Merge bb31f8b707 into b08debceca

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

.coderabbit.yaml

@@ -4,12 +4,12 @@ early_access: false
 tone_instructions: "Only comment on issues introduced by this PR's changes. Do not flag pre-existing problems in moved, re-indented, or reformatted code."
 
 reviews:
-  profile: "chill"
-  request_changes_workflow: false
+  profile: "assertive"
+  request_changes_workflow: true
   high_level_summary: false
   poem: false
   review_status: false
-  review_details: false
+  review_details: true
   commit_status: true
   collapse_walkthrough: true
   changed_files_summary: false
@@ -39,6 +39,14 @@ reviews:
     - path: "**"
       instructions: |
         IMPORTANT: Only comment on issues directly introduced by this PR's code changes.
+        Treat AGENTS.md as mandatory repository policy, not optional style guidance.
+        Flag PR changes that violate AGENTS.md even when the code is otherwise functional.
+        In particular, enforce architecture boundaries, dtype/device/memory rules,
+        interface contracts, import style, no unnecessary try/except blocks, no inline
+        imports, no outbound internet paths in core ComfyUI, and narrow scoped fixes.
+        Prefer direct findings over suggestions when a rule is violated. Only ignore
+        AGENTS.md when it clearly conflicts with a newer explicit maintainer instruction
+        in the PR.
         Do NOT flag pre-existing issues in code that was merely moved, re-indented,
         de-indented, or reformatted without logic changes. If code appears in the diff
         only due to whitespace or structural reformatting (e.g., removing a `with:` block),
@@ -123,5 +131,10 @@ chat:
 
 knowledge_base:
   opt_out: false
+  code_guidelines:
+    enabled: true
+    filePatterns:
+      - files: "AGENTS.md"
+        applyTo: "**"
   learnings:
     scope: "auto"

CLAUDE.md

@@ -0,0 +1 @@
+AGENTS.md

comfy/model_management.py

@@ -1860,7 +1860,21 @@ def supports_fp8_compute(device=None):
     if SUPPORT_FP8_OPS:
         return True
 
-    if not is_nvidia():
+    if device is None:
+        device = get_torch_device()
+
+    if is_device_cpu(device) or is_device_mps(device):
+        return False
+
+    # Per-device check instead of the global is_nvidia(). On ROCm builds,
+    # is_device_cuda() returns True (AMD GPUs appear as cuda:N via HIP) but
+    # torch.version.cuda is None, so this correctly returns False for AMD.
+    # If PyTorch ever supports mixed-vendor GPUs in one process, these
+    # per-device checks remain correct unlike the global is_nvidia().
+    if not is_device_cuda(device):
+        return False
+
+    if not torch.version.cuda:
         return False
 
     props = torch.cuda.get_device_properties(device)
@@ -1881,7 +1895,10 @@ def supports_fp8_compute(device=None):
     return True
 
 def supports_nvfp4_compute(device=None):
-    if not is_nvidia():
+    if device is None:
+        device = get_torch_device()
+
+    if not is_device_cuda(device) or not torch.version.cuda:
         return False
 
     props = torch.cuda.get_device_properties(device)

comfy/sd1_clip.py

@@ -543,18 +543,24 @@ class SDTokenizer:
     def _try_get_embedding(self, embedding_name:str):
         '''
         Takes a potential embedding name and tries to retrieve it.
-        Returns a Tuple consisting of the embedding and any leftover string, embedding can be None.
+        Returns a Tuple consisting of the embedding, the cleaned embedding name, and any leftover string, embedding can be None.
         '''
         split_embed = embedding_name.split()
         embedding_name = split_embed[0]
         leftover = ' '.join(split_embed[1:])
+
+        match = re.search(r'[<\[]', embedding_name)
+        if match is not None:
+            leftover = embedding_name[match.start():] + (" " + leftover if leftover else "")
+            embedding_name = embedding_name[:match.start()]
+
         embed = load_embed(embedding_name, self.embedding_directory, self.embedding_size, self.embedding_key)
         if embed is None:
             stripped = embedding_name.strip(',')
             if len(stripped) < len(embedding_name):
                 embed = load_embed(stripped, self.embedding_directory, self.embedding_size, self.embedding_key)
-                return (embed, "{} {}".format(embedding_name[len(stripped):], leftover))
-        return (embed, leftover)
+                return (embed, embedding_name, "{} {}".format(embedding_name[len(stripped):], leftover))
+        return (embed, embedding_name, leftover)
 
