[BUG]: LowLevelZeroPlugin with use_fp8=true causes `shape mismatch error: shape [32, 512] is invalid for input of size 512`

[sdjasj]

Is there an existing issue for this bug?

• I have searched the existing issues

The bug has not been fixed in the latest main branch

• I have checked the latest main branch

Do you feel comfortable sharing a concise (minimal) script that reproduces the error? :)

Yes, I will share a minimal reproducible script.

🐛 Describe the bug

It seems that the fp8 feature in LowLevelZeroPlugin is not working properly, as it raises the following error:
shape '[32, 512]' is invalid for input of size 512.

The following script can reproduce the issue. In this script, use_fp8=true is set when using LowLevelZeroPlugin:

import torch
import torch.nn as nn
import torch.optim as optim
from torch.utils.data import Dataset, DataLoader
import colossalai
from colossalai.booster import Booster
from colossalai.nn.optimizer import CPUAdam, Lamb
from colossalai.booster.plugin import GeminiPlugin, LowLevelZeroPlugin

class RandomDataset(Dataset):
    def __init__(self, num_samples=32 * 100, input_dim=1024, num_classes=10):
        self.x = torch.randn(num_samples, input_dim)
        self.y = torch.randint(0, num_classes, (num_samples,))

    def __len__(self):
        return len(self.x)

    def __getitem__(self, idx):
        return self.x[idx], self.y[idx]


class MLP(nn.Module):
    def __init__(self, input_dim=1024, hidden_dim=512, num_layers=10, num_classes=10):
        super().__init__()
        layers = []
        for i in range(num_layers):
            in_dim = input_dim if i == 0 else hidden_dim
            layers.append(nn.Linear(in_dim, hidden_dim))
            layers.append(nn.ReLU())
        layers.append(nn.Linear(hidden_dim, num_classes))
        self.net = nn.Sequential(*layers)

    def forward(self, x):
        return self.net(x)



def main():
    seed = 1024
    colossalai.launch_from_torch(seed=seed)
    plugin = LowLevelZeroPlugin(
        use_fp8=True
    )

    booster = Booster(plugin=plugin)

    model = MLP()
    optimizer = optim.Adam(model.parameters(), lr=1e-3)
    criterion = nn.CrossEntropyLoss()

    dataset = RandomDataset()
    train_dataloader = DataLoader(dataset, batch_size=32, shuffle=False)

    model, optimizer, criterion, train_dataloader, _ = booster.boost(model, optimizer, criterion, train_dataloader)

    precision = getattr(plugin, "precision", "fp16")
    dtype_map = {"fp16": torch.float16, "bf16": torch.bfloat16, "fp32": torch.float32}
    dtype = dtype_map.get(precision, torch.float16)

    device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
    model.train()
    for epoch in range(1):
        total_loss = 0
        for step, (x, y) in enumerate(train_dataloader):

            x = x.to(device=device, dtype=dtype)
            y = y.to(device=device)

            optimizer.zero_grad()
            output = model(x)
            loss = criterion(output, y)

            booster.backward(loss, optimizer)
            optimizer.step()

            total_loss += loss.item()
            print(f"[Epoch {epoch}] step {step}, loss = {loss.item():.4f}")

        avg_loss = total_loss / len(train_dataloader)
        print(f"Epoch {epoch} finished, average loss = {avg_loss:.4f}")


if __name__ == "__main__":
    main()

Running the command:

colossalai run --nproc_per_node 4 main.py

produces the following error log (only a portion is shown due to length):

