NotImplementedError: No operator found for `memory_efficient_attention_forward` with inputs #130

zmy-08 · 2024-04-25T06:55:07Z

zmy-08
Apr 25, 2024

Now I want to use PixArt to train my own 512*512 data set. But I encountered the following problem. Is it because my equipment requirements are not up to standard? I am using V100 32G, so what should I do? How can I run my dataset on this device? Or what device should I use? Looking forward to your answer, thank you! ! !

(pixart) 2023028046zmy@sdnu-SYS-4028GR-TR:/abc/l2/zmy/PixArt-alpha$ python -m torch.distributed.launch --nproc_per_45 /abc/l2/zmy/PixArt-alpha/train_scripts/train.py /abc/l2/zmy/PixArt-alpha/configs/pixart_config/PixArt_xl2_img512output/train_SAM_256
/icislab/volume4/zmy/Myhome/2023028046zmy/envs/pixart/lib/python3.9/site-packages/torch/distributed/launch.py:181: e torch.distributed.launch is deprecated
and will be removed in future. Use torchrun.
Note that --use-env is set by default in torchrun.
If your script expects `--local-rank` argument to be set, please
change it to read from `os.environ['LOCAL_RANK']` instead. See
https://pytorch.org/docs/stable/distributed.html#launch-utility for
further instructions

  warnings.warn(
[2024-04-25 10:39:42,159] torch.distributed.run: [WARNING]
[2024-04-25 10:39:42,159] torch.distributed.run: [WARNING] *****************************************
[2024-04-25 10:39:42,159] torch.distributed.run: [WARNING] Setting OMP_NUM_THREADS environment variable for each pr, to avoid your system being overloaded, please further tune the variable for optimal performance in your applicati
[2024-04-25 10:39:42,159] torch.distributed.run: [WARNING] *****************************************
WARNING[XFORMERS]: xFormers can't load C++/CUDA extensions. xFormers was built for:
    PyTorch 2.1.1+cu121 with CUDA 1201 (you have 2.1.1+cu118)
    Python  3.9.18 (you have 3.9.0)
  Please reinstall xformers (see https://github.com/facebookresearch/xformers#installing-xformers)
  Memory-efficient attention, SwiGLU, sparse and more won't be available.
  Set XFORMERS_MORE_DETAILS=1 for more details
WARNING[XFORMERS]: xFormers can't load C++/CUDA extensions. xFormers was built for:
    PyTorch 2.1.1+cu121 with CUDA 1201 (you have 2.1.1+cu118)
    Python  3.9.18 (you have 3.9.0)
  Please reinstall xformers (see https://github.com/facebookresearch/xformers#installing-xformers)
  Memory-efficient attention, SwiGLU, sparse and more won't be available.
  Set XFORMERS_MORE_DETAILS=1 for more details
/icislab/volume4/zmy/Myhome/2023028046zmy/envs/pixart/lib/python3.9/site-packages/mmcv/__init__.py:20: UserWarning:V will release v2.0.0, in which it will remove components related to the training process and add a data transforma, it will rename the package names mmcv to mmcv-lite and mmcv-full to mmcv. See https://github.com/open-mmlab/mmcv/atibility.md for more details.
  warnings.warn(
/icislab/volume4/zmy/Myhome/2023028046zmy/envs/pixart/lib/python3.9/site-packages/mmcv/__init__.py:20: UserWarning:V will release v2.0.0, in which it will remove components related to the training process and add a data transforma, it will rename the package names mmcv to mmcv-lite and mmcv-full to mmcv. See https://github.com/open-mmlab/mmcv/atibility.md for more details.
  warnings.warn(
Detected kernel version 4.15.0, which is below the recommended minimum of 5.5.0; this can cause the process to hangpgrade the kernel to the minimum version or higher.
2024-04-25 10:39:47,967 - PixArt - INFO - Config:
data_root = '/abc/l2/zmy/PixArt-alpha/data/food_test'
data = dict(
    type='InternalData',
    root='InternData',
    image_list_json=['data_info.json'],
    transform='default_train',
    load_vae_feat=True)
image_size = 512
train_batch_size = 1
eval_batch_size = 16
use_fsdp = False
valid_num = 0
model = 'PixArt_XL_2'
aspect_ratio_type = None
multi_scale = False
lewei_scale = 1.0
num_workers = 10
train_sampling_steps = 1000
eval_sampling_steps = 200
model_max_length = 120
lora_rank = 4
num_epochs = 200
gradient_accumulation_steps = 1
grad_checkpointing = True
gradient_clip = 0.01
gc_step = 1
auto_lr = dict(rule='sqrt')
optimizer = dict(type='AdamW', lr=2e-05, weight_decay=0.03, eps=1e-10)
lr_schedule = 'constant'
lr_schedule_args = dict(num_warmup_steps=1000)
save_image_epochs = 1
save_model_epochs = 1
save_model_steps = 1000000
sample_posterior = True
mixed_precision = 'fp16'
scale_factor = 0.18215
ema_rate = 0.9999
tensorboard_mox_interval = 50
log_interval = 20
cfg_scale = 4
mask_type = 'null'
num_group_tokens = 0
mask_loss_coef = 0.0
load_mask_index = False
vae_pretrained = 'output/pretrained_models/sd-vae-ft-ema'
load_from = None
resume_from = dict(
    checkpoint=None,
    load_ema=False,
    resume_optimizer=True,
    resume_lr_scheduler=True)
snr_loss = False
work_dir = 'output/train_SAM_256'
s3_work_dir = None
seed = 43
image_list_json = ['data_info.json']
window_block_indexes = []
window_size = 0
use_rel_pos = False
fp32_attention = True

