Fix typo in error message tips for CLIP nodes

comfy/ldm/chroma/layers.py

@@ -3,6 +3,7 @@ from torch import Tensor, nn
 
 from comfy.ldm.flux.layers import (
     MLPEmbedder,
+    RMSNorm,
     ModulationOut,
 )
 
@@ -28,7 +29,7 @@ class Approximator(nn.Module):
         super().__init__()
         self.in_proj = operations.Linear(in_dim, hidden_dim, bias=True, dtype=dtype, device=device)
         self.layers = nn.ModuleList([MLPEmbedder(hidden_dim, hidden_dim, dtype=dtype, device=device, operations=operations) for x in range( n_layers)])
-        self.norms = nn.ModuleList([operations.RMSNorm(hidden_dim, dtype=dtype, device=device) for x in range( n_layers)])
+        self.norms = nn.ModuleList([RMSNorm(hidden_dim, dtype=dtype, device=device, operations=operations) for x in range( n_layers)])
         self.out_proj = operations.Linear(hidden_dim, out_dim, dtype=dtype, device=device)
 
     @property

comfy/ldm/chroma_radiance/layers.py

@@ -4,6 +4,8 @@ from functools import lru_cache
 import torch
 from torch import nn
 
+from comfy.ldm.flux.layers import RMSNorm
+
 
 class NerfEmbedder(nn.Module):
     """
@@ -143,7 +145,7 @@ class NerfGLUBlock(nn.Module):
         # We now need to generate parameters for 3 matrices.
         total_params = 3 * hidden_size_x**2 * mlp_ratio
         self.param_generator = operations.Linear(hidden_size_s, total_params, dtype=dtype, device=device)
-        self.norm = operations.RMSNorm(hidden_size_x, dtype=dtype, device=device)
+        self.norm = RMSNorm(hidden_size_x, dtype=dtype, device=device, operations=operations)
         self.mlp_ratio = mlp_ratio
 
 
@@ -176,7 +178,7 @@ class NerfGLUBlock(nn.Module):
 class NerfFinalLayer(nn.Module):
     def __init__(self, hidden_size, out_channels, dtype=None, device=None, operations=None):
         super().__init__()
-        self.norm = operations.RMSNorm(hidden_size, dtype=dtype, device=device)
+        self.norm = RMSNorm(hidden_size, dtype=dtype, device=device, operations=operations)
         self.linear = operations.Linear(hidden_size, out_channels, dtype=dtype, device=device)
 
     def forward(self, x: torch.Tensor) -> torch.Tensor:
@@ -188,7 +190,7 @@ class NerfFinalLayer(nn.Module):
 class NerfFinalLayerConv(nn.Module):
     def __init__(self, hidden_size: int, out_channels: int, dtype=None, device=None, operations=None):
         super().__init__()
-        self.norm = operations.RMSNorm(hidden_size, dtype=dtype, device=device)
+        self.norm = RMSNorm(hidden_size, dtype=dtype, device=device, operations=operations)
         self.conv = operations.Conv2d(
             in_channels=hidden_size,
             out_channels=out_channels,

comfy/ldm/flux/layers.py

@@ -5,6 +5,8 @@ import torch
 from torch import Tensor, nn
 
 from .math import attention, rope
+import comfy.ops
+import comfy.ldm.common_dit
 
 
 class EmbedND(nn.Module):
@@ -85,12 +87,20 @@ def build_mlp(hidden_size, mlp_hidden_dim, mlp_silu_act=False, yak_mlp=False, dt
             operations.Linear(mlp_hidden_dim, hidden_size, bias=True, dtype=dtype, device=device),
         )
 
+class RMSNorm(torch.nn.Module):
+    def __init__(self, dim: int, dtype=None, device=None, operations=None):
+        super().__init__()
+        self.scale = nn.Parameter(torch.empty((dim), dtype=dtype, device=device))
+
+    def forward(self, x: Tensor):
+        return comfy.ldm.common_dit.rms_norm(x, self.scale, 1e-6)
+
 
 class QKNorm(torch.nn.Module):
     def __init__(self, dim: int, dtype=None, device=None, operations=None):
         super().__init__()
-        self.query_norm = operations.RMSNorm(dim, dtype=dtype, device=device)
-        self.key_norm = operations.RMSNorm(dim, dtype=dtype, device=device)
+        self.query_norm = RMSNorm(dim, dtype=dtype, device=device, operations=operations)
+        self.key_norm = RMSNorm(dim, dtype=dtype, device=device, operations=operations)
 
     def forward(self, q: Tensor, k: Tensor, v: Tensor) -> tuple:
         q = self.query_norm(q)
@@ -159,7 +169,7 @@ class SiLUActivation(nn.Module):
 
 
 class DoubleStreamBlock(nn.Module):
-    def __init__(self, hidden_size: int, num_heads: int, mlp_ratio: float, qkv_bias: bool = False, modulation=True, mlp_silu_act=False, proj_bias=True, yak_mlp=False, dtype=None, device=None, operations=None):
+    def __init__(self, hidden_size: int, num_heads: int, mlp_ratio: float, qkv_bias: bool = False, flipped_img_txt=False, modulation=True, mlp_silu_act=False, proj_bias=True, yak_mlp=False, dtype=None, device=None, operations=None):
         super().__init__()
 
         mlp_hidden_dim = int(hidden_size * mlp_ratio)
@@ -187,6 +197,8 @@ class DoubleStreamBlock(nn.Module):
 
         self.txt_mlp = build_mlp(hidden_size, mlp_hidden_dim, mlp_silu_act=mlp_silu_act, yak_mlp=yak_mlp, dtype=dtype, device=device, operations=operations)
 