W1101 19:36:20.780868 936565 site-packages/torch/distributed/run.py:793] 
W1101 19:36:20.780868 936565 site-packages/torch/distributed/run.py:793] *****************************************
W1101 19:36:20.780868 936565 site-packages/torch/distributed/run.py:793] Setting OMP_NUM_THREADS environment variable for each process to be 1 in default, to avoid your system being overloaded, please further tune the variable for optimal performance in your application as needed. 
W1101 19:36:20.780868 936565 site-packages/torch/distributed/run.py:793] *****************************************
[11/01/25 19:36:24] INFO     colossalai - colossalai - INFO:                    
                             /home/yanzhen/miniconda3/envs/colossal/lib/python3.
                             9/site-packages/colossalai/initialize.py:75 launch 
                    INFO     colossalai - colossalai - INFO: Distributed        
                             environment is initialized, world size: 4          
[rank1]:E1101 19:36:25.141706 937254 site-packages/torch/_subclasses/fake_tensor.py:2017] [0/0] failed while attempting to run meta for aten.view.default
[rank1]:E1101 19:36:25.141706 937254 site-packages/torch/_subclasses/fake_tensor.py:2017] [0/0] Traceback (most recent call last):
[rank1]:E1101 19:36:25.141706 937254 site-packages/torch/_subclasses/fake_tensor.py:2017] [0/0]   File "/home/yanzhen/miniconda3/envs/colossal/lib/python3.9/site-packages/torch/_subclasses/fake_tensor.py", line 2013, in _dispatch_impl
[rank1]:E1101 19:36:25.141706 937254 site-packages/torch/_subclasses/fake_tensor.py:2017] [0/0]     r = func(*args, **kwargs)
[rank1]:E1101 19:36:25.141706 937254 site-packages/torch/_subclasses/fake_tensor.py:2017] [0/0]   File "/home/yanzhen/miniconda3/envs/colossal/lib/python3.9/site-packages/torch/_ops.py", line 716, in __call__
[rank1]:E1101 19:36:25.141706 937254 site-packages/torch/_subclasses/fake_tensor.py:2017] [0/0]     return self._op(*args, **kwargs)
[rank1]:E1101 19:36:25.141706 937254 site-packages/torch/_subclasses/fake_tensor.py:2017] [0/0]   File "/home/yanzhen/miniconda3/envs/colossal/lib/python3.9/site-packages/torch/_refs/__init__.py", line 4591, in view
[rank1]:E1101 19:36:25.141706 937254 site-packages/torch/_subclasses/fake_tensor.py:2017] [0/0]     return _reshape_view_helper(a, *shape, allow_copy=False)
[rank1]:E1101 19:36:25.141706 937254 site-packages/torch/_subclasses/fake_tensor.py:2017] [0/0]   File "/home/yanzhen/miniconda3/envs/colossal/lib/python3.9/site-packages/torch/_refs/__init__.py", line 3659, in _reshape_view_helper
[rank1]:E1101 19:36:25.141706 937254 site-packages/torch/_subclasses/fake_tensor.py:2017] [0/0]     shape = utils.infer_size(shape, a.numel())
[rank1]:E1101 19:36:25.141706 937254 site-packages/torch/_subclasses/fake_tensor.py:2017] [0/0]   File "/home/yanzhen/miniconda3/envs/colossal/lib/python3.9/site-packages/torch/_prims_common/__init__.py", line 901, in infer_size
[rank1]:E1101 19:36:25.141706 937254 site-packages/torch/_subclasses/fake_tensor.py:2017] [0/0]     torch._check(
[rank1]:E1101 19:36:25.141706 937254 site-packages/torch/_subclasses/fake_tensor.py:2017] [0/0]   File "/home/yanzhen/miniconda3/envs/colossal/lib/python3.9/site-packages/torch/__init__.py", line 1564, in _check
[rank1]:E1101 19:36:25.141706 937254 site-packages/torch/_subclasses/fake_tensor.py:2017] [0/0]     _check_with(RuntimeError, cond, message)
[rank1]:E1101 19:36:25.141706 937254 site-packages/torch/_subclasses/fake_tensor.py:2017] [0/0]   File "/home/yanzhen/miniconda3/envs/colossal/lib/python3.9/site-packages/torch/__init__.py", line 1546, in _check_with