2024-04-25 10:39:47,968 - PixArt - INFO - World_size: 2, seed: 43
2024-04-25 10:39:47,968 - PixArt - INFO - Initializing: DDP for training
2024-04-25 10:39:59,446 - PixArt - WARNING - lewei scale: (1.0,), base size: 32
2024-04-25 10:39:59,461 - PixArt - INFO - PixArt Model Parameters: 610,856,096
2024-04-25 10:40:03,037 - PixArt - INFO - Constructing dataset InternalData...
2024-04-25 10:40:03,038 - PixArt - INFO - T5 max token length: 120
2024-04-25 10:40:03,039 - PixArt - INFO - Dataset InternalData constructed. time: 0.00 s, length (use/ori): 5/5
2024-04-25 10:40:03,040 - PixArt - INFO - Automatically adapt lr to 0.0000018 (using sqrt scaling rule).
2024-04-25 10:40:03,120 - PixArt - INFO - Optimizer: total 435 param groups, 435 are learnable, 0 are fix. Lr group0000; Weight decay group: 435 params with weight decay 0.03.
2024-04-25 10:40:03,120 - PixArt - INFO - Lr schedule: constant, num_warmup_steps:1000.
Traceback (most recent call last):
  File "/abc/l2/zmy/PixArt-alpha/train_scripts/train.py", line 336, in <module>
    train()
  File "/abc/l2/zmy/PixArt-alpha/train_scripts/train.py", line 93, in train
    loss_term = train_diffusion.training_losses(model, clean_images, timesteps, model_kwargs=dict(y=y, mask=y_mask,
  File "/abc/l2/zmy/PixArt-alpha/diffusion/model/respace.py", line 94, in training_losses
    return super().training_losses(self._wrap_model(model), *args, **kwargs)
  File "/abc/l2/zmy/PixArt-alpha/diffusion/model/gaussian_diffusion.py", line 750, in training_losses
    model_output = model(x_t, t, **model_kwargs)
  File "/abc/l2/zmy/PixArt-alpha/diffusion/model/respace.py", line 131, in __call__
    return self.model(x, timestep=new_ts, **kwargs)
  File "/icislab/volume4/zmy/Myhome/2023028046zmy/envs/pixart/lib/python3.9/site-packages/torch/nn/modules/module.ped_call_impl
    return self._call_impl(*args, **kwargs)
  File "/icislab/volume4/zmy/Myhome/2023028046zmy/envs/pixart/lib/python3.9/site-packages/torch/nn/modules/module.pimpl
    return forward_call(*args, **kwargs)
  File "/icislab/volume4/zmy/Myhome/2023028046zmy/envs/pixart/lib/python3.9/site-packages/torch/nn/parallel/distribforward
    else self._run_ddp_forward(*inputs, **kwargs)
  File "/icislab/volume4/zmy/Myhome/2023028046zmy/envs/pixart/lib/python3.9/site-packages/torch/nn/parallel/distrib_run_ddp_forward
Traceback (most recent call last):
  File "/abc/l2/zmy/PixArt-alpha/train_scripts/train.py", line 336, in <module>
    train()
  File "/abc/l2/zmy/PixArt-alpha/train_scripts/train.py", line 93, in train
    return self.module(*inputs, **kwargs)  # type: ignore[index]
  File "/icislab/volume4/zmy/Myhome/2023028046zmy/envs/pixart/lib/python3.9/site-packages/torch/nn/modules/module.ped_call_impl
    loss_term = train_diffusion.training_losses(model, clean_images, timesteps, model_kwargs=dict(y=y, mask=y_mask,
  File "/abc/l2/zmy/PixArt-alpha/diffusion/model/respace.py", line 94, in training_losses
    return super().training_losses(self._wrap_model(model), *args, **kwargs)
  File "/abc/l2/zmy/PixArt-alpha/diffusion/model/gaussian_diffusion.py", line 750, in training_losses
    model_output = model(x_t, t, **model_kwargs)
  File "/abc/l2/zmy/PixArt-alpha/diffusion/model/respace.py", line 131, in __call__