+        self.flipped_img_txt = flipped_img_txt
+
     def forward(self, img: Tensor, txt: Tensor, vec: Tensor, pe: Tensor, attn_mask=None, modulation_dims_img=None, modulation_dims_txt=None, transformer_options={}):
         if self.modulation:
             img_mod1, img_mod2 = self.img_mod(vec)
@@ -212,17 +224,32 @@ class DoubleStreamBlock(nn.Module):
         del txt_qkv
         txt_q, txt_k = self.txt_attn.norm(txt_q, txt_k, txt_v)
 
-        q = torch.cat((txt_q, img_q), dim=2)
-        del txt_q, img_q
-        k = torch.cat((txt_k, img_k), dim=2)
-        del txt_k, img_k
-        v = torch.cat((txt_v, img_v), dim=2)
-        del txt_v, img_v
-        # run actual attention
-        attn = attention(q, k, v, pe=pe, mask=attn_mask, transformer_options=transformer_options)
-        del q, k, v
+        if self.flipped_img_txt:
+            q = torch.cat((img_q, txt_q), dim=2)
+            del img_q, txt_q
+            k = torch.cat((img_k, txt_k), dim=2)
+            del img_k, txt_k
+            v = torch.cat((img_v, txt_v), dim=2)
+            del img_v, txt_v
+            # run actual attention
+            attn = attention(q, k, v,
+                             pe=pe, mask=attn_mask, transformer_options=transformer_options)
+            del q, k, v
 
-        txt_attn, img_attn = attn[:, : txt.shape[1]], attn[:, txt.shape[1]:]
+            img_attn, txt_attn = attn[:, : img.shape[1]], attn[:, img.shape[1]:]
+        else:
+            q = torch.cat((txt_q, img_q), dim=2)
+            del txt_q, img_q
+            k = torch.cat((txt_k, img_k), dim=2)
+            del txt_k, img_k
+            v = torch.cat((txt_v, img_v), dim=2)
+            del txt_v, img_v
+            # run actual attention
+            attn = attention(q, k, v,
+                             pe=pe, mask=attn_mask, transformer_options=transformer_options)
+            del q, k, v
+
+            txt_attn, img_attn = attn[:, : txt.shape[1]], attn[:, txt.shape[1]:]
 
         # calculate the img bloks
         img += apply_mod(self.img_attn.proj(img_attn), img_mod1.gate, None, modulation_dims_img)

comfy/ldm/flux/model.py

@@ -16,6 +16,7 @@ from .layers import (
     SingleStreamBlock,
     timestep_embedding,
     Modulation,
+    RMSNorm
 )
 
 @dataclass
@@ -80,7 +81,7 @@ class Flux(nn.Module):
         self.txt_in = operations.Linear(params.context_in_dim, self.hidden_size, bias=params.ops_bias, dtype=dtype, device=device)
 
         if params.txt_norm:
-            self.txt_norm = operations.RMSNorm(params.context_in_dim, dtype=dtype, device=device)
+            self.txt_norm = RMSNorm(params.context_in_dim, dtype=dtype, device=device, operations=operations)
         else:
             self.txt_norm = None
 

comfy/ldm/hunyuan_video/model.py

@@ -241,6 +241,7 @@ class HunyuanVideo(nn.Module):
                     self.num_heads,
                     mlp_ratio=params.mlp_ratio,
                     qkv_bias=params.qkv_bias,
+                    flipped_img_txt=True,
                     dtype=dtype, device=device, operations=operations
                 )
                 for _ in range(params.depth)
@@ -377,14 +378,14 @@ class HunyuanVideo(nn.Module):
             extra_txt_ids = torch.zeros((txt_ids.shape[0], txt_vision_states.shape[1], txt_ids.shape[-1]), device=txt_ids.device, dtype=txt_ids.dtype)
             txt_ids = torch.cat((txt_ids, extra_txt_ids), dim=1)
 
-        ids = torch.cat((txt_ids, img_ids), dim=1)
+        ids = torch.cat((img_ids, txt_ids), dim=1)
         pe = self.pe_embedder(ids)
 
         img_len = img.shape[1]
         if txt_mask is not None:
             attn_mask_len = img_len + txt.shape[1]
             attn_mask = torch.zeros((1, 1, attn_mask_len), dtype=img.dtype, device=img.device)
-            attn_mask[:, 0, :txt.shape[1]] = txt_mask
+            attn_mask[:, 0, img_len:] = txt_mask
         else:
             attn_mask = None
 
@@ -412,7 +413,7 @@ class HunyuanVideo(nn.Module):
                     if add is not None:
                         img += add
 
-        img = torch.cat((txt, img), 1)
+        img = torch.cat((img, txt), 1)
 
         transformer_options["total_blocks"] = len(self.single_blocks)
         transformer_options["block_type"] = "single"
@@ -434,9 +435,9 @@ class HunyuanVideo(nn.Module):
                 if i < len(control_o):
                     add = control_o[i]
                     if add is not None:
-                        img[:, txt.shape[1]: img_len + txt.shape[1]] += add
+                        img[:, : img_len] += add
 