     def pad_tokens(self, tokens, amount):
         if self.pad_left:
@@ -585,7 +591,7 @@ class SDTokenizer:
         tokens = []
         for weighted_segment, weight in parsed_weights:
             to_tokenize = unescape_important(weighted_segment)
-            split = re.split(' {0}|\n{0}'.format(self.embedding_identifier), to_tokenize)
+            split = re.split(r'(?<=\s){}'.format(re.escape(self.embedding_identifier)), to_tokenize)
             to_tokenize = [split[0]]
             for i in range(1, len(split)):
                 to_tokenize.append("{}{}".format(self.embedding_identifier, split[i]))
@@ -595,7 +601,7 @@ class SDTokenizer:
                 # if we find an embedding, deal with the embedding
                 if word.startswith(self.embedding_identifier) and self.embedding_directory is not None:
                     embedding_name = word[len(self.embedding_identifier):].strip('\n')
-                    embed, leftover = self._try_get_embedding(embedding_name)
+                    embed, embedding_name, leftover = self._try_get_embedding(embedding_name)
                     if embed is None:
                         logging.warning(f"warning, embedding:{embedding_name} does not exist, ignoring")
                     else:

comfy/text_encoders/llama.py

@@ -937,22 +937,41 @@ class BaseGenerate:
             return torch.argmax(logits, dim=-1, keepdim=True)
 
         # Sampling mode
-        if repetition_penalty != 1.0:
-            for i in range(logits.shape[0]):
-                for token_id in set(token_history):
-                    logits[i, token_id] *= repetition_penalty if logits[i, token_id] < 0 else 1/repetition_penalty
-
-        if presence_penalty is not None and presence_penalty != 0.0:
-            for i in range(logits.shape[0]):
-                for token_id in set(token_history):
-                    logits[i, token_id] -= presence_penalty
+        if len(token_history) > 0 and (repetition_penalty != 1.0 or (presence_penalty is not None and presence_penalty != 0.0)):
+            token_ids = torch.tensor(list(set(token_history)), device=logits.device)
+            token_logits = logits[:, token_ids]
+            if repetition_penalty != 1.0:
+                token_logits = torch.where(token_logits < 0, token_logits * repetition_penalty, token_logits / repetition_penalty)
+            if presence_penalty is not None and presence_penalty != 0.0:
+                token_logits = token_logits - presence_penalty
+            logits[:, token_ids] = token_logits
 
         if temperature != 1.0:
             logits = logits / temperature
 
         if top_k > 0:
-            indices_to_remove = logits < torch.topk(logits, top_k)[0][..., -1, None]
-            logits[indices_to_remove] = torch.finfo(logits.dtype).min
+            top_k = min(top_k, logits.shape[-1])
+            logits, top_indices = torch.topk(logits, top_k)
+
+            if min_p > 0.0:
+                probs_before_filter = torch.nn.functional.softmax(logits, dim=-1)
+                top_probs, _ = probs_before_filter.max(dim=-1, keepdim=True)
+                min_threshold = min_p * top_probs
+                indices_to_remove = probs_before_filter < min_threshold
+                logits[indices_to_remove] = torch.finfo(logits.dtype).min
+
+            if top_p < 1.0:
+                sorted_logits, sorted_indices = torch.sort(logits, descending=True)
+                cumulative_probs = torch.cumsum(torch.nn.functional.softmax(sorted_logits, dim=-1), dim=-1)
+                sorted_indices_to_remove = cumulative_probs > top_p
+                sorted_indices_to_remove[..., 0] = False
+                indices_to_remove = torch.zeros_like(logits, dtype=torch.bool)
+                indices_to_remove.scatter_(1, sorted_indices, sorted_indices_to_remove)
+                logits[indices_to_remove] = torch.finfo(logits.dtype).min
+
+            probs = torch.nn.functional.softmax(logits, dim=-1)
+            next_token = torch.multinomial(probs, num_samples=1, generator=generator)
+            return top_indices.gather(1, next_token)
 
         if min_p > 0.0:
             probs_before_filter = torch.nn.functional.softmax(logits, dim=-1)

comfy_api_nodes/util/request_logger.py

@@ -9,6 +9,7 @@ from typing import Any
 import folder_paths
 
 logger = logging.getLogger(__name__)
+_SENSITIVE_HEADERS = {"authorization", "x-api-key"}
 
 
 def get_log_directory():
@@ -73,6 +74,10 @@ def _format_data_for_logging(data: Any) -> str:
     return str(data)
 