    return self._call_impl(*args, **kwargs)
  File "/icislab/volume4/zmy/Myhome/2023028046zmy/envs/pixart/lib/python3.9/site-packages/torch/nn/modules/module.pimpl
    return self.model(x, timestep=new_ts, **kwargs)
  File "/icislab/volume4/zmy/Myhome/2023028046zmy/envs/pixart/lib/python3.9/site-packages/torch/nn/modules/module.ped_call_impl
    return self._call_impl(*args, **kwargs)
  File "/icislab/volume4/zmy/Myhome/2023028046zmy/envs/pixart/lib/python3.9/site-packages/torch/nn/modules/module.pimpl
    return forward_call(*args, **kwargs)
  File "/icislab/volume4/zmy/Myhome/2023028046zmy/envs/pixart/lib/python3.9/site-packages/accelerate/utils/operatioward
    return model_forward(*args, **kwargs)
  File "/icislab/volume4/zmy/Myhome/2023028046zmy/envs/pixart/lib/python3.9/site-packages/accelerate/utils/operatioall__
    return forward_call(*args, **kwargs)
  File "/icislab/volume4/zmy/Myhome/2023028046zmy/envs/pixart/lib/python3.9/site-packages/torch/nn/parallel/distribforward
    return convert_to_fp32(self.model_forward(*args, **kwargs))
  File "/icislab/volume4/zmy/Myhome/2023028046zmy/envs/pixart/lib/python3.9/site-packages/torch/amp/autocast_mode.p_autocast
    return func(*args, **kwargs)
  File "/abc/l2/zmy/PixArt-alpha/diffusion/model/nets/PixArt.py", line 134, in forward
    else self._run_ddp_forward(*inputs, **kwargs)
  File "/icislab/volume4/zmy/Myhome/2023028046zmy/envs/pixart/lib/python3.9/site-packages/torch/nn/parallel/distrib_run_ddp_forward
    x = auto_grad_checkpoint(block, x, y, t0, y_lens)  # (N, T, D) #support grad checkpoint
  File "/abc/l2/zmy/PixArt-alpha/diffusion/model/utils.py", line 41, in auto_grad_checkpoint
    return checkpoint(module, *args, **kwargs)
  File "/icislab/volume4/zmy/Myhome/2023028046zmy/envs/pixart/lib/python3.9/site-packages/torch/_compile.py", line
    return torch._dynamo.disable(fn, recursive)(*args, **kwargs)
  File "/icislab/volume4/zmy/Myhome/2023028046zmy/envs/pixart/lib/python3.9/site-packages/torch/_dynamo/eval_frame.
    return self.module(*inputs, **kwargs)  # type: ignore[index]
  File "/icislab/volume4/zmy/Myhome/2023028046zmy/envs/pixart/lib/python3.9/site-packages/torch/nn/modules/module.ped_call_impl
    return fn(*args, **kwargs)
  File "/icislab/volume4/zmy/Myhome/2023028046zmy/envs/pixart/lib/python3.9/site-packages/torch/_dynamo/external_uter
    return fn(*args, **kwargs)
  File "/icislab/volume4/zmy/Myhome/2023028046zmy/envs/pixart/lib/python3.9/site-packages/torch/utils/checkpoint.pynt
    return CheckpointFunction.apply(function, preserve, *args)
  File "/icislab/volume4/zmy/Myhome/2023028046zmy/envs/pixart/lib/python3.9/site-packages/torch/autograd/function.p
    return self._call_impl(*args, **kwargs)
  File "/icislab/volume4/zmy/Myhome/2023028046zmy/envs/pixart/lib/python3.9/site-packages/torch/nn/modules/module.pimpl
    return super().apply(*args, **kwargs)  # type: ignore[misc]
  File "/icislab/volume4/zmy/Myhome/2023028046zmy/envs/pixart/lib/python3.9/site-packages/torch/utils/checkpoint.py
    outputs = run_function(*args)
  File "/icislab/volume4/zmy/Myhome/2023028046zmy/envs/pixart/lib/python3.9/site-packages/torch/nn/modules/module.ped_call_impl
    return forward_call(*args, **kwargs)
