torch.AcceleratorError: CUDA error: invalid argument on 9060xt 16g

### Custom Node Testing

- [x] I have tried disabling custom nodes and the issue persists (see [how to disable custom nodes](https://docs.comfy.org/troubleshooting/custom-node-issues#step-1%3A-test-with-all-custom-nodes-disabled) if you need help)

### Your question
torch.AcceleratorError: CUDA error: invalid argument on 9060xt 16g when running minimaxh3 official t2v workflow

### Logs

```powershell
(comfyui) D:\ComfyUI_all>python -s "ComfyUI\main.py" --windows-standalone-build --enable-dynamic-vram --disable-smart-memory --disable-pinned-memory --reserve-vram 3 --use-ck-attention --disable-auto-launch --cuda-device 1
[INFO] setup plugin alembic.autogenerate.schemas
[INFO] setup plugin alembic.autogenerate.tables
[INFO] setup plugin alembic.autogenerate.types
[INFO] setup plugin alembic.autogenerate.constraints
[INFO] setup plugin alembic.autogenerate.defaults
[INFO] setup plugin alembic.autogenerate.comments
[INFO] setup plugin alembic.autogenerate.checkconstraint_byname
[INFO] comfy-aimdo failed to load: Could not find module 'E:\miniconda3\envs\comfyui\Lib\site-packages\comfy_aimdo\aimdo_rocm.dll' (or one of its dependencies). Try using the full path with constructor syntax.
[INFO] NOTE: comfy-aimdo currently only supports Nvidia and AMD GPUs
[INFO] Set cuda device to: 1
[START] Security scan
[DONE] Security scan
[ComfyUI-Manager] Logging failed: [WinError 32] 另一个程序正在使用此文件，进程无法访问。: 'D:\\ComfyUI_all\\ComfyUI\\user\\comfyui.log' -> 'D:\\ComfyUI_all\\ComfyUI\\user\\comfyui.prev.log'
## ComfyUI-Manager: installing dependencies done.
** ComfyUI startup time: 2026-08-16 01:39:11.651
** Platform: Windows
** Python version: 3.12.13 | packaged by Anaconda, Inc. | (main, Jul 9 2026, 14:26:47) [MSC v.1942 64 bit (AMD64)]
** Python executable: E:\miniconda3\envs\comfyui\python.exe
** ComfyUI Path: D:\ComfyUI_all\ComfyUI
** ComfyUI Base Folder Path: D:\ComfyUI_all\ComfyUI
** User directory: D:\ComfyUI_all\ComfyUI\user
** ComfyUI-Manager config path: D:\ComfyUI_all\ComfyUI\user\__manager\config.ini
** Log path: D:\ComfyUI_all\ComfyUI\user\comfyui.log
[INFO]
Prestartup times for custom nodes:
[INFO] 0.0 seconds: D:\ComfyUI_all\ComfyUI\custom_nodes\rgthree-comfy
[INFO] 3.0 seconds: D:\ComfyUI_all\ComfyUI\custom_nodes\ComfyUI-Manager
[INFO]
[INFO] Found comfy_kitchen backend cuda: {'available': False, 'disabled': True, 'unavailable_reason': 'Extension file not found: E:\\miniconda3\\envs\\comfyui\\Lib\\site-packages\\comfy_kitchen\\backends\\cuda\\_C.abi3.pyd', 'capabilities': []}
[INFO] Found comfy_kitchen backend eager: {'available': True, 'disabled': False, 'unavailable_reason': None, 'capabilities': ['adaln', 'apply_rope', 'apply_rope1', 'apply_rope1_', 'apply_rope_', 'apply_rope_split_half', 'apply_rope_split_half1', 'apply_rope_split_half1_', 'apply_rope_split_half_', 'convrot_w4a4_linear', 'dequantize_convrot_w4a4_weight', 'dequantize_int8_convrot_weight', 'dequantize_int8_convrot_weight_dtype', 'dequantize_int8_embedding', 'dequantize_int8_simple', 'dequantize_int8_simple_dtype', 'dequantize_mxfp8', 'dequantize_nvfp4', 'dequantize_per_tensor_fp8', 