 
+def _redact_headers(headers: dict) -> dict:
+    return {k: ("***" if k.lower() in _SENSITIVE_HEADERS else v) for k, v in headers.items()}
+
+
 def log_request_response(
     operation_id: str,
     request_method: str,
@@ -101,7 +106,7 @@ def log_request_response(
         log_content.append(f"Method: {request_method}")
         log_content.append(f"URL: {request_url}")
         if request_headers:
-            log_content.append(f"Headers:\n{_format_data_for_logging(request_headers)}")
+            log_content.append(f"Headers:\n{_format_data_for_logging(_redact_headers(request_headers))}")
         if request_params:
             log_content.append(f"Params:\n{_format_data_for_logging(request_params)}")
         if request_data is not None:

comfy_extras/nodes_color.py

@@ -16,23 +16,30 @@ class ColorToRGBInt(io.ComfyNode):
             ],
             outputs=[
                 io.Int.Output(display_name="rgb_int"),
-                io.Color.Output(display_name="hex")
+                io.Color.Output(display_name="hex"),
+                io.Float.Output(display_name="alpha"),
             ],
         )
 
     @classmethod
     def execute(cls, color: str) -> io.NodeOutput:
-        # expect format #RRGGBB
-        if len(color) != 7 or color[0] != "#":
-            raise ValueError("Color must be in format #RRGGBB")
+        # expect format #RRGGBB or #RRGGBBAA
+        if len(color) not in (7, 9) or color[0] != "#":
+            raise ValueError("Color must be in format #RRGGBB or #RRGGBBAA")
         try:
             int(color[1:], 16)
         except ValueError:
-            raise ValueError("Color must be in format #RRGGBB") from None
+            raise ValueError("Color must be in format #RRGGBB or #RRGGBBAA") from None
+
+        alpha = 1.0
+        if len(color) == 9:
+            alpha = int(color[7:9], 16) / 255.0
+            color = color[:7]
+
         r, g, b = hex_to_rgb(color)
 
         rgb_int = r * 256 * 256 + g * 256 + b
-        return io.NodeOutput(rgb_int, color)
+        return io.NodeOutput(rgb_int, color, alpha)
 
 
 class ColorExtension(ComfyExtension):

requirements.txt

@@ -1,6 +1,6 @@
 comfyui-frontend-package==1.45.20
 comfyui-workflow-templates==0.11.2
-comfyui-embedded-docs==0.5.6
+comfyui-embedded-docs==0.5.7
 torch
 torchsde
 torchvision

