Layout.quantize with stochastic_rounding=0 uses ck.quantize_per_tensor_fp8."""
+        x = _make_mps_tensor((64, 64), dtype=torch.bfloat16)
+        qdata, params = TensorCoreFP8E4M3Layout.quantize(
+            x, scale="recalculate", stochastic_rounding=0
+        )
+
+        assert qdata.dtype == torch.float8_e4m3fn
+
+    def test_fp8_e5m2_layout_quantize_on_mps(self):
+        """TensorCoreFP8E5M2Layout.quantize must work with MPS tensors."""
+        x = _make_mps_tensor((64, 64), dtype=torch.float32)
+        qdata, params = TensorCoreFP8E5M2Layout.quantize(
+            x, scale="recalculate", stochastic_rounding=42
+        )
+
+        assert qdata.dtype == torch.float8_e5m2
+
+
+if __name__ == "__main__":
+    pytest.main([__file__, "-v", "--tb=short"])