[rank1]:E1101 19:36:25.141706 937254 site-packages/torch/_subclasses/fake_tensor.py:2017] [0/0]     raise error_type(message_evaluated)
[rank1]:E1101 19:36:25.141706 937254 site-packages/torch/_subclasses/fake_tensor.py:2017] [0/0] RuntimeError: shape '[32, 512]' is invalid for input of size 512
[rank1]: Traceback (most recent call last):
[rank1]:   File "/home/yanzhen/distributed_test/colossalAI/test/bug2.py", line 88, in <module>
[rank1]:     main()
[rank1]:   File "/home/yanzhen/distributed_test/colossalAI/test/bug2.py", line 74, in main
[rank1]:     output = model(x)
[rank1]:   File "/home/yanzhen/miniconda3/envs/colossal/lib/python3.9/site-packages/torch/nn/modules/module.py", line 1736, in _wrapped_call_impl
[rank1]:     return self._call_impl(*args, **kwargs)
[rank1]:   File "/home/yanzhen/miniconda3/envs/colossal/lib/python3.9/site-packages/torch/nn/modules/module.py", line 1747, in _call_impl
[rank1]:     return forward_call(*args, **kwargs)
[rank1]:   File "/home/yanzhen/miniconda3/envs/colossal/lib/python3.9/site-packages/colossalai/booster/plugin/low_level_zero_plugin.py", line 107, in forward
[rank1]:     return super().forward(*args, **kwargs)
[rank1]:   File "/home/yanzhen/miniconda3/envs/colossal/lib/python3.9/site-packages/colossalai/interface/model.py", line 127, in forward
[rank1]:     return self.module(*args, **kwargs)
[rank1]:   File "/home/yanzhen/miniconda3/envs/colossal/lib/python3.9/site-packages/torch/nn/modules/module.py", line 1736, in _wrapped_call_impl
[rank1]:     return self._call_impl(*args, **kwargs)
[rank1]:   File "/home/yanzhen/miniconda3/envs/colossal/lib/python3.9/site-packages/torch/nn/modules/module.py", line 1747, in _call_impl
[rank1]:     return forward_call(*args, **kwargs)
[rank1]:   File "/home/yanzhen/distributed_test/colossalAI/test/bug2.py", line 38, in forward
[rank1]:     return self.net(x)
[rank1]:   File "/home/yanzhen/miniconda3/envs/colossal/lib/python3.9/site-packages/torch/nn/modules/module.py", line 1736, in _wrapped_call_impl
[rank1]:     return self._call_impl(*args, **kwargs)
[rank1]:   File "/home/yanzhen/miniconda3/envs/colossal/lib/python3.9/site-packages/torch/nn/modules/module.py", line 1747, in _call_impl
[rank1]:     return forward_call(*args, **kwargs)
[rank1]:   File "/home/yanzhen/miniconda3/envs/colossal/lib/python3.9/site-packages/torch/nn/modules/container.py", line 250, in forward
[rank1]:     input = module(input)
[rank1]:   File "/home/yanzhen/miniconda3/envs/colossal/lib/python3.9/site-packages/torch/nn/modules/module.py", line 1736, in _wrapped_call_impl
[rank1]:     return self._call_impl(*args, **kwargs)
[rank1]:   File "/home/yanzhen/miniconda3/envs/colossal/lib/python3.9/site-packages/torch/nn/modules/module.py", line 1747, in _call_impl
[rank1]:     return forward_call(*args, **kwargs)
[rank1]:   File "/home/yanzhen/miniconda3/envs/colossal/lib/python3.9/site-packages/torch/nn/modules/linear.py", line 125, in forward
[rank1]:     return F.linear(input, self.weight, self.bias)
[rank1]:   File "/home/yanzhen/miniconda3/envs/colossal/lib/python3.9/site-packages/colossalai/tensor/colo_parameter.py", line 66, in __torch_function__
[rank1]:     ret = super().__torch_function__(func, types, args, kwargs)
[rank1]:   File "/home/yanzhen/miniconda3/envs/colossal/lib/python3.9/site-packages/colossalai/tensor/colo_tensor.py", line 91, in __torch_function__
[rank1]:     ret = func(*args, **kwargs)
[rank1]:   File "/home/yanzhen/miniconda3/envs/colossal/lib/python3.9/site-packages/colossalai/quantization/fp8.py", line 845, in linear_fp8