'dequantize_w4a8_int8_weight', 'gemv_awq_w4a16', 'int8_linear', 'na3d', 'prepare_int4_weight_for_int8_linear', 'quantize_and_rotate_rowwise', 'quantize_convrot_w4a4_weight', 'quantize_int8_convrot_weight', 'quantize_int8_rowwise', 'quantize_int8_tensorwise', 'quantize_mxfp8', 'quantize_nvfp4', 'quantize_per_tensor_fp8', 'quantize_svdquant_w4a4', 'quantize_w4a8_int8_weight', 'rms_adaln', 'rms_rope', 'rms_rope1', 'rms_rope1_', 'rms_rope_', 'rms_rope_split_half', 'rms_rope_split_half1', 'rms_rope_split_half1_', 'rms_rope_split_half_', 'rotate_int8_convrot_weight', 'scaled_mm_mxfp8', 'scaled_mm_nvfp4', 'scaled_mm_svdquant_w4a4', 'stochastic_rounding_fp8', 'w4a8_int8_linear']}
[INFO] Found comfy_kitchen backend triton: {'available': True, 'disabled': True, 'unavailable_reason': None, 'capabilities': ['adaln', 'apply_rope', 'apply_rope1', 'apply_rope1_', 'apply_rope_', 'apply_rope_split_half', 'apply_rope_split_half1', 'apply_rope_split_half1_', 'apply_rope_split_half_', 'dequantize_nvfp4', 'dequantize_per_tensor_fp8', 'int8_linear', 'na3d', 'quantize_and_rotate_rowwise', 'quantize_int8_rowwise', 'quantize_mxfp8', 'quantize_nvfp4', 'quantize_per_tensor_fp8', 'rms_adaln', 'rms_rope', 'rms_rope1', 'rms_rope1_', 'rms_rope_', 'rms_rope_split_half', 'rms_rope_split_half1', 'rms_rope_split_half1_', 'rms_rope_split_half_', 'w4a8_int8_linear']}
[INFO] Found comfy_kitchen backend hip: {'available': True, 'disabled': False, 'unavailable_reason': None, 'capabilities': ['adaln', 'apply_rope', 'apply_rope1', 'apply_rope1_', 'apply_rope_', 'apply_rope_split_half', 'apply_rope_split_half1', 'apply_rope_split_half1_', 'apply_rope_split_half_', 'convrot_w4a4_linear', 'dequantize_convrot_w4a4_weight', 'dequantize_int8_convrot_weight_dtype', 'dequantize_int8_simple_dtype', 'dequantize_per_tensor_fp8', 'dequantize_w4a8_int8_weight', 'gemv_awq_w4a16', 'int8_linear', 'na3d', 'quantize_and_rotate_rowwise', 'quantize_convrot_w4a4_weight', 'quantize_int8_convrot_weight', 'quantize_int8_rowwise', 'quantize_int8_tensorwise', 'quantize_per_tensor_fp8', 'quantize_svdquant_w4a4', 'quantize_w4a8_int8_weight', 'rms_adaln', 'rms_rope', 'rms_rope1', 'rms_rope1_', 'rms_rope_', 'rms_rope_split_half', 'rms_rope_split_half1', 'rms_rope_split_half1_', 'rms_rope_split_half_', 'scaled_mm_svdquant_w4a4', 'stochastic_rounding_fp8', 'w4a8_int8_linear']}
[INFO] Checkpoint files will always be loaded safely.
[INFO] Total VRAM 16304 MB, total RAM 31350 MB
[INFO] pytorch version: 2.12.0+rocm7.14.0
[INFO] Set: torch.backends.cudnn.enabled = False for better AMD performance.
[INFO] AMD arch: gfx1200
[INFO] ROCm version: (7, 14)
[INFO] Set vram state to: NORMAL_VRAM
[INFO] Disabling smart memory management
[INFO] Device: cuda:0 AMD Radeon RX 9060 XT : native
[INFO] Using async weight offloading with 2 streams
E:\miniconda3\envs\comfyui\Lib\site-packages\flash_attn\flash_attn_interface.py:17: UserWarning: flash_attn_2_cuda (which has ROCm/HIP kernels) not found, falling back to Triton implementation
warnings.warn("flash_attn_2_cuda (which has ROCm/HIP kernels) not found, falling back to Triton implementation")
[aiter] Triton ops only: CK and HIP ops (and their JIT build) are skipped.
