RuntimeError: xformers::efficient_attention_forward_cutlass() expected at most 8 argument(s) but received 13 argument(s).

CUDA SETUP: Loading binary E:\AI\kohya_ss\venv\lib\site-packages\bitsandbytes\libbitsandbytes_cuda116.dll...
use 8-bit AdamW optimizer | {}
running training / 学習開始
num train images * repeats / 学習画像の数×繰り返し回数: 675
num reg images / 正則化画像の数: 0
num batches per epoch / 1epochのバッチ数: 338
num epochs / epoch数: 5
batch size per device / バッチサイズ: 2
gradient accumulation steps / 勾配を合計するステップ数 = 1
total optimization steps / 学習ステップ数: 1688
steps: 0%| | 0/1688 [00:00
train(args)
File "E:\AI\kohya_ss\train_network.py", line 583, in train
noise_pred = unet(noisy_latents, timesteps, encoder_hidden_states).sample
File "E:\AI\kohya_ss\venv\lib\site-packages\torch\nn\modules\module.py", line 1130, in _call_impl
return forward_call(*input, **kwargs)
File "E:\AI\kohya_ss\venv\lib\site-packages\accelerate\utils\operations.py", line 490, in __call__
return convert_to_fp32(self.model_forward(*args, **kwargs))
File "E:\AI\kohya_ss\venv\lib\site-packages\torch\amp\autocast_mode.py", line 12, in decorate_autocast
return func(*args, **kwargs)
File "E:\AI\kohya_ss\venv\lib\site-packages\diffusers\models\unet_2d_condition.py", line 381, in forward
sample, res_samples = downsample_block(
File "E:\AI\kohya_ss\venv\lib\site-packages\torch\nn\modules\module.py", line 1130, in _call_impl
return forward_call(*input, **kwargs)
File "E:\AI\kohya_ss\venv\lib\site-packages\diffusers\models\unet_2d_blocks.py", line 612, in forward
hidden_states = attn(hidden_states, encoder_hidden_states=encoder_hidden_states).sample
File "E:\AI\kohya_ss\venv\lib\site-packages\torch\nn\modules\module.py", line 1130, in _call_impl
return forward_call(*input, **kwargs)
File "E:\AI\kohya_ss\venv\lib\site-packages\diffusers\models\attention.py", line 216, in forward
hidden_states = block(hidden_states, context=encoder_hidden_states, timestep=timestep)
File "E:\AI\kohya_ss\venv\lib\site-packages\torch\nn\modules\module.py", line 1130, in _call_impl
return forward_call(*input, **kwargs)
File "E:\AI\kohya_ss\venv\lib\site-packages\diffusers\models\attention.py", line 484, in forward
hidden_states = self.attn1(norm_hidden_states) + hidden_states
File "E:\AI\kohya_ss\venv\lib\site-packages\torch\nn\modules\module.py", line 1130, in _call_impl
return forward_call(*input, **kwargs)
File "E:\AI\kohya_ss\library\train_util.py", line 1792, in forward_xformers
out = xformers.ops.memory_efficient_attention(q, k, v, attn_bias=None) # 最適なのを選んでくれる
File "E:\AI\kohya_ss\venv\lib\site-packages\xformers\ops\fmha\__init__.py", line 196, in memory_efficient_attention
return _memory_efficient_attention(
File "E:\AI\kohya_ss\venv\lib\site-packages\xformers\ops\fmha\__init__.py", line 299, in _memory_efficient_attention
return _fMHA.apply(
File "E:\AI\kohya_ss\venv\lib\site-packages\xformers\ops\fmha\__init__.py", line 41, in forward
out, op_ctx = _memory_efficient_attention_forward_requires_grad(
File "E:\AI\kohya_ss\venv\lib\site-packages\xformers\ops\fmha\__init__.py", line 327, in _memory_efficient_attention_forward_requires_grad
out = op.apply(inp, needs_gradient=True)
File "E:\AI\kohya_ss\venv\lib\site-packages\xformers\ops\fmha\cutlass.py", line 175, in apply
out, lse, rng_seed, rng_offset = cls.OPERATOR(
File "E:\AI\kohya_ss\venv\lib\site-packages\torch\_ops.py", line 143, in __call__
return self._op(*args, **kwargs or {})
RuntimeError: xformers::efficient_attention_forward_cutlass() expected at most 8 argument(s) but received 13 argument(s). Declaration: xformers::efficient_attention_forward_cutlass(Tensor query, Tensor key, Tensor value, Tensor? cu_seqlens_q, Tensor? cu_seqlens_k, int? max_seqlen_q, bool compute_logsumexp, bool causal) -> (Tensor, Tensor)
steps: 0%| | 0/1688 [00:09
File "E:\AI\kohya_ss\venv\lib\site-packages\accelerate\commands\accelerate_cli.py", line 45, in main
args.func(args)
File "E:\AI\kohya_ss\venv\lib\site-packages\accelerate\commands\launch.py", line 1104, in launch_command
simple_launcher(args)
File "E:\AI\kohya_ss\venv\lib\site-packages\accelerate\commands\launch.py", line 567, in simple_launcher
raise subprocess.CalledProcessError(returncode=process.returncode, cmd=cmd)

Start at library/train_util.py:1792 and trace the xformers attention call through the installed xformers and PyTorch operator declaration shown in the traceback. Reproduce the training run with the reported environment and determine which dependency or call signature is inconsistent. Done means the training path proceeds past the attention operation without this RuntimeError.

Written by the indexing model from the issue text.