-        img = img[:, txt.shape[1]: img_len + txt.shape[1]]
+        img = img[:, : img_len]
         if ref_latent is not None:
             img = img[:, ref_latent.shape[1]:]
 

comfy/model_detection.py

@@ -19,12 +19,6 @@ def count_blocks(state_dict_keys, prefix_string):
         count += 1
     return count
 
-def any_suffix_in(keys, prefix, main, suffix_list=[]):
-    for x in suffix_list:
-        if "{}{}{}".format(prefix, main, x) in keys:
-            return True
-    return False
-
 def calculate_transformer_depth(prefix, state_dict_keys, state_dict):
     context_dim = None
     use_linear_in_transformer = False
@@ -192,7 +186,7 @@ def detect_unet_config(state_dict, key_prefix, metadata=None):
             dit_config["meanflow_sum"] = False
         return dit_config
 
-    if any_suffix_in(state_dict_keys, key_prefix, 'double_blocks.0.img_attn.norm.key_norm.', ["weight", "scale"]) and ('{}img_in.weight'.format(key_prefix) in state_dict_keys or any_suffix_in(state_dict_keys, key_prefix, 'distilled_guidance_layer.norms.0.', ["weight", "scale"])): #Flux, Chroma or Chroma Radiance (has no img_in.weight)
+    if '{}double_blocks.0.img_attn.norm.key_norm.scale'.format(key_prefix) in state_dict_keys and ('{}img_in.weight'.format(key_prefix) in state_dict_keys or f"{key_prefix}distilled_guidance_layer.norms.0.scale" in state_dict_keys): #Flux, Chroma or Chroma Radiance (has no img_in.weight)
         dit_config = {}
         if '{}double_stream_modulation_img.lin.weight'.format(key_prefix) in state_dict_keys:
             dit_config["image_model"] = "flux2"
@@ -247,8 +241,7 @@ def detect_unet_config(state_dict, key_prefix, metadata=None):
 
         dit_config["depth"] = count_blocks(state_dict_keys, '{}double_blocks.'.format(key_prefix) + '{}.')
         dit_config["depth_single_blocks"] = count_blocks(state_dict_keys, '{}single_blocks.'.format(key_prefix) + '{}.')
-
-        if any_suffix_in(state_dict_keys, key_prefix, 'distilled_guidance_layer.0.norms.0.', ["weight", "scale"]) or any_suffix_in(state_dict_keys, key_prefix, 'distilled_guidance_layer.norms.0.', ["weight", "scale"]): #Chroma
+        if '{}distilled_guidance_layer.0.norms.0.scale'.format(key_prefix) in state_dict_keys or '{}distilled_guidance_layer.norms.0.scale'.format(key_prefix) in state_dict_keys: #Chroma
             dit_config["image_model"] = "chroma"
             dit_config["in_channels"] = 64
             dit_config["out_channels"] = 64
@@ -256,8 +249,7 @@ def detect_unet_config(state_dict, key_prefix, metadata=None):
             dit_config["out_dim"] = 3072
             dit_config["hidden_dim"] = 5120
             dit_config["n_layers"] = 5
-
-            if any_suffix_in(state_dict_keys, key_prefix, 'nerf_blocks.0.norm.', ["weight", "scale"]): #Chroma Radiance
+            if f"{key_prefix}nerf_blocks.0.norm.scale" in state_dict_keys: #Chroma Radiance
                 dit_config["image_model"] = "chroma_radiance"
                 dit_config["in_channels"] = 3
                 dit_config["out_channels"] = 3
@@ -267,7 +259,7 @@ def detect_unet_config(state_dict, key_prefix, metadata=None):
                 dit_config["nerf_depth"] = 4
                 dit_config["nerf_max_freqs"] = 8
                 dit_config["nerf_tile_size"] = 512
-                dit_config["nerf_final_head_type"] = "conv" if any_suffix_in(state_dict_keys, key_prefix, 'nerf_final_layer_conv.norm.', ["weight", "scale"]) else "linear"
+                dit_config["nerf_final_head_type"] = "conv" if f"{key_prefix}nerf_final_layer_conv.norm.scale" in state_dict_keys else "linear"
                 dit_config["nerf_embedder_dtype"] = torch.float32
                 if "{}__x0__".format(key_prefix) in state_dict_keys: # x0 pred
                     dit_config["use_x0"] = True
@@ -276,7 +268,7 @@ def detect_unet_config(state_dict, key_prefix, metadata=None):
         else:
             dit_config["guidance_embed"] = "{}guidance_in.in_layer.weight".format(key_prefix) in state_dict_keys
             dit_config["yak_mlp"] = '{}double_blocks.0.img_mlp.gate_proj.weight'.format(key_prefix) in state_dict_keys
-            dit_config["txt_norm"] = any_suffix_in(state_dict_keys, key_prefix, 'txt_norm.', ["weight", "scale"])
+            dit_config["txt_norm"] = "{}txt_norm.scale".format(key_prefix) in state_dict_keys
             if dit_config["yak_mlp"] and dit_config["txt_norm"]:  # Ovis model
                 dit_config["txt_ids_dims"] = [1, 2]
 

comfy/supported_models.py

@@ -710,15 +710,6 @@ class Flux(supported_models_base.BASE):
 
     supported_inference_dtypes = [torch.bfloat16, torch.float16, torch.float32]
 
-    def process_unet_state_dict(self, state_dict):
-        out_sd = {}
-        for k in list(state_dict.keys()):
-            key_out = k
-            if key_out.endswith("_norm.scale"):
-                key_out = "{}.weight".format(key_out[:-len(".scale")])
-            out_sd[key_out] = state_dict[k]
-        return out_sd
-
     vae_key_prefix = ["vae."]
     text_encoder_key_prefix = ["text_encoders."]
 