  File "/icislab/volume4/zmy/Myhome/2023028046zmy/envs/pixart/lib/python3.9/site-packages/accelerate/utils/operatioward
    return model_forward(*args, **kwargs)
  File "/icislab/volume4/zmy/Myhome/2023028046zmy/envs/pixart/lib/python3.9/site-packages/accelerate/utils/operatioall__
    return self._call_impl(*args, **kwargs)
  File "/icislab/volume4/zmy/Myhome/2023028046zmy/envs/pixart/lib/python3.9/site-packages/torch/nn/modules/module.pimpl
    return convert_to_fp32(self.model_forward(*args, **kwargs))
  File "/icislab/volume4/zmy/Myhome/2023028046zmy/envs/pixart/lib/python3.9/site-packages/torch/amp/autocast_mode.p_autocast
    return func(*args, **kwargs)
  File "/abc/l2/zmy/PixArt-alpha/diffusion/model/nets/PixArt.py", line 134, in forward
    return forward_call(*args, **kwargs)
  File "/abc/l2/zmy/PixArt-alpha/diffusion/model/nets/PixArt.py", line 50, in forward
    x = auto_grad_checkpoint(block, x, y, t0, y_lens)  # (N, T, D) #support grad checkpoint
  File "/abc/l2/zmy/PixArt-alpha/diffusion/model/utils.py", line 41, in auto_grad_checkpoint
    x = x + self.drop_path(gate_msa * self.attn(t2i_modulate(self.norm1(x), shift_msa, scale_msa)).reshape(B, N, C)
  File "/icislab/volume4/zmy/Myhome/2023028046zmy/envs/pixart/lib/python3.9/site-packages/torch/nn/modules/module.ped_call_impl
    return checkpoint(module, *args, **kwargs)
  File "/icislab/volume4/zmy/Myhome/2023028046zmy/envs/pixart/lib/python3.9/site-packages/torch/_compile.py", line
    return torch._dynamo.disable(fn, recursive)(*args, **kwargs)
  File "/icislab/volume4/zmy/Myhome/2023028046zmy/envs/pixart/lib/python3.9/site-packages/torch/_dynamo/eval_frame.
    return fn(*args, **kwargs)
  File "/icislab/volume4/zmy/Myhome/2023028046zmy/envs/pixart/lib/python3.9/site-packages/torch/_dynamo/external_uter
    return self._call_impl(*args, **kwargs)
  File "/icislab/volume4/zmy/Myhome/2023028046zmy/envs/pixart/lib/python3.9/site-packages/torch/nn/modules/module.pimpl
    return fn(*args, **kwargs)
  File "/icislab/volume4/zmy/Myhome/2023028046zmy/envs/pixart/lib/python3.9/site-packages/torch/utils/checkpoint.pynt
    return CheckpointFunction.apply(function, preserve, *args)
  File "/icislab/volume4/zmy/Myhome/2023028046zmy/envs/pixart/lib/python3.9/site-packages/torch/autograd/function.p
    return forward_call(*args, **kwargs)
  File "/abc/l2/zmy/PixArt-alpha/diffusion/model/nets/PixArt_blocks.py", line 120, in forward
    x = xformers.ops.memory_efficient_attention(q, k, v, p=self.attn_drop.p, attn_bias=attn_bias)
  File "/icislab/volume4/zmy/Myhome/2023028046zmy/envs/pixart/lib/python3.9/site-packages/xformers/ops/fmha/__init_ry_efficient_attention
    return super().apply(*args, **kwargs)  # type: ignore[misc]
      File "/icislab/volume4/zmy/Myhome/2023028046zmy/envs/pixart/lib/python3.9/site-packages/torch/utils/checkpoinard
return _memory_efficient_attention(
  File "/icislab/volume4/zmy/Myhome/2023028046zmy/envs/pixart/lib/python3.9/site-packages/xformers/ops/fmha/__init_ory_efficient_attention
    return _memory_efficient_attention_forward(
    outputs = run_function(*args)  File "/icislab/volume4/zmy/Myhome/2023028046zmy/envs/pixart/lib/python3.9/site-pa/__init__.py", line 337, in _memory_efficient_attention_forward