[rank1]:     out = _linear_fp8(input, weight, bias)
[rank1]:   File "/home/yanzhen/miniconda3/envs/colossal/lib/python3.9/site-packages/torch/_dynamo/eval_frame.py", line 465, in _fn
[rank1]:     return fn(*args, **kwargs)
[rank1]:   File "/home/yanzhen/miniconda3/envs/colossal/lib/python3.9/site-packages/torch/_dynamo/convert_frame.py", line 1269, in __call__
[rank1]:     return self._torchdynamo_orig_callable(
[rank1]:   File "/home/yanzhen/miniconda3/envs/colossal/lib/python3.9/site-packages/torch/_dynamo/convert_frame.py", line 1064, in __call__
[rank1]:     result = self._inner_convert(
[rank1]:   File "/home/yanzhen/miniconda3/envs/colossal/lib/python3.9/site-packages/torch/_dynamo/convert_frame.py", line 526, in __call__
[rank1]:     return _compile(
[rank1]:   File "/home/yanzhen/miniconda3/envs/colossal/lib/python3.9/site-packages/torch/_dynamo/convert_frame.py", line 924, in _compile
[rank1]:     guarded_code = compile_inner(code, one_graph, hooks, transform)
[rank1]:   File "/home/yanzhen/miniconda3/envs/colossal/lib/python3.9/site-packages/torch/_dynamo/convert_frame.py", line 666, in compile_inner
[rank1]:     return _compile_inner(code, one_graph, hooks, transform)
[rank1]:   File "/home/yanzhen/miniconda3/envs/colossal/lib/python3.9/site-packages/torch/_utils_internal.py", line 87, in wrapper_function
[rank1]:     return function(*args, **kwargs)
[rank1]:   File "/home/yanzhen/miniconda3/envs/colossal/lib/python3.9/site-packages/torch/_dynamo/convert_frame.py", line 699, in _compile_inner
[rank1]:     out_code = transform_code_object(code, transform)
[rank1]:   File "/home/yanzhen/miniconda3/envs/colossal/lib/python3.9/site-packages/torch/_dynamo/bytecode_transformation.py", line 1322, in transform_code_object
[rank1]:     transformations(instructions, code_options)
[rank1]:   File "/home/yanzhen/miniconda3/envs/colossal/lib/python3.9/site-packages/torch/_dynamo/convert_frame.py", line 219, in _fn
[rank1]:     return fn(*args, **kwargs)
[rank1]:   File "/home/yanzhen/miniconda3/envs/colossal/lib/python3.9/site-packages/torch/_dynamo/convert_frame.py", line 634, in transform
[rank1]:     tracer.run()
[rank1]:   File "/home/yanzhen/miniconda3/envs/colossal/lib/python3.9/site-packages/torch/_dynamo/symbolic_convert.py", line 2796, in run
[rank1]:     super().run()
[rank1]:   File "/home/yanzhen/miniconda3/envs/colossal/lib/python3.9/site-packages/torch/_dynamo/symbolic_convert.py", line 983, in run
[rank1]:     while self.step():
[rank1]:   File "/home/yanzhen/miniconda3/envs/colossal/lib/python3.9/site-packages/torch/_dynamo/symbolic_convert.py", line 895, in step
[rank1]:     self.dispatch_table[inst.opcode](self, inst)
[rank1]:   File "/home/yanzhen/miniconda3/envs/colossal/lib/python3.9/site-packages/torch/_dynamo/symbolic_convert.py", line 582, in wrapper
[rank1]:     return inner_fn(self, inst)
[rank1]:   File "/home/yanzhen/miniconda3/envs/colossal/lib/python3.9/site-packages/torch/_dynamo/symbolic_convert.py", line 1602, in CALL_FUNCTION
[rank1]:     self.call_function(fn, args, {})
[rank1]:   File "/home/yanzhen/miniconda3/envs/colossal/lib/python3.9/site-packages/torch/_dynamo/symbolic_convert.py", line 830, in call_function
[rank1]:     self.push(fn.call_function(self, args, kwargs))  # type: ignore[arg-type]
[rank1]:   File "/home/yanzhen/miniconda3/envs/colossal/lib/python3.9/site-packages/torch/_dynamo/variables/misc.py", line 1024, in call_function
[rank1]:     return self.obj.call_method(tx, self.name, args, kwargs)
[rank1]:   File "/home/yanzhen/miniconda3/envs/colossal/lib/python3.9/site-packages/torch/_dynamo/variables/misc.py", line 774, in call_method