@@ -907,13 +898,11 @@ class HunyuanVideo(supported_models_base.BASE):
             key_out = key_out.replace("txt_in.c_embedder.linear_1.", "txt_in.c_embedder.in_layer.").replace("txt_in.c_embedder.linear_2.", "txt_in.c_embedder.out_layer.")
             key_out = key_out.replace("_mod.linear.", "_mod.lin.").replace("_attn_qkv.", "_attn.qkv.")
             key_out = key_out.replace("mlp.fc1.", "mlp.0.").replace("mlp.fc2.", "mlp.2.")
-            key_out = key_out.replace("_attn_q_norm.weight", "_attn.norm.query_norm.weight").replace("_attn_k_norm.weight", "_attn.norm.key_norm.weight")
-            key_out = key_out.replace(".q_norm.weight", ".norm.query_norm.weight").replace(".k_norm.weight", ".norm.key_norm.weight")
+            key_out = key_out.replace("_attn_q_norm.weight", "_attn.norm.query_norm.scale").replace("_attn_k_norm.weight", "_attn.norm.key_norm.scale")
+            key_out = key_out.replace(".q_norm.weight", ".norm.query_norm.scale").replace(".k_norm.weight", ".norm.key_norm.scale")
             key_out = key_out.replace("_attn_proj.", "_attn.proj.")
             key_out = key_out.replace(".modulation.linear.", ".modulation.lin.")
             key_out = key_out.replace("_in.mlp.2.", "_in.out_layer.").replace("_in.mlp.0.", "_in.in_layer.")
-            if key_out.endswith(".scale"):
-                key_out = "{}.weight".format(key_out[:-len(".scale")])
             out_sd[key_out] = state_dict[k]
         return out_sd
 
@@ -1275,15 +1264,6 @@ class Hunyuan3Dv2(supported_models_base.BASE):
 
     latent_format = latent_formats.Hunyuan3Dv2
 
-    def process_unet_state_dict(self, state_dict):
-        out_sd = {}
-        for k in list(state_dict.keys()):
-            key_out = k
-            if key_out.endswith(".scale"):
-                key_out = "{}.weight".format(key_out[:-len(".scale")])
-            out_sd[key_out] = state_dict[k]
-        return out_sd
-
     def process_unet_state_dict_for_saving(self, state_dict):
         replace_prefix = {"": "model."}
         return utils.state_dict_prefix_replace(state_dict, replace_prefix)
@@ -1361,14 +1341,6 @@ class Chroma(supported_models_base.BASE):
 
     supported_inference_dtypes = [torch.bfloat16, torch.float16, torch.float32]
 
-    def process_unet_state_dict(self, state_dict):
-        out_sd = {}
-        for k in list(state_dict.keys()):
-            key_out = k
-            if key_out.endswith(".scale"):
-                key_out = "{}.weight".format(key_out[:-len(".scale")])
-            out_sd[key_out] = state_dict[k]
-        return out_sd
 
     def get_model(self, state_dict, prefix="", device=None):
         out = model_base.Chroma(self, device=device)

comfy/utils.py

@@ -675,10 +675,10 @@ def flux_to_diffusers(mmdit_config, output_prefix=""):
                         "ff_context.linear_in.bias": "txt_mlp.0.bias",
                         "ff_context.linear_out.weight": "txt_mlp.2.weight",
                         "ff_context.linear_out.bias": "txt_mlp.2.bias",
-                        "attn.norm_q.weight": "img_attn.norm.query_norm.weight",
-                        "attn.norm_k.weight": "img_attn.norm.key_norm.weight",
-                        "attn.norm_added_q.weight": "txt_attn.norm.query_norm.weight",
-                        "attn.norm_added_k.weight": "txt_attn.norm.key_norm.weight",
+                        "attn.norm_q.weight": "img_attn.norm.query_norm.scale",
+                        "attn.norm_k.weight": "img_attn.norm.key_norm.scale",
+                        "attn.norm_added_q.weight": "txt_attn.norm.query_norm.scale",
+                        "attn.norm_added_k.weight": "txt_attn.norm.key_norm.scale",
                     }
 
         for k in block_map:
@@ -701,8 +701,8 @@ def flux_to_diffusers(mmdit_config, output_prefix=""):
                         "norm.linear.bias": "modulation.lin.bias",
                         "proj_out.weight": "linear2.weight",
                         "proj_out.bias": "linear2.bias",
-                        "attn.norm_q.weight": "norm.query_norm.weight",
-                        "attn.norm_k.weight": "norm.key_norm.weight",
+                        "attn.norm_q.weight": "norm.query_norm.scale",
+                        "attn.norm_k.weight": "norm.key_norm.scale",
                         "attn.to_qkv_mlp_proj.weight": "linear1.weight", # Flux 2
                         "attn.to_out.weight": "linear2.weight", # Flux 2
                     }

comfy_api_nodes/util/client.py

@@ -57,7 +57,6 @@ class _RequestConfig:
     files: dict[str, Any] | list[tuple[str, Any]] | None
     multipart_parser: Callable | None
     max_retries: int
-    max_retries_on_rate_limit: int
     retry_delay: float
     retry_backoff: float
     wait_label: str = "Waiting"
@@ -66,7 +65,6 @@ class _RequestConfig:
     final_label_on_success: str | None = "Completed"
     progress_origin_ts: float | None = None
     price_extractor: Callable[[dict[str, Any]], float | None] | None = None
-    is_rate_limited: Callable[[int, Any], bool] | None = None
 