[INFO] Using pytorch attention
[INFO] Using Comfy Kitchen attention
[WARNING] No working comfy-aimdo install detected. DynamicVRAM support disabled. Falling back to legacy ModelPatcher. VRAM estimates may be unreliable especially on Windows
[INFO] Python version: 3.12.13 | packaged by Anaconda, Inc. | (main, Jul 9 2026, 14:26:47) [MSC v.1942 64 bit (AMD64)]
[INFO] ComfyUI version: 0.33.0
[INFO] comfy-aimdo version: 0.4.13
[INFO] comfy-kitchen version: 0.2.31
[INFO] [Prompt Server] web root: E:\miniconda3\envs\comfyui\Lib\site-packages\comfyui_frontend_package\static
[INFO] Asset seeder disabled
[INFO] No OpenGL_accelerate module loaded: No module named 'OpenGL_accelerate'
[INFO] ### Loading: ComfyUI-Manager (V3.41)
[INFO] [ComfyUI-Manager] network_mode: public
[INFO] [ComfyUI-Manager] ComfyUI per-queue preview override detected (PR #11261). Manager's preview method feature is disabled. Use ComfyUI's --preview-method CLI option or 'Settings > Execution > Live preview method'.
[INFO] ### ComfyUI Version: v0.33.0-8-g3d0e5c34 on 'local' | Released on '2026-08-14'

[rgthree-comfy] Loaded 48 epic nodes. 🎉Duplicate of #

[rgthree-comfy] ComfyUI's new Node 2.0 rendering may be incompatible with some rgthree-comfy nodes and features, breaking some rendering as well as losing the ability to access a node's properties (a vital part of many nodes). It also appears to run MUCH more slowly spiking CPU usage and causing jankiness and unresponsiveness, especially with large workflows. Personally I am not planning to use the new Nodes 2.0 and, unfortunately, am not able to invest the time to investigate and overhaul rgthree-comfy where needed. If you have issues when Nodes 2.0 is enabled, I'd urge you to switch it off as well and join me in hoping ComfyUI is not planning to deprecate the existing, stable canvas rendering all together.

🔧 torchaudio.save/load globally patched (scipy for WAV files, no TorchCodec)
🔇 Suppressed torchaudio 2.9 migration warnings
[INFO] [ComfyUI-Manager] default cache updated: https://raw.githubusercontent.com/ltdrdata/ComfyUI-Manager/main/alter-list.json
[INFO] [ComfyUI-Manager] default cache updated: https://raw.githubusercontent.com/ltdrdata/ComfyUI-Manager/main/model-list.json
[INFO] [ComfyUI-Manager] default cache updated: https://raw.githubusercontent.com/ltdrdata/ComfyUI-Manager/main/github-stats.json
🔬 Numba compatibility setup: 0.85s
✅ Numba JIT working properly (tested in 0.85s)
ℹ️ Critical package versions: NumPy 2.2.6, Librosa 0.11.0, Numba 0.66.0, PyTorch 2.12.0+rocm7.14.0, TorchAudio 2.11.0+rocm7.14.0, Transformers 5.14.1, Accelerate 1.14.0, SoundFile 0.14.0
⚠️ FFmpeg not found - using fallback audio processing (reduced quality)
💡 Install FFmpeg for optimal performance: https://ffmpeg.org/download.html
[INFO] [ComfyUI-Manager] default cache updated: https://raw.githubusercontent.com/ltdrdata/ComfyUI-Manager/main/extension-node-map.json
[INFO] [ComfyUI-Manager] default cache updated: https://raw.githubusercontent.com/ltdrdata/ComfyUI-Manager/main/custom-node-list.json
======================================================================
🚀 TTS Audio Suite v5.8.1
Universal multi-engine TTS extension for ComfyUI
✅ TTS Audio Suite v5.8.1 loaded with 59 nodes:
• ⚙️ ChatterBox Official 23-Lang Engine
• ⚙️ ChatterBox TTS Engine
• ⚙️ CosyVoice3 Engine
• ⚙️ Dots TTS Engine
• ⚙️ DramaBox Engine
• ⚙️ Echo-TTS Engine
• ⚙️ F5 TTS Engine
• ⚙️ Fish Audio S2 Pro Engine
• ⚙️ Granite ASR Engine
• ⚙️ Higgs Audio 2 Engine
• ⚙️ Higgs Audio v3 Engine
• ⚙️ IndexTTS 2 / 2.5 Engine