[rank1]:     return self.call_apply(tx, args, kwargs)
[rank1]:   File "/home/yanzhen/miniconda3/envs/colossal/lib/python3.9/site-packages/torch/_dynamo/variables/misc.py", line 694, in call_apply
[rank1]:     val = AutogradFunctionApplyVariable(
[rank1]:   File "/home/yanzhen/miniconda3/envs/colossal/lib/python3.9/site-packages/torch/_dynamo/variables/higher_order_ops.py", line 2015, in call_function
[rank1]:     (fwd_out, _), fwd_graph, fwd_freevars = speculate_subgraph(
[rank1]:   File "/home/yanzhen/miniconda3/envs/colossal/lib/python3.9/site-packages/torch/_dynamo/variables/higher_order_ops.py", line 462, in speculate_subgraph
[rank1]:     output = f.call_function(tx, args, sub_kwargs)
[rank1]:   File "/home/yanzhen/miniconda3/envs/colossal/lib/python3.9/site-packages/torch/_dynamo/variables/functions.py", line 324, in call_function
[rank1]:     return super().call_function(tx, args, kwargs)
[rank1]:   File "/home/yanzhen/miniconda3/envs/colossal/lib/python3.9/site-packages/torch/_dynamo/variables/functions.py", line 111, in call_function
[rank1]:     return tx.inline_user_function_return(self, [*self.self_args(), *args], kwargs)
[rank1]:   File "/home/yanzhen/miniconda3/envs/colossal/lib/python3.9/site-packages/torch/_dynamo/symbolic_convert.py", line 836, in inline_user_function_return
[rank1]:     return InliningInstructionTranslator.inline_call(self, fn, args, kwargs)
[rank1]:   File "/home/yanzhen/miniconda3/envs/colossal/lib/python3.9/site-packages/torch/_dynamo/symbolic_convert.py", line 3011, in inline_call
[rank1]:     return cls.inline_call_(parent, func, args, kwargs)
[rank1]:   File "/home/yanzhen/miniconda3/envs/colossal/lib/python3.9/site-packages/torch/_dynamo/symbolic_convert.py", line 3139, in inline_call_
[rank1]:     tracer.run()
[rank1]:   File "/home/yanzhen/miniconda3/envs/colossal/lib/python3.9/site-packages/torch/_dynamo/symbolic_convert.py", line 983, in run
[rank1]:     while self.step():
[rank1]:   File "/home/yanzhen/miniconda3/envs/colossal/lib/python3.9/site-packages/torch/_dynamo/symbolic_convert.py", line 895, in step
[rank1]:     self.dispatch_table[inst.opcode](self, inst)
[rank1]:   File "/home/yanzhen/miniconda3/envs/colossal/lib/python3.9/site-packages/torch/_dynamo/symbolic_convert.py", line 582, in wrapper
[rank1]:     return inner_fn(self, inst)
[rank1]:   File "/home/yanzhen/miniconda3/envs/colossal/lib/python3.9/site-packages/torch/_dynamo/symbolic_convert.py", line 1680, in CALL_FUNCTION_EX
[rank1]:     self.call_function(fn, argsvars.items, kwargsvars)
[rank1]:   File "/home/yanzhen/miniconda3/envs/colossal/lib/python3.9/site-packages/torch/_dynamo/symbolic_convert.py", line 830, in call_function
[rank1]:     self.push(fn.call_function(self, args, kwargs))  # type: ignore[arg-type]
[rank1]:   File "/home/yanzhen/miniconda3/envs/colossal/lib/python3.9/site-packages/torch/_dynamo/variables/misc.py", line 1024, in call_function
[rank1]:     return self.obj.call_method(tx, self.name, args, kwargs)
[rank1]:   File "/home/yanzhen/miniconda3/envs/colossal/lib/python3.9/site-packages/torch/_dynamo/variables/tensor.py", line 535, in call_method
[rank1]:     return wrap_fx_proxy(
[rank1]:   File "/home/yanzhen/miniconda3/envs/colossal/lib/python3.9/site-packages/torch/_dynamo/variables/builder.py", line 2037, in wrap_fx_proxy
[rank1]:     return wrap_fx_proxy_cls(target_cls=TensorVariable, **kwargs)
[rank1]:   File "/home/yanzhen/miniconda3/envs/colossal/lib/python3.9/site-packages/torch/_dynamo/variables/builder.py", line 2124, in wrap_fx_proxy_cls
