Trying to convert Float8_e4m3fn to the MPS backend but it does not have support for that dtype.

[ZeeMenng]

换了各种weight_type都不行。

[wailovet]

关掉fp8_fast_mode可以吗

[ZeeMenng]

不行，关掉后显示RuntimeError: unsupported scalarType

got prompt
transformer type: 5b
GGUF: False
model weight dtype: torch.float8_e4m3fn manual cast dtype: torch.float16
Encoded latents shape: torch.Size([1, 1, 16, 60, 90])
/opt/homebrew/Caskroom/miniconda/base/lib/python3.12/site-packages/transformers/tokenization_utils_base.py:1601: FutureWarning: clean_up_tokenization_spaces was not set. It will be set to True by default. This behavior will be depracted in transformers v4.45, and will be then set to False by default. For more details check this issue: huggingface/transformers#31884
warnings.warn(
Requested to load SD3ClipModel_
Loading 1 new model
loaded completely 0.0 4541.693359375 True
!!! Exception during processing !!! unsupported scalarType
Traceback (most recent call last):
File "/Users/ZeeMenng/Project/ComfyUI/execution.py", line 323, in execute
output_data, output_ui, has_subgraph = get_output_data(obj, input_data_all, execution_block_cb=execution_block_cb, pre_execute_cb=pre_execute_cb)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/Users/ZeeMenng/Project/ComfyUI/execution.py", line 198, in get_output_data
return_values = _map_node_over_list(obj, input_data_all, obj.FUNCTION, allow_interrupt=True, execution_block_cb=execution_block_cb, pre_execute_cb=pre_execute_cb)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/Users/ZeeMenng/Project/ComfyUI/execution.py", line 169, in _map_node_over_list
process_inputs(input_dict, i)
File "/Users/ZeeMenng/Project/ComfyUI/execution.py", line 158, in process_inputs
results.append(getattr(obj, func)(**inputs))
^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/Users/ZeeMenng/Project/ComfyUI/custom_nodes/ComfyUI-CogVideoXWrapper/nodes.py", line 841, in process
autocast_context = torch.autocast(mm.get_autocast_device(device)) if autocastcondition else nullcontext()
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/opt/homebrew/Caskroom/miniconda/base/lib/python3.12/site-packages/torch/amp/autocast_mode.py", line 229, in init
dtype = torch.get_autocast_dtype(device_type)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
RuntimeError: unsupported scalarType

[wailovet]

更新一下ComfyUI-CogVideoXWrapper试试

[ZeeMenng]

全都升级过了，一样的错误，试了各种办法不行，很奇怪。

我的是M1 Pro，Python我用3.12和3.11两个版本分别作了尝试，都不行。不知道哪个依赖包版本有问题，还是怎样的。

[wailovet]

目前看起来支持应该有些问题, 可以看 kijai/ComfyUI-CogVideoXWrapper#59

[YAY-3M-TA3]

For MPS, you can try:
in CogVideoXLoader node in ComfyUI, set
weight_type to bf16
enable_vae_encode_tiling to true

then a code change:
custom_cogvideox_transformer_3d.py

Line: 103, change to this:
query.to(torch.bfloat16), key.to(torch.bfloat16), value.to(torch.bfloat16), attn_mask=attention_mask, dropout_p=0.0, is_causal=False

[LONGG1126]

你解决这个问题了吗

[lvboy1]

我是m1max运行flux模型的时候出现这个问题，解决方式是打开comfyui的时候用cpu运行
python main.py --force-fp16 --use-split-cross-attention --cpu
切换cpu运行以后可以完整运行流程，不过速度很慢很慢=。=
感觉是mac支持的问题，要是想快只能换电脑