• ⚙️ MOSS SoundEffect v2 Engine
• ⚙️ MOSS-TTS Engine
• ⚙️ OmniVoice Engine
• ⚙️ Qwen3-TTS Engine
• ⚙️ RVC Engine
• ⚙️ Step Audio EditX Engine
• ⚙️ VibeVoice Engine
• ♻️ Refresh Voice Cache
• ✏️ ASR Transcribe
• 🌈 IndexTTS-2 Emotion Vectors
• 🌈 IndexTTS-2 Text Emotion
• 🌊 Audio Wave Analyzer
• 🌩️ Sound Effects
• 🎓 Model Training
• 🎙️ Voice Capture
• 🎛️ DramaBox Training Config
• 🎛️ MOSS Training Config
• 🎛️ RVC Training Config
• 🎞️ Training Clip Staging
• 🎤 TTS Text
• 🎨 Step Audio EditX - Audio Editor
• 🎨 Voice Designer
• 🎭 Character Voices
• 🎭 Load RVC Character Model
• 🏷️ Multiline TTS Tag Editor
• 👄 F5-TTS Speech Editor
• 💾 Save Character Voice
• 📐 Visual Tag Builder
• 📝 ASR Punctuation / Truecase
• 📝 Phoneme Text Normalizer
• 📦 DramaBox Dataset Prep
• 📦 MOSS Dataset Prep
• 📦 RVC Dataset Prep
• 📺 TTS SRT
• 📺 Text to SRT Builder
• 🔄 Voice Changer
• 🔧 Audio Analyzer Options
• 🔧 F5-TTS Edit Options
• 🔧 RVC Pitch Extraction Options
• 🔧 SRT Advanced Options
• 🔧 Viseme Mouth Shape Options
• 🗣️ Silent Speech Analyzer
• 🤐 Noise or Vocal Removal
• 🤐 Voice Fixer
• 🥪 Merge Audio
• 🧾 DramaBox Dataset Rows
• 🧾 MOSS Dataset Rows
======================================================================
[INFO]
Import times for custom nodes:
[INFO] 0.0 seconds: D:\ComfyUI_all\ComfyUI\custom_nodes\websocket_image_save.py
[INFO] 0.0 seconds: D:\ComfyUI_all\ComfyUI\custom_nodes\comfyui-inpaint-cropandstitch
[INFO] 0.0 seconds: D:\ComfyUI_all\ComfyUI\custom_nodes\Anomalous_Model_Browser
[INFO] 0.0 seconds: D:\ComfyUI_all\ComfyUI\custom_nodes\ComfyUI-SolAttn_triton
[INFO] 0.0 seconds: D:\ComfyUI_all\ComfyUI\custom_nodes\ComfyUI-MiniMaxH3-Director
[INFO] 0.0 seconds: D:\ComfyUI_all\ComfyUI\custom_nodes\rgthree-comfy
[INFO] 0.0 seconds: D:\ComfyUI_all\ComfyUI\custom_nodes\ComfyUI-KJNodes
[INFO] 0.6 seconds: D:\ComfyUI_all\ComfyUI\custom_nodes\ComfyUI-Manager
[INFO] 3.3 seconds: D:\ComfyUI_all\ComfyUI\custom_nodes\TTS-Audio-Suite
[INFO]
[INFO] Context impl SQLiteImpl.
[INFO] Will assume non-transactional DDL.
[INFO] Using RAM pressure cache.
[INFO] Starting server

[INFO] To see the GUI go to: http://127.0.0.1:8188
[INFO] got prompt
[INFO] VAE load device: cuda:0, offload device: cpu, dtype: torch.float32
[INFO] VAE load device: cuda:0, offload device: cpu, dtype: torch.float16
[INFO] Found quantization metadata version 1
[INFO] Using MixedPrecisionOps for text encoder
[INFO] CLIP/text encoder model load device: cuda:0, offload device: cpu, current: cpu, dtype: torch.float16
[INFO] Requested to load MiniMaxH3TEModel_
[INFO] loaded partially; 12269.80 MB usable, 11394.51 MB loaded, 14488.75 MB offloaded, 875.29 MB buffer reserved, lowvram patches: 0
D:\ComfyUI_all\ComfyUI\comfy\text_encoders\llama.py:462: UserWarning: bgemm_internal_cublaslt error: HIPBLAS_STATUS_NOT_SUPPORTED when calling hipblasLtMatmul with transpose_mat1 n transpose_mat2 n m 388 n 64 k 1 lda 388 ldb 1 ldc 388 abType 0 cType 0 computeType 6 scaleType 0. Will attempt to recover by calling cublas instead. (Triggered internally at B:\src\pytorch\aten\src\ATen\hip\HIPBlas.cpp:602.)
freqs = (inv_freq_expanded.float() @ position_ids_expanded.float()).transpose(1, 2)
D:\ComfyUI_all\ComfyUI\comfy\ops.py:1328: UserWarning: bgemm_internal_cublaslt error: HIPBLAS_STATUS_NOT_SUPPORTED whencalling hipblasLtMatmul with transpose_mat1 t transpose_mat2 n m 8192 n 388 k 5120 lda 5120 ldb 5120 ldc 8192 abType 0 cType 0 computeType 6 scaleType 0. Will attempt to recover by calling cublas instead. (Triggered internally at B:\src\pytorch\aten\src\ATen\hip\HIPBlas.cpp:602.)