[rank1]:     example_value = get_fake_value(proxy.node, tx, allow_non_graph_fake=True)
[rank1]:   File "/home/yanzhen/miniconda3/envs/colossal/lib/python3.9/site-packages/torch/_dynamo/utils.py", line 2082, in get_fake_value
[rank1]:     raise TorchRuntimeError(str(e)).with_traceback(e.__traceback__) from None
[rank1]:   File "/home/yanzhen/miniconda3/envs/colossal/lib/python3.9/site-packages/torch/_dynamo/utils.py", line 2017, in get_fake_value
[rank1]:     ret_val = wrap_fake_exception(
[rank1]:   File "/home/yanzhen/miniconda3/envs/colossal/lib/python3.9/site-packages/torch/_dynamo/utils.py", line 1574, in wrap_fake_exception
[rank1]:     return fn()
[rank1]:   File "/home/yanzhen/miniconda3/envs/colossal/lib/python3.9/site-packages/torch/_dynamo/utils.py", line 2018, in <lambda>
[rank1]:     lambda: run_node(tx.output, node, args, kwargs, nnmodule)
[rank1]:   File "/home/yanzhen/miniconda3/envs/colossal/lib/python3.9/site-packages/torch/_dynamo/utils.py", line 2150, in run_node
[rank1]:     raise RuntimeError(make_error_message(e)).with_traceback(
[rank1]:   File "/home/yanzhen/miniconda3/envs/colossal/lib/python3.9/site-packages/torch/_dynamo/utils.py", line 2134, in run_node
[rank1]:     return getattr(args[0], node.target)(*args[1:], **kwargs)
[rank1]:   File "/home/yanzhen/miniconda3/envs/colossal/lib/python3.9/site-packages/torch/utils/_stats.py", line 21, in wrapper
[rank1]:     return fn(*args, **kwargs)
[rank1]:   File "/home/yanzhen/miniconda3/envs/colossal/lib/python3.9/site-packages/torch/_subclasses/fake_tensor.py", line 1238, in __torch_dispatch__
[rank1]:     return self.dispatch(func, types, args, kwargs)
[rank1]:   File "/home/yanzhen/miniconda3/envs/colossal/lib/python3.9/site-packages/torch/_subclasses/fake_tensor.py", line 1692, in dispatch
[rank1]:     return self._cached_dispatch_impl(func, types, args, kwargs)
[rank1]:   File "/home/yanzhen/miniconda3/envs/colossal/lib/python3.9/site-packages/torch/_subclasses/fake_tensor.py", line 1339, in _cached_dispatch_impl
[rank1]:     output = self._dispatch_impl(func, types, args, kwargs)
[rank1]:   File "/home/yanzhen/miniconda3/envs/colossal/lib/python3.9/site-packages/torch/_subclasses/fake_tensor.py", line 2013, in _dispatch_impl
[rank1]:     r = func(*args, **kwargs)
[rank1]:   File "/home/yanzhen/miniconda3/envs/colossal/lib/python3.9/site-packages/torch/_ops.py", line 716, in __call__
[rank1]:     return self._op(*args, **kwargs)
[rank1]:   File "/home/yanzhen/miniconda3/envs/colossal/lib/python3.9/site-packages/torch/_refs/__init__.py", line 4591, in view
[rank1]:     return _reshape_view_helper(a, *shape, allow_copy=False)
[rank1]:   File "/home/yanzhen/miniconda3/envs/colossal/lib/python3.9/site-packages/torch/_refs/__init__.py", line 3659, in _reshape_view_helper
[rank1]:     shape = utils.infer_size(shape, a.numel())
[rank1]:   File "/home/yanzhen/miniconda3/envs/colossal/lib/python3.9/site-packages/torch/_prims_common/__init__.py", line 901, in infer_size
[rank1]:     torch._check(
[rank1]:   File "/home/yanzhen/miniconda3/envs/colossal/lib/python3.9/site-packages/torch/__init__.py", line 1564, in _check
[rank1]:     _check_with(RuntimeError, cond, message)
[rank1]:   File "/home/yanzhen/miniconda3/envs/colossal/lib/python3.9/site-packages/torch/__init__.py", line 1546, in _check_with
[rank1]:     raise error_type(message_evaluated)
[rank1]: torch._dynamo.exc.TorchRuntimeError: Failed running call_method reshape(*(FakeTensor(..., device='cuda:1', size=(512,), dtype=torch.float16,
[rank1]:            grad_fn=<SelectBackward0>), 32, 512), **{}):
[rank1]: shape '[32, 512]' is invalid for input of size 512