  File "/icislab/volume4/zmy/Myhome/2023028046zmy/envs/pixart/lib/python3.9/site-packages/torch/nn/modules/module.ped_call_impl
    op = _dispatch_fw(inp, False)
  File "/icislab/volume4/zmy/Myhome/2023028046zmy/envs/pixart/lib/python3.9/site-packages/xformers/ops/fmha/dispatcpatch_fw
    return _run_priority_list(
  File "/icislab/volume4/zmy/Myhome/2023028046zmy/envs/pixart/lib/python3.9/site-packages/xformers/ops/fmha/dispatcpriority_list
    raise NotImplementedError(msg)
NotImplementedError: No operator found for `memory_efficient_attention_forward` with inputs:
     query       : shape=(1, 1024, 16, 72) (torch.float32)
     key         : shape=(1, 1024, 16, 72) (torch.float32)
     value       : shape=(1, 1024, 16, 72) (torch.float32)
     attn_bias   : <class 'NoneType'>
     p           : 0.0
`decoderF` is not supported because:
    xFormers wasn't build with CUDA support
    attn_bias type is <class 'NoneType'>
    operator wasn't built - see `python -m xformers.info` for more info
`[email protected]` is not supported because:
    xFormers wasn't build with CUDA support
    requires device with capability > (8, 0) but your GPU has capability (7, 0) (too old)
    dtype=torch.float32 (supported: {torch.float16, torch.bfloat16})
    operator wasn't built - see `python -m xformers.info` for more info
`tritonflashattF` is not supported because:
    xFormers wasn't build with CUDA support
    requires device with capability > (8, 0) but your GPU has capability (7, 0) (too old)
    dtype=torch.float32 (supported: {torch.float16, torch.bfloat16})
    operator wasn't built - see `python -m xformers.info` for more info
    triton is not available
    requires GPU with sm80 minimum compute capacity, e.g., A100/H100/L4
    Only work on pre-MLIR triton for now
`cutlassF` is not supported because:
    xFormers wasn't build with CUDA support
    operator wasn't built - see `python -m xformers.info` for more info
`smallkF` is not supported because:
    max(query.shape[-1] != value.shape[-1]) > 32
    xFormers wasn't build with CUDA support
    operator wasn't built - see `python -m xformers.info` for more info
    unsupported embed per head: 72
    return self._call_impl(*args, **kwargs)
  File "/icislab/volume4/zmy/Myhome/2023028046zmy/envs/pixart/lib/python3.9/site-packages/torch/nn/modules/module.pimpl
    return forward_call(*args, **kwargs)
  File "/abc/l2/zmy/PixArt-alpha/diffusion/model/nets/PixArt.py", line 50, in forward
    x = x + self.drop_path(gate_msa * self.attn(t2i_modulate(self.norm1(x), shift_msa, scale_msa)).reshape(B, N, C)
  File "/icislab/volume4/zmy/Myhome/2023028046zmy/envs/pixart/lib/python3.9/site-packages/torch/nn/modules/module.ped_call_impl
    return self._call_impl(*args, **kwargs)
  File "/icislab/volume4/zmy/Myhome/2023028046zmy/envs/pixart/lib/python3.9/site-packages/torch/nn/modules/module.pimpl