return torch.nn.functional.linear(input, weight, bias)
D:\ComfyUI_all\ComfyUI\comfy\ops.py:1328: UserWarning: bgemm_internal_cublaslt error: HIPBLAS_STATUS_NOT_SUPPORTED whencalling hipblasLtMatmul with transpose_mat1 t transpose_mat2 n m 1024 n 388 k 5120 lda 5120 ldb 5120 ldc 1024 abType 0 cType 0 computeType 6 scaleType 0. Will attempt to recover by calling cublas instead. (Triggered internally at B:\src\pytorch\aten\src\ATen\hip\HIPBlas.cpp:602.)
return torch.nn.functional.linear(input, weight, bias)
[ERROR] !!! Exception during processing !!! CUDA error: invalid argument
Search for `hipErrorInvalidValue' in https://rocm.docs.amd.com/projects/HIP/en/latest/index.html for more information.
CUDA kernel errors might be asynchronously reported at some other API call, so the stacktrace below might be incorrect.
For debugging consider passing AMD_SERIALIZE_KERNEL=3
Device-side assertion tracking was not enabled by user.
[ERROR] Traceback (most recent call last):
File "D:\ComfyUI_all\ComfyUI\execution.py", line 545, in execute
output_data, output_ui, has_subgraph, has_pending_tasks = await get_output_data(prompt_id, unique_id, obj, input_data_all, execution_block_cb=execution_block_cb, pre_execute_cb=pre_execute_cb, v3_data=v3_data)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "D:\ComfyUI_all\ComfyUI\execution.py", line 344, in get_output_data
return_values = await _async_map_node_over_list(prompt_id, unique_id, obj, input_data_all, obj.FUNCTION, allow_interrupt=True, execution_block_cb=execution_block_cb, pre_execute_cb=pre_execute_cb, v3_data=v3_data)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "D:\ComfyUI_all\ComfyUI\execution.py", line 318, in _async_map_node_over_list
await process_inputs(input_dict, i)
File "D:\ComfyUI_all\ComfyUI\execution.py", line 306, in process_inputs
result = f(**inputs)
^^^^^^^^^^^
File "D:\ComfyUI_all\ComfyUI\comfy_api\internal\__init__.py", line 149, in wrapped_func
return method(locked_class, **inputs)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "D:\ComfyUI_all\ComfyUI\comfy_api\latest\_io.py", line 1990, in EXECUTE_NORMALIZED
to_return = cls.execute(*args, **kwargs)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "D:\ComfyUI_all\ComfyUI\comfy_extras\nodes_minimax_h3.py", line 165, in execute
cond = clip.encode_from_tokens_scheduled(tokens)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "D:\ComfyUI_all\ComfyUI\comfy\sd.py", line 340, in encode_from_tokens_scheduled
pooled_dict = self.encode_from_tokens(tokens, return_pooled=return_pooled, return_dict=True)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "D:\ComfyUI_all\ComfyUI\comfy\sd.py", line 409, in encode_from_tokens
o = self.cond_stage_model.encode_token_weights(tokens)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "D:\ComfyUI_all\ComfyUI\comfy\sd1_clip.py", line 743, in encode_token_weights
out = getattr(self, self.clip).encode_token_weights(token_weight_pairs)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "D:\ComfyUI_all\ComfyUI\comfy\text_encoders\minimax.py", line 110, in encode_token_weights
out = super().encode_token_weights(token_weight_pairs)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "D:\ComfyUI_all\ComfyUI\comfy\sd1_clip.py", line 45, in encode_token_weights
o = self.encode(to_encode)
^^^^^^^^^^^^^^^^^^^^^^
File "D:\ComfyUI_all\ComfyUI\comfy\sd1_clip.py", line 306, in encode
return self(tokens)
^^^^^^^^^^^^
File "E:\miniconda3\envs\comfyui\Lib\site-packages\torch\nn\modules\module.py", line 1778, in _wrapped_call_impl
return self._call_impl(*args, **kwargs)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "E:\miniconda3\envs\comfyui\Lib\site-packages\torch\nn\modules\module.py", line 1789, in _call_impl
return forward_call(*args, **kwargs)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "D:\ComfyUI_all\ComfyUI\comfy\sd1_clip.py", line 279, in forward
outputs = self.transformer(None, attention_mask_model, embeds=embeds, num_tokens=num_tokens, intermediate_output=intermediate_output, final_layer_norm_intermediate=self.layer_norm_hidden_state, dtype=torch.float32, embeds_info=embeds_info)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "E:\miniconda3\envs\comfyui\Lib\site-packages\torch\nn\modules\module.py", line 1778, in _wrapped_call_impl