 
 @dataclass
@@ -80,7 +78,7 @@ class _PollUIState:
     active_since: float | None = None  # start time of current active interval (None if queued)
 
 
-_RETRY_STATUS = {408, 500, 502, 503, 504}  # status 429 is handled separately
+_RETRY_STATUS = {408, 429, 500, 502, 503, 504}
 COMPLETED_STATUSES = ["succeeded", "succeed", "success", "completed", "finished", "done", "complete"]
 FAILED_STATUSES = ["cancelled", "canceled", "canceling", "fail", "failed", "error"]
 QUEUED_STATUSES = ["created", "queued", "queueing", "submitted", "initializing"]
@@ -105,8 +103,6 @@ async def sync_op(
     final_label_on_success: str | None = "Completed",
     progress_origin_ts: float | None = None,
     monitor_progress: bool = True,
-    max_retries_on_rate_limit: int = 16,
-    is_rate_limited: Callable[[int, Any], bool] | None = None,
 ) -> M:
     raw = await sync_op_raw(
         cls,
@@ -126,8 +122,6 @@ async def sync_op(
         final_label_on_success=final_label_on_success,
         progress_origin_ts=progress_origin_ts,
         monitor_progress=monitor_progress,
-        max_retries_on_rate_limit=max_retries_on_rate_limit,
-        is_rate_limited=is_rate_limited,
     )
     if not isinstance(raw, dict):
         raise Exception("Expected JSON response to validate into a Pydantic model, got non-JSON (binary or text).")
@@ -200,8 +194,6 @@ async def sync_op_raw(
     final_label_on_success: str | None = "Completed",
     progress_origin_ts: float | None = None,
     monitor_progress: bool = True,
-    max_retries_on_rate_limit: int = 16,
-    is_rate_limited: Callable[[int, Any], bool] | None = None,
 ) -> dict[str, Any] | bytes:
     """
     Make a single network request.
@@ -230,8 +222,6 @@ async def sync_op_raw(
         final_label_on_success=final_label_on_success,
         progress_origin_ts=progress_origin_ts,
         price_extractor=price_extractor,
-        max_retries_on_rate_limit=max_retries_on_rate_limit,
-        is_rate_limited=is_rate_limited,
     )
     return await _request_base(cfg, expect_binary=as_binary)
 
@@ -516,7 +506,7 @@ def _friendly_http_message(status: int, body: Any) -> str:
     if status == 409:
         return "There is a problem with your account. Please contact support@comfy.org."
     if status == 429:
-        return "Rate Limit Exceeded: The server returned 429 after all retry attempts. Please wait and try again."
+        return "Rate Limit Exceeded: Please try again later."
     try:
         if isinstance(body, dict):
             err = body.get("error")
@@ -596,8 +586,6 @@ async def _request_base(cfg: _RequestConfig, expect_binary: bool):
     start_time = cfg.progress_origin_ts if cfg.progress_origin_ts is not None else time.monotonic()
     attempt = 0
     delay = cfg.retry_delay
-    rate_limit_attempts = 0
-    rate_limit_delay = cfg.retry_delay
     operation_succeeded: bool = False
     final_elapsed_seconds: int | None = None
     extracted_price: float | None = None
@@ -665,14 +653,17 @@ async def _request_base(cfg: _RequestConfig, expect_binary: bool):
                 payload_headers["Content-Type"] = "application/json"
                 payload_kw["json"] = cfg.data or {}
 
-            request_logger.log_request_response(
-                operation_id=operation_id,
-                request_method=method,
-                request_url=url,
-                request_headers=dict(payload_headers) if payload_headers else None,
-                request_params=dict(params) if params else None,
-                request_data=request_body_log,
-            )
+            try:
+                request_logger.log_request_response(
+                    operation_id=operation_id,
+                    request_method=method,
+                    request_url=url,
+                    request_headers=dict(payload_headers) if payload_headers else None,
+                    request_params=dict(params) if params else None,
+                    request_data=request_body_log,
+                )
+            except Exception as _log_e:
+                logging.debug("[DEBUG] request logging failed: %s", _log_e)
 