    return forward_call(*args, **kwargs)
  File "/abc/l2/zmy/PixArt-alpha/diffusion/model/nets/PixArt_blocks.py", line 120, in forward
    x = xformers.ops.memory_efficient_attention(q, k, v, p=self.attn_drop.p, attn_bias=attn_bias)
  File "/icislab/volume4/zmy/Myhome/2023028046zmy/envs/pixart/lib/python3.9/site-packages/xformers/ops/fmha/__init_ry_efficient_attention
    return _memory_efficient_attention(
  File "/icislab/volume4/zmy/Myhome/2023028046zmy/envs/pixart/lib/python3.9/site-packages/xformers/ops/fmha/__init_ory_efficient_attention
    return _memory_efficient_attention_forward(
  File "/icislab/volume4/zmy/Myhome/2023028046zmy/envs/pixart/lib/python3.9/site-packages/xformers/ops/fmha/__init_ory_efficient_attention_forward
    op = _dispatch_fw(inp, False)
  File "/icislab/volume4/zmy/Myhome/2023028046zmy/envs/pixart/lib/python3.9/site-packages/xformers/ops/fmha/dispatcpatch_fw
    return _run_priority_list(
  File "/icislab/volume4/zmy/Myhome/2023028046zmy/envs/pixart/lib/python3.9/site-packages/xformers/ops/fmha/dispatcpriority_list
    raise NotImplementedError(msg)
NotImplementedError: No operator found for `memory_efficient_attention_forward` with inputs:
     query       : shape=(1, 1024, 16, 72) (torch.float32)
     key         : shape=(1, 1024, 16, 72) (torch.float32)
     value       : shape=(1, 1024, 16, 72) (torch.float32)
     attn_bias   : <class 'NoneType'>
     p           : 0.0
`decoderF` is not supported because:
    xFormers wasn't build with CUDA support
    attn_bias type is <class 'NoneType'>
    operator wasn't built - see `python -m xformers.info` for more info
`[email protected]` is not supported because:
    xFormers wasn't build with CUDA support
    requires device with capability > (8, 0) but your GPU has capability (7, 0) (too old)
    dtype=torch.float32 (supported: {torch.float16, torch.bfloat16})
    operator wasn't built - see `python -m xformers.info` for more info
`tritonflashattF` is not supported because:
    xFormers wasn't build with CUDA support
    requires device with capability > (8, 0) but your GPU has capability (7, 0) (too old)
    dtype=torch.float32 (supported: {torch.float16, torch.bfloat16})
    operator wasn't built - see `python -m xformers.info` for more info
    triton is not available
    requires GPU with sm80 minimum compute capacity, e.g., A100/H100/L4
    Only work on pre-MLIR triton for now
`cutlassF` is not supported because:
    xFormers wasn't build with CUDA support
    operator wasn't built - see `python -m xformers.info` for more info
`smallkF` is not supported because:
    max(query.shape[-1] != value.shape[-1]) > 32
    xFormers wasn't build with CUDA support
    operator wasn't built - see `python -m xformers.info` for more info
    unsupported embed per head: 72