return self._call_impl(*args, **kwargs)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "E:\miniconda3\envs\comfyui\Lib\site-packages\torch\nn\modules\module.py", line 1789, in _call_impl
return forward_call(*args, **kwargs)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "D:\ComfyUI_all\ComfyUI\comfy\text_encoders\minimax.py", line 96, in forward
return super().forward(input_ids, attention_mask=attention_mask, embeds=embeds,
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "D:\ComfyUI_all\ComfyUI\comfy\text_encoders\qwen3vl.py", line 100, in forward
return self.model(
^^^^^^^^^^^
File "E:\miniconda3\envs\comfyui\Lib\site-packages\torch\nn\modules\module.py", line 1778, in _wrapped_call_impl
return self._call_impl(*args, **kwargs)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "E:\miniconda3\envs\comfyui\Lib\site-packages\torch\nn\modules\module.py", line 1789, in _call_impl
return forward_call(*args, **kwargs)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "D:\ComfyUI_all\ComfyUI\comfy\text_encoders\llama.py", line 913, in forward
comfy.model_prefetch.prefetch_queue_pop(prefetch_queue, x.device, layer, x.dtype, core=core, enable_graph=enable_graph)
File "D:\ComfyUI_all\ComfyUI\comfy\model_prefetch.py", line 66, in prefetch_queue_pop
core()
File "D:\ComfyUI_all\ComfyUI\comfy\text_encoders\llama.py", line 903, in core
x, current_kv = layer(
^^^^^^
File "E:\miniconda3\envs\comfyui\Lib\site-packages\torch\nn\modules\module.py", line 1778, in _wrapped_call_impl
return self._call_impl(*args, **kwargs)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "E:\miniconda3\envs\comfyui\Lib\site-packages\torch\nn\modules\module.py", line 1789, in _call_impl
return forward_call(*args, **kwargs)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "D:\ComfyUI_all\ComfyUI\comfy\text_encoders\llama.py", line 665, in forward
x, present_key_value = self.self_attn(
^^^^^^^^^^^^^^^
File "E:\miniconda3\envs\comfyui\Lib\site-packages\torch\nn\modules\module.py", line 1778, in _wrapped_call_impl
return self._call_impl(*args, **kwargs)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "E:\miniconda3\envs\comfyui\Lib\site-packages\torch\nn\modules\module.py", line 1789, in _call_impl
return forward_call(*args, **kwargs)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "D:\ComfyUI_all\ComfyUI\comfy\text_encoders\llama.py", line 616, in forward
return self.o_proj(output), present_key_value
^^^^^^^^^^^^^^^^^^^
File "E:\miniconda3\envs\comfyui\Lib\site-packages\torch\nn\modules\module.py", line 1778, in _wrapped_call_impl
return self._call_impl(*args, **kwargs)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "E:\miniconda3\envs\comfyui\Lib\site-packages\torch\nn\modules\module.py", line 1789, in _call_impl
return forward_call(*args, **kwargs)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "D:\ComfyUI_all\ComfyUI\comfy\ops.py", line 1414, in forward
output = self.forward_comfy_cast_weights(
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "D:\ComfyUI_all\ComfyUI\comfy\ops.py", line 1338, in forward_comfy_cast_weights
with CastBiasWeightContext(
^^^^^^^^^^^^^^^^^^^^^^
File "D:\ComfyUI_all\ComfyUI\comfy\ops.py", line 466, in __init__
self.state = (None, None) if self.slf is None else cast_bias_weight(*args, **kwargs)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "D:\ComfyUI_all\ComfyUI\comfy\ops.py", line 404, in cast_bias_weight
cast_buffer = comfy.model_management.get_cast_buffer(offload_stream, device, cast_buffer_size, s)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "D:\ComfyUI_all\ComfyUI\comfy\model_management.py", line 1403, in get_cast_buffer
cast_buffer = torch.empty((size), dtype=torch.int8, device=device)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
torch.AcceleratorError: CUDA error: invalid argument
Search for `hipErrorInvalidValue' in https://rocm.docs.amd.com/projects/HIP/en/latest/index.html for more information.
CUDA kernel errors might be asynchronously reported at some other API call, so the stacktrace below might be incorrect.
For debugging consider passing AMD_SERIALIZE_KERNEL=3
Device-side assertion tracking was not enabled by user.