             req_coro = sess.request(method, url, params=params, **payload_kw)
             req_task = asyncio.create_task(req_coro)
@@ -697,33 +688,41 @@ async def _request_base(cfg: _RequestConfig, expect_binary: bool):
                         body = await resp.json()
                     except (ContentTypeError, json.JSONDecodeError):
                         body = await resp.text()
-                    should_retry = False
-                    wait_time = 0.0
-                    retry_label = ""
-                    is_rl = resp.status == 429 or (
-                        cfg.is_rate_limited is not None and cfg.is_rate_limited(resp.status, body)
-                    )
-                    if is_rl and rate_limit_attempts < cfg.max_retries_on_rate_limit:
-                        rate_limit_attempts += 1
-                        wait_time = min(rate_limit_delay, 30.0)
-                        rate_limit_delay *= cfg.retry_backoff
-                        retry_label = f"rate-limit retry {rate_limit_attempts} of {cfg.max_retries_on_rate_limit}"
-                        should_retry = True
-                    elif resp.status in _RETRY_STATUS and (attempt - rate_limit_attempts) <= cfg.max_retries:
-                        wait_time = delay
-                        delay *= cfg.retry_backoff
-                        retry_label = f"retry {attempt - rate_limit_attempts} of {cfg.max_retries}"
-                        should_retry = True
-
-                    if should_retry:
+                    if resp.status in _RETRY_STATUS and attempt <= cfg.max_retries:
                         logging.warning(
-                            "HTTP %s %s -> %s. Waiting %.2fs (%s).",
+                            "HTTP %s %s -> %s. Retrying in %.2fs (retry %d of %d).",
                             method,
                             url,
                             resp.status,
-                            wait_time,
-                            retry_label,
+                            delay,
+                            attempt,
+                            cfg.max_retries,
                         )
+                        try:
+                            request_logger.log_request_response(
+                                operation_id=operation_id,
+                                request_method=method,
+                                request_url=url,
+                                response_status_code=resp.status,
+                                response_headers=dict(resp.headers),
+                                response_content=body,
+                                error_message=_friendly_http_message(resp.status, body),
+                            )
+                        except Exception as _log_e:
+                            logging.debug("[DEBUG] response logging failed: %s", _log_e)
+
+                        await sleep_with_interrupt(
+                            delay,
+                            cfg.node_cls,
+                            cfg.wait_label if cfg.monitor_progress else None,
+                            start_time if cfg.monitor_progress else None,
+                            cfg.estimated_total,
+                            display_callback=_display_time_progress if cfg.monitor_progress else None,
+                        )
+                        delay *= cfg.retry_backoff
+                        continue
+                    msg = _friendly_http_message(resp.status, body)
+                    try:
                         request_logger.log_request_response(
                             operation_id=operation_id,
                             request_method=method,
@@ -731,27 +730,10 @@ async def _request_base(cfg: _RequestConfig, expect_binary: bool):
                             response_status_code=resp.status,
                             response_headers=dict(resp.headers),
                             response_content=body,
-                            error_message=f"HTTP {resp.status} ({retry_label}, will retry in {wait_time:.1f}s)",
+                            error_message=msg,
                         )
-                        await sleep_with_interrupt(
-                            wait_time,
-                            cfg.node_cls,
-                            cfg.wait_label if cfg.monitor_progress else None,
-                            start_time if cfg.monitor_progress else None,
-                            cfg.estimated_total,
-                            display_callback=_display_time_progress if cfg.monitor_progress else None,
-                        )
-                        continue
-                    msg = _friendly_http_message(resp.status, body)
-                    request_logger.log_request_response(
-                        operation_id=operation_id,
-                        request_method=method,
-                        request_url=url,
-                        response_status_code=resp.status,
-                        response_headers=dict(resp.headers),
-                        response_content=body,
-                        error_message=msg,
-                    )
+                    except Exception as _log_e:
+                        logging.debug("[DEBUG] response logging failed: %s", _log_e)
                     raise Exception(msg)
 
                 if expect_binary:
@@ -771,14 +753,17 @@ async def _request_base(cfg: _RequestConfig, expect_binary: bool):
                     bytes_payload = bytes(buff)
                     operation_succeeded = True
                     final_elapsed_seconds = int(time.monotonic() - start_time)
-                    request_logger.log_request_response(
-                        operation_id=operation_id,
-                        request_method=method,
-                        request_url=url,
-                        response_status_code=resp.status,
-                        response_headers=dict(resp.headers),
-                        response_content=bytes_payload,
-                    )
+                    try:
+                        request_logger.log_request_response(
+                            operation_id=operation_id,
+                            request_method=method,
+                            request_url=url,
+                            response_status_code=resp.status,
+                            response_headers=dict(resp.headers),
+                            response_content=bytes_payload,
+                        )
+                    except Exception as _log_e:
+                        logging.debug("[DEBUG] response logging failed: %s", _log_e)
                     return bytes_payload
                 else:
                     try:
@@ -795,39 +780,45 @@ async def _request_base(cfg: _RequestConfig, expect_binary: bool):
                         extracted_price = cfg.price_extractor(payload) if cfg.price_extractor else None
                     operation_succeeded = True
                     final_elapsed_seconds = int(time.monotonic() - start_time)
-                    request_logger.log_request_response(
-                        operation_id=operation_id,
-                        request_method=method,
-                        request_url=url,
-                        response_status_code=resp.status,
-                        response_headers=dict(resp.headers),
-                        response_content=response_content_to_log,
-                    )
+                    try:
+                        request_logger.log_request_response(
+                            operation_id=operation_id,
+                            request_method=method,
+                            request_url=url,
+                            response_status_code=resp.status,
+                            response_headers=dict(resp.headers),
+                            response_content=response_content_to_log,
+                        )
+                    except Exception as _log_e:
+                        logging.debug("[DEBUG] response logging failed: %s", _log_e)
                     return payload
 