zmy-08 · 2024-04-25T11:35:35Z

zmy-08
Apr 25, 2024
Author

My gpu compute power is 7, for this training pixArt always error. This is my statement: python -m torch.distributed.launch --nproc_per_node=2 --master_port=12345 /abc/l2/zmy/PixArt-alpha /train_scripts/train. py /abc/l2/zmy/PixArt-alpha/configs/pixart_config/PixArt_xl2_img512_internal.py --work-dir output/train_A
Should I change the server? I have 4 V100 32g ,if training pixArt_lcm, I wonder if it is feasible? I see that lcm is only training 1024 images, if I want to train 512 images, is it ok to just modify the InternalData.py file under datasets 、 PixArt_xl2_img1024_lcm.py under configs? I just changed the image_size from 1024 to 512 and some basic changes, but it still reports an error.

 (pixart) 2023028046zmy@sdnu-SYS-4028GR-TR:/abc/l2/zmy/PixArt-alpha$ python -m xformers.info                  WARNING[XFORMERS]: xFormers can't load C++/CUDA extensions. xFormers was built for:
    PyTorch 2.1.1+cu121 with CUDA 1201 (you have 2.1.1+cu118)
    Python  3.9.18 (you have 3.9.0)
  Please reinstall xformers (see https://github.com/facebookresearch/xformers#installing-xformers)
  Memory-efficient attention, SwiGLU, sparse and more won't be available.
  Set XFORMERS_MORE_DETAILS=1 for more details
xFormers 0.0.23
memory_efficient_attention.cutlassF:               unavailable
memory_efficient_attention.cutlassB:               unavailable
memory_efficient_attention.decoderF:               unavailable
[email protected]:         unavailable
[email protected]:         unavailable
memory_efficient_attention.smallkF:                unavailable
memory_efficient_attention.smallkB:                unavailable
memory_efficient_attention.tritonflashattF:        unavailable
memory_efficient_attention.tritonflashattB:        unavailable
memory_efficient_attention.triton_splitKF:         available
indexing.scaled_index_addF:                        available
indexing.scaled_index_addB:                        available
indexing.index_select:                             available
swiglu.dual_gemm_silu:                             unavailable
swiglu.gemm_fused_operand_sum:                     unavailable
swiglu.fused.p.cpp:                                not built
is_triton_available:                               True
pytorch.version:                                   2.1.1+cu118
pytorch.cuda:                                      available
gpu.compute_capability:                            7.0
gpu.name:                                          Tesla V100-PCIE-32GB
build.info:                                        available
build.cuda_version:                                1201
build.python_version:                              3.9.18
build.torch_version:                               2.1.1+cu121
build.env.TORCH_CUDA_ARCH_LIST:                    5.0+PTX 6.0 6.1 7.0 7.5 8.0+PTX 9.0
build.env.XFORMERS_BUILD_TYPE:                     Release
build.env.XFORMERS_ENABLE_DEBUG_ASSERTIONS:        None
build.env.NVCC_FLAGS:                              None
build.env.XFORMERS_PACKAGE_FROM:                   wheel-v0.0.23
source.privacy:                                    open source

0 replies

lawrence-cj · 2024-04-26T06:53:28Z

lawrence-cj
Apr 26, 2024
Maintainer

Can you try xformers==0.19.0 and torch==2.0u117?

0 replies

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

PixArt

NotImplementedError: No operator found for `memory_efficient_attention_forward` with inputs #130

{{title}}

Replies: 2 comments

{{title}}

{{title}}

Select a reply

PixArt

NotImplementedError: No operator found for memory_efficient_attention_forward with inputs #130

zmy-08 Apr 25, 2024

Replies: 2 comments

zmy-08 Apr 25, 2024 Author

lawrence-cj Apr 26, 2024 Maintainer

NotImplementedError: No operator found for `memory_efficient_attention_forward` with inputs #130

zmy-08
Apr 25, 2024

zmy-08
Apr 25, 2024
Author

lawrence-cj
Apr 26, 2024
Maintainer