[INFO] Prompt executed in 31.43 seconds
[INFO] got prompt
[INFO] Requested to load MiniMaxH3TEModel_
[INFO] loaded partially; 11979.45 MB usable, 11104.16 MB loaded, 14779.52 MB offloaded, 875.29 MB buffer reserved, lowvram patches: 0
[ERROR] !!! Exception during processing !!! CUDA error: invalid argument
Search for `hipErrorInvalidValue' in https://rocm.docs.amd.com/projects/HIP/en/latest/index.html for more information.
CUDA kernel errors might be asynchronously reported at some other API call, so the stacktrace below might be incorrect.
For debugging consider passing AMD_SERIALIZE_KERNEL=3
Device-side assertion tracking was not enabled by user.
[ERROR] Traceback (most recent call last):
File "D:\ComfyUI_all\ComfyUI\execution.py", line 545, in execute
output_data, output_ui, has_subgraph, has_pending_tasks = await get_output_data(prompt_id, unique_id, obj, input_data_all, execution_block_cb=execution_block_cb, pre_execute_cb=pre_execute_cb, v3_data=v3_data)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "D:\ComfyUI_all\ComfyUI\execution.py", line 344, in get_output_data
return_values = await _async_map_node_over_list(prompt_id, unique_id, obj, input_data_all, obj.FUNCTION, allow_interrupt=True, execution_block_cb=execution_block_cb, pre_execute_cb=pre_execute_cb, v3_data=v3_data)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "D:\ComfyUI_all\ComfyUI\execution.py", line 318, in _async_map_node_over_list
await process_inputs(input_dict, i)
File "D:\ComfyUI_all\ComfyUI\execution.py", line 306, in process_inputs
result = f(**inputs)
^^^^^^^^^^^
File "D:\ComfyUI_all\ComfyUI\comfy_api\internal\__init__.py", line 149, in wrapped_func
return method(locked_class, **inputs)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "D:\ComfyUI_all\ComfyUI\comfy_api\latest\_io.py", line 1990, in EXECUTE_NORMALIZED
to_return = cls.execute(*args, **kwargs)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "D:\ComfyUI_all\ComfyUI\comfy_extras\nodes_minimax_h3.py", line 165, in execute
cond = clip.encode_from_tokens_scheduled(tokens)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "D:\ComfyUI_all\ComfyUI\comfy\sd.py", line 340, in encode_from_tokens_scheduled
pooled_dict = self.encode_from_tokens(tokens, return_pooled=return_pooled, return_dict=True)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "D:\ComfyUI_all\ComfyUI\comfy\sd.py", line 409, in encode_from_tokens
o = self.cond_stage_model.encode_token_weights(tokens)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "D:\ComfyUI_all\ComfyUI\comfy\sd1_clip.py", line 743, in encode_token_weights
out = getattr(self, self.clip).encode_token_weights(token_weight_pairs)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "D:\ComfyUI_all\ComfyUI\comfy\text_encoders\minimax.py", line 110, in encode_token_weights
out = super().encode_token_weights(token_weight_pairs)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "D:\ComfyUI_all\ComfyUI\comfy\sd1_clip.py", line 45, in encode_token_weights
o = self.encode(to_encode)
^^^^^^^^^^^^^^^^^^^^^^
File "D:\ComfyUI_all\ComfyUI\comfy\sd1_clip.py", line 306, in encode
return self(tokens)
^^^^^^^^^^^^
File "E:\miniconda3\envs\comfyui\Lib\site-packages\torch\nn\modules\module.py", line 1778, in _wrapped_call_impl
return self._call_impl(*args, **kwargs)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "E:\miniconda3\envs\comfyui\Lib\site-packages\torch\nn\modules\module.py", line 1789, in _call_impl
return forward_call(*args, **kwargs)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "D:\ComfyUI_all\ComfyUI\comfy\sd1_clip.py", line 279, in forward
outputs = self.transformer(None, attention_mask_model, embeds=embeds, num_tokens=num_tokens, intermediate_output=intermediate_output, final_layer_norm_intermediate=self.layer_norm_hidden_state, dtype=torch.float32, embeds_info=embeds_info)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "E:\miniconda3\envs\comfyui\Lib\site-packages\torch\nn\modules\module.py", line 1778, in _wrapped_call_impl
return self._call_impl(*args, **kwargs)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "E:\miniconda3\envs\comfyui\Lib\site-packages\torch\nn\modules\module.py", line 1789, in _call_impl
return forward_call(*args, **kwargs)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "D:\ComfyUI_all\ComfyUI\comfy\text_encoders\minimax.py", line 96, in forward