         except ProcessingInterrupted:
             logging.debug("Polling was interrupted by user")
             raise
         except (ClientError, OSError) as e:
-            if (attempt - rate_limit_attempts) <= cfg.max_retries:
+            if attempt <= cfg.max_retries:
                 logging.warning(
                     "Connection error calling %s %s. Retrying in %.2fs (%d/%d): %s",
                     method,
                     url,
                     delay,
-                    attempt - rate_limit_attempts,
+                    attempt,
                     cfg.max_retries,
                     str(e),
                 )
-                request_logger.log_request_response(
-                    operation_id=operation_id,
-                    request_method=method,
-                    request_url=url,
-                    request_headers=dict(payload_headers) if payload_headers else None,
-                    request_params=dict(params) if params else None,
-                    request_data=request_body_log,
-                    error_message=f"{type(e).__name__}: {str(e)} (will retry)",
-                )
+                try:
+                    request_logger.log_request_response(
+                        operation_id=operation_id,
+                        request_method=method,
+                        request_url=url,
+                        request_headers=dict(payload_headers) if payload_headers else None,
+                        request_params=dict(params) if params else None,
+                        request_data=request_body_log,
+                        error_message=f"{type(e).__name__}: {str(e)} (will retry)",
+                    )
+                except Exception as _log_e:
+                    logging.debug("[DEBUG] request error logging failed: %s", _log_e)
                 await sleep_with_interrupt(
                     delay,
                     cfg.node_cls,
@@ -840,6 +831,23 @@ async def _request_base(cfg: _RequestConfig, expect_binary: bool):
                 continue
             diag = await _diagnose_connectivity()
             if not diag["internet_accessible"]:
+                try:
+                    request_logger.log_request_response(
+                        operation_id=operation_id,
+                        request_method=method,
+                        request_url=url,
+                        request_headers=dict(payload_headers) if payload_headers else None,
+                        request_params=dict(params) if params else None,
+                        request_data=request_body_log,
+                        error_message=f"LocalNetworkError: {str(e)}",
+                    )
+                except Exception as _log_e:
+                    logging.debug("[DEBUG] final error logging failed: %s", _log_e)
+                raise LocalNetworkError(
+                    "Unable to connect to the API server due to local network issues. "
+                    "Please check your internet connection and try again."
+                ) from e
+            try:
                 request_logger.log_request_response(
                     operation_id=operation_id,
                     request_method=method,
@@ -847,21 +855,10 @@ async def _request_base(cfg: _RequestConfig, expect_binary: bool):
                     request_headers=dict(payload_headers) if payload_headers else None,
                     request_params=dict(params) if params else None,
                     request_data=request_body_log,
-                    error_message=f"LocalNetworkError: {str(e)}",
+                    error_message=f"ApiServerError: {str(e)}",
                 )
-                raise LocalNetworkError(
-                    "Unable to connect to the API server due to local network issues. "
-                    "Please check your internet connection and try again."
-                ) from e
-            request_logger.log_request_response(
-                operation_id=operation_id,
-                request_method=method,
-                request_url=url,
-                request_headers=dict(payload_headers) if payload_headers else None,
-                request_params=dict(params) if params else None,
-                request_data=request_body_log,
-                error_message=f"ApiServerError: {str(e)}",
-            )
+            except Exception as _log_e:
+                logging.debug("[DEBUG] final error logging failed: %s", _log_e)
             raise ApiServerError(
                 f"The API server at {default_base_url()} is currently unreachable. "
                 f"The service may be experiencing issues."

comfy_api_nodes/util/download_helpers.py

@@ -167,25 +167,27 @@ async def download_url_to_bytesio(
                     with contextlib.suppress(Exception):
                         dest.seek(0)
 
-                request_logger.log_request_response(
-                    operation_id=op_id,
-                    request_method="GET",
-                    request_url=url,
-                    response_status_code=resp.status,
-                    response_headers=dict(resp.headers),
-                    response_content=f"[streamed {written} bytes to dest]",
-                )
+                with contextlib.suppress(Exception):
+                    request_logger.log_request_response(
+                        operation_id=op_id,
+                        request_method="GET",
+                        request_url=url,
+                        response_status_code=resp.status,
+                        response_headers=dict(resp.headers),
+                        response_content=f"[streamed {written} bytes to dest]",
+                    )
                 return
         except asyncio.CancelledError:
             raise ProcessingInterrupted("Task cancelled") from None
         except (ClientError, OSError) as e:
             if attempt <= max_retries:
-                request_logger.log_request_response(
-                    operation_id=op_id,
-                    request_method="GET",
-                    request_url=url,
-                    error_message=f"{type(e).__name__}: {str(e)} (will retry)",
-                )
+                with contextlib.suppress(Exception):
+                    request_logger.log_request_response(
+                        operation_id=op_id,
+                        request_method="GET",
+                        request_url=url,
+                        error_message=f"{type(e).__name__}: {str(e)} (will retry)",
+                    )
                 await sleep_with_interrupt(delay, cls, None, None, None)
                 delay *= retry_backoff
                 continue

comfy_api_nodes/util/request_logger.py

@@ -8,6 +8,7 @@ from typing import Any
 
 import folder_paths
 
+# Get the logger instance
 logger = logging.getLogger(__name__)
 