tests-unit/comfy_test/model_management_test.py

@@ -0,0 +1,109 @@
+import pytest
+from unittest.mock import patch, MagicMock
+import torch
+
+import comfy.model_management as mm
+
+
+class FakeDeviceProps:
+    """Minimal stand-in for torch.cuda.get_device_properties return value."""
+    def __init__(self, major, minor, name="FakeGPU"):
+        self.major = major
+        self.minor = minor
+        self.name = name
+
+
+class TestSupportsFp8Compute:
+    """Tests for per-device fp8 compute capability detection."""
+
+    def test_cpu_device_returns_false(self):
+        assert mm.supports_fp8_compute(torch.device("cpu")) is False
+
+    @pytest.mark.skipif(not hasattr(torch.backends, "mps"), reason="MPS backend not available")
+    def test_mps_device_returns_false(self):
+        assert mm.supports_fp8_compute(torch.device("mps")) is False
+
+    @patch("comfy.model_management.SUPPORT_FP8_OPS", True)
+    def test_cli_override_returns_true(self):
+        assert mm.supports_fp8_compute(torch.device("cpu")) is True
+
+    @patch("comfy.model_management.get_torch_device", return_value=torch.device("cpu"))
+    def test_none_device_defaults_to_get_torch_device(self, mock_get):
+        result = mm.supports_fp8_compute(None)
+        mock_get.assert_called_once()
+        assert result is False
+
+    @pytest.mark.skipif(not torch.cuda.is_available(), reason="CUDA not available")
+    def test_each_cuda_device_checked_independently(self):
+        """On a multi-GPU system, each device should be queried for its own capabilities."""
+        count = torch.cuda.device_count()
+        if count < 2:
+            pytest.skip("Need 2+ CUDA devices for multi-GPU test")
+        results = {}
+        for i in range(count):
+            dev = torch.device(f"cuda:{i}")
+            results[i] = mm.supports_fp8_compute(dev)
+            props = torch.cuda.get_device_properties(dev)
+            # Verify the result is consistent with the device's compute capability
+            if props.major >= 9:
+                assert results[i] is True, f"cuda:{i} ({props.name}) has SM {props.major}.{props.minor}, should support fp8"
+            elif props.major < 8 or props.minor < 9:
+                assert results[i] is False, f"cuda:{i} ({props.name}) has SM {props.major}.{props.minor}, should not support fp8"
+
+    @patch("torch.version.cuda", None)
+    @patch("comfy.model_management.SUPPORT_FP8_OPS", False)
+    def test_rocm_build_returns_false(self):
+        """On ROCm, devices appear as cuda:N via HIP but torch.version.cuda is None."""
+        dev = MagicMock()
+        dev.type = "cuda"
+        assert mm.supports_fp8_compute(dev) is False
+
+    @patch("torch.version.cuda", "12.4")
+    @patch("comfy.model_management.SUPPORT_FP8_OPS", False)
+    @patch("torch.cuda.get_device_properties")
+    def test_sm89_supports_fp8(self, mock_props):
+        """Ada Lovelace (SM 8.9, e.g. RTX 4080) should support fp8."""
+        mock_props.return_value = FakeDeviceProps(major=8, minor=9)
+        dev = torch.device("cuda:0")
+        assert mm.supports_fp8_compute(dev) is True
+
+    @patch("torch.version.cuda", "12.4")
+    @patch("comfy.model_management.SUPPORT_FP8_OPS", False)
+    @patch("torch.cuda.get_device_properties")
+    def test_sm86_does_not_support_fp8(self, mock_props):
+        """Ampere (SM 8.6, e.g. RTX 3090) should not support fp8."""
+        mock_props.return_value = FakeDeviceProps(major=8, minor=6)
+        dev = torch.device("cuda:0")
+        assert mm.supports_fp8_compute(dev) is False
+
+    @patch("torch.version.cuda", "12.4")
+    @patch("comfy.model_management.SUPPORT_FP8_OPS", False)
+    @patch("torch.cuda.get_device_properties")
+    def test_sm90_supports_fp8(self, mock_props):
+        """Hopper (SM 9.0) and above should support fp8."""
+        mock_props.return_value = FakeDeviceProps(major=9, minor=0)
+        dev = torch.device("cuda:0")
+        assert mm.supports_fp8_compute(dev) is True
+
+
+class TestSupportsNvfp4Compute:
+    """Tests for per-device nvfp4 compute capability detection."""
+
+    def test_cpu_device_returns_false(self):
+        assert mm.supports_nvfp4_compute(torch.device("cpu")) is False
+
+    @patch("torch.version.cuda", "12.4")
+    @patch("torch.cuda.get_device_properties")
+    def test_sm100_supports_nvfp4(self, mock_props):
+        """Blackwell (SM 10.0) should support nvfp4."""
+        mock_props.return_value = FakeDeviceProps(major=10, minor=0)
+        dev = torch.device("cuda:0")
+        assert mm.supports_nvfp4_compute(dev) is True
+
+    @patch("torch.version.cuda", "12.4")
+    @patch("torch.cuda.get_device_properties")
+    def test_sm89_does_not_support_nvfp4(self, mock_props):
+        """Ada Lovelace (SM 8.9) should not support nvfp4."""
+        mock_props.return_value = FakeDeviceProps(major=8, minor=9)
+        dev = torch.device("cuda:0")
+        assert mm.supports_nvfp4_compute(dev) is False