return super().forward(input_ids, attention_mask=attention_mask, embeds=embeds,
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "D:\ComfyUI_all\ComfyUI\comfy\text_encoders\qwen3vl.py", line 100, in forward
return self.model(
^^^^^^^^^^^
File "E:\miniconda3\envs\comfyui\Lib\site-packages\torch\nn\modules\module.py", line 1778, in _wrapped_call_impl
return self._call_impl(*args, **kwargs)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "E:\miniconda3\envs\comfyui\Lib\site-packages\torch\nn\modules\module.py", line 1789, in _call_impl
return forward_call(*args, **kwargs)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "D:\ComfyUI_all\ComfyUI\comfy\text_encoders\llama.py", line 913, in forward
comfy.model_prefetch.prefetch_queue_pop(prefetch_queue, x.device, layer, x.dtype, core=core, enable_graph=enable_graph)
File "D:\ComfyUI_all\ComfyUI\comfy\model_prefetch.py", line 66, in prefetch_queue_pop
core()
File "D:\ComfyUI_all\ComfyUI\comfy\text_encoders\llama.py", line 903, in core
x, current_kv = layer(
^^^^^^
File "E:\miniconda3\envs\comfyui\Lib\site-packages\torch\nn\modules\module.py", line 1778, in _wrapped_call_impl
return self._call_impl(*args, **kwargs)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "E:\miniconda3\envs\comfyui\Lib\site-packages\torch\nn\modules\module.py", line 1789, in _call_impl
return forward_call(*args, **kwargs)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "D:\ComfyUI_all\ComfyUI\comfy\text_encoders\llama.py", line 665, in forward
x, present_key_value = self.self_attn(
^^^^^^^^^^^^^^^
File "E:\miniconda3\envs\comfyui\Lib\site-packages\torch\nn\modules\module.py", line 1778, in _wrapped_call_impl
return self._call_impl(*args, **kwargs)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "E:\miniconda3\envs\comfyui\Lib\site-packages\torch\nn\modules\module.py", line 1789, in _call_impl
return forward_call(*args, **kwargs)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "D:\ComfyUI_all\ComfyUI\comfy\text_encoders\llama.py", line 616, in forward
return self.o_proj(output), present_key_value
^^^^^^^^^^^^^^^^^^^
File "E:\miniconda3\envs\comfyui\Lib\site-packages\torch\nn\modules\module.py", line 1778, in _wrapped_call_impl
return self._call_impl(*args, **kwargs)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "E:\miniconda3\envs\comfyui\Lib\site-packages\torch\nn\modules\module.py", line 1789, in _call_impl
return forward_call(*args, **kwargs)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "D:\ComfyUI_all\ComfyUI\comfy\ops.py", line 1414, in forward
output = self.forward_comfy_cast_weights(
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "D:\ComfyUI_all\ComfyUI\comfy\ops.py", line 1338, in forward_comfy_cast_weights
with CastBiasWeightContext(
^^^^^^^^^^^^^^^^^^^^^^
File "D:\ComfyUI_all\ComfyUI\comfy\ops.py", line 466, in __init__
self.state = (None, None) if self.slf is None else cast_bias_weight(*args, **kwargs)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "D:\ComfyUI_all\ComfyUI\comfy\ops.py", line 404, in cast_bias_weight
cast_buffer = comfy.model_management.get_cast_buffer(offload_stream, device, cast_buffer_size, s)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "D:\ComfyUI_all\ComfyUI\comfy\model_management.py", line 1403, in get_cast_buffer
cast_buffer = torch.empty((size), dtype=torch.int8, device=device)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
torch.AcceleratorError: CUDA error: invalid argument
Search for `hipErrorInvalidValue' in https://rocm.docs.amd.com/projects/HIP/en/latest/index.html for more information.
CUDA kernel errors might be asynchronously reported at some other API call, so the stacktrace below might be incorrect.
For debugging consider passing AMD_SERIALIZE_KERNEL=3
Device-side assertion tracking was not enabled by user.

[INFO] Prompt executed in 4.66 seconds

```

### Other

_No response_

Start with ComfyUI\main.py and reproduce the official MiniMaxH3 text-to-video workflow on the AMD Radeon RX 9060 XT with custom nodes disabled. Capture the complete traceback around the CUDA invalid-argument error and compare the logged PyTorch/ROCm, attention, and Comfy Kitchen backend settings. Done means the workflow runs successfully or the failure is narrowed to a specific component.

Scritto dal modello di indicizzazione a partire dal testo della issue.