 
@@ -90,41 +91,38 @@ def log_request_response(
     Filenames are sanitized and length-limited for cross-platform safety.
     If we still fail to write, we fall back to appending into api.log.
     """
+    log_dir = get_log_directory()
+    filepath = _build_log_filepath(log_dir, operation_id, request_url)
+
+    log_content: list[str] = []
+    log_content.append(f"Timestamp: {datetime.datetime.now().isoformat()}")
+    log_content.append(f"Operation ID: {operation_id}")
+    log_content.append("-" * 30 + " REQUEST " + "-" * 30)
+    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)}")
+    if request_params:
+        log_content.append(f"Params:\n{_format_data_for_logging(request_params)}")
+    if request_data is not None:
+        log_content.append(f"Data/Body:\n{_format_data_for_logging(request_data)}")
+
+    log_content.append("\n" + "-" * 30 + " RESPONSE " + "-" * 30)
+    if response_status_code is not None:
+        log_content.append(f"Status Code: {response_status_code}")
+    if response_headers:
+        log_content.append(f"Headers:\n{_format_data_for_logging(response_headers)}")
+    if response_content is not None:
+        log_content.append(f"Content:\n{_format_data_for_logging(response_content)}")
+    if error_message:
+        log_content.append(f"Error:\n{error_message}")
+
     try:
-        log_dir = get_log_directory()
-        filepath = _build_log_filepath(log_dir, operation_id, request_url)
-
-        log_content: list[str] = []
-        log_content.append(f"Timestamp: {datetime.datetime.now().isoformat()}")
-        log_content.append(f"Operation ID: {operation_id}")
-        log_content.append("-" * 30 + " REQUEST " + "-" * 30)
-        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)}")
-        if request_params:
-            log_content.append(f"Params:\n{_format_data_for_logging(request_params)}")
-        if request_data is not None:
-            log_content.append(f"Data/Body:\n{_format_data_for_logging(request_data)}")
-
-        log_content.append("\n" + "-" * 30 + " RESPONSE " + "-" * 30)
-        if response_status_code is not None:
-            log_content.append(f"Status Code: {response_status_code}")
-        if response_headers:
-            log_content.append(f"Headers:\n{_format_data_for_logging(response_headers)}")
-        if response_content is not None:
-            log_content.append(f"Content:\n{_format_data_for_logging(response_content)}")
-        if error_message:
-            log_content.append(f"Error:\n{error_message}")
-
-        try:
-            with open(filepath, "w", encoding="utf-8") as f:
-                f.write("\n".join(log_content))
-            logger.debug("API log saved to: %s", filepath)
-        except Exception as e:
-            logger.error("Error writing API log to %s: %s", filepath, str(e))
-    except Exception as _log_e:
-        logging.debug("[DEBUG] log_request_response failed: %s", _log_e)
+        with open(filepath, "w", encoding="utf-8") as f:
+            f.write("\n".join(log_content))
+        logger.debug("API log saved to: %s", filepath)
+    except Exception as e:
+        logger.error("Error writing API log to %s: %s", filepath, str(e))
 
 
 if __name__ == '__main__':

comfy_api_nodes/util/upload_helpers.py

@@ -255,14 +255,17 @@ async def upload_file(
         monitor_task = asyncio.create_task(_monitor())
         sess: aiohttp.ClientSession | None = None
         try:
-            request_logger.log_request_response(
-                operation_id=operation_id,
-                request_method="PUT",
-                request_url=upload_url,
-                request_headers=headers or None,
-                request_params=None,
-                request_data=f"[File data {len(data)} bytes]",
-            )
+            try:
+                request_logger.log_request_response(
+                    operation_id=operation_id,
+                    request_method="PUT",
+                    request_url=upload_url,
+                    request_headers=headers or None,
+                    request_params=None,
+                    request_data=f"[File data {len(data)} bytes]",
+                )
+            except Exception as e:
+                logging.debug("[DEBUG] upload request logging failed: %s", e)
 
             sess = aiohttp.ClientSession(timeout=timeout)
             req = sess.put(upload_url, data=data, headers=headers, skip_auto_headers=skip_auto_headers)
@@ -308,27 +311,31 @@ async def upload_file(
                         delay *= retry_backoff
                         continue
                     raise Exception(f"Failed to upload (HTTP {resp.status}).")
-                request_logger.log_request_response(
-                    operation_id=operation_id,
-                    request_method="PUT",
-                    request_url=upload_url,
-                    response_status_code=resp.status,
-                    response_headers=dict(resp.headers),
-                    response_content="File uploaded successfully.",
-                )
+                try:
+                    request_logger.log_request_response(
+                        operation_id=operation_id,
+                        request_method="PUT",
+                        request_url=upload_url,
+                        response_status_code=resp.status,
+                        response_headers=dict(resp.headers),
+                        response_content="File uploaded successfully.",
+                    )
+                except Exception as e:
+                    logging.debug("[DEBUG] upload response logging failed: %s", e)
                 return
         except asyncio.CancelledError:
             raise ProcessingInterrupted("Task cancelled") from None
         except (aiohttp.ClientError, OSError) as e:
             if attempt <= max_retries:
-                request_logger.log_request_response(
-                    operation_id=operation_id,
-                    request_method="PUT",
-                    request_url=upload_url,
-                    request_headers=headers or None,
-                    request_data=f"[File data {len(data)} bytes]",
-                    error_message=f"{type(e).__name__}: {str(e)} (will retry)",
-                )
+                with contextlib.suppress(Exception):
+                    request_logger.log_request_response(
+                        operation_id=operation_id,
+                        request_method="PUT",
+                        request_url=upload_url,
+                        request_headers=headers or None,
+                        request_data=f"[File data {len(data)} bytes]",
+                        error_message=f"{type(e).__name__}: {str(e)} (will retry)",
+                    )
                 await sleep_with_interrupt(
                     delay,
                     cls,