[rank1]: from user code:
[rank1]:    File "/home/yanzhen/miniconda3/envs/colossal/lib/python3.9/site-packages/colossalai/quantization/fp8.py", line 839, in _linear_fp8
[rank1]:     return _LinearFp8.apply(input, weight, bias)
[rank1]:   File "/home/yanzhen/miniconda3/envs/colossal/lib/python3.9/site-packages/colossalai/quantization/fp8.py", line 809, in forward
[rank1]:     return out.reshape(*ctx.x_shape[:-1], w.shape[0])

[rank1]: Set TORCH_LOGS="+dynamo" and TORCHDYNAMO_VERBOSE=1 for more information


[rank1]: You can suppress this exception and fall back to eager by setting:
[rank1]:     import torch._dynamo
[rank1]:     torch._dynamo.config.suppress_errors = True

Environment

PyTorch version: 2.5.1+cu124
Is debug build: False
CUDA used to build PyTorch: 12.4
ROCM used to build PyTorch: N/A

OS: Ubuntu 22.04.5 LTS (x86_64)
GCC version: (Ubuntu 11.4.0-1ubuntu1~22.04.2) 11.4.0
Clang version: 18.1.3 (1ubuntu1)
CMake version: version 3.28.3
Libc version: glibc-2.39

Python version: 3.9.23 (main, Jun  5 2025, 13:40:20)  [GCC 11.2.0] (64-bit runtime)
Python platform: Linux-6.5.0-18-generic-x86_64-with-glibc2.39
Is CUDA available: True
CUDA runtime version: 12.4.99
CUDA_MODULE_LOADING set to: LAZY
GPU models and configuration: 
GPU 0: NVIDIA GeForce RTX 4090
GPU 1: NVIDIA GeForce RTX 4090
GPU 2: NVIDIA GeForce RTX 4090
GPU 3: NVIDIA GeForce RTX 4090

Nvidia driver version: 580.65.06
cuDNN version: Probably one of the following:
/usr/local/cuda-11.6/targets/x86_64-linux/lib/libcudnn.so.8
/usr/local/cuda-11.6/targets/x86_64-linux/lib/libcudnn_adv_infer.so.8
/usr/local/cuda-11.6/targets/x86_64-linux/lib/libcudnn_adv_train.so.8
/usr/local/cuda-11.6/targets/x86_64-linux/lib/libcudnn_cnn_infer.so.8
/usr/local/cuda-11.6/targets/x86_64-linux/lib/libcudnn_cnn_train.so.8
/usr/local/cuda-11.6/targets/x86_64-linux/lib/libcudnn_ops_infer.so.8
/usr/local/cuda-11.6/targets/x86_64-linux/lib/libcudnn_ops_train.so.8
HIP runtime version: N/A
MIOpen runtime version: N/A
Is XNNPACK available: True

CPU:
Architecture:                       x86_64
CPU op-mode(s):                     32-bit, 64-bit
Address sizes:                      48 bits physical, 48 bits virtual
Byte Order:                         Little Endian
CPU(s):                             256
On-line CPU(s) list:                0-255
Vendor ID:                          AuthenticAMD
Model name:                         AMD EPYC 7773X 64-Core Processor
CPU family:                         25
Model:                              1
Thread(s) per core:                 2
Core(s) per socket:                 64
Socket(s):                          2
Stepping:                           2
Frequency boost:                    enabled
CPU max MHz:                        3527.7339
CPU min MHz:                        1500.0000
BogoMIPS:                           4400.15
Flags:                              fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt pdpe1gb rdtscp lm constant_tsc rep_good nopl nonstop_tsc cpuid extd_apicid aperfmperf rapl pni pclmulqdq monitor ssse3 fma cx16 pcid sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand lahf_lm cmp_legacy svm extapic cr8_legacy abm sse4a misalignsse 3dnowprefetch osvw ibs skinit wdt tce topoext perfctr_core perfctr_nb bpext perfctr_llc mwaitx cpb cat_l3 cdp_l3 invpcid_single hw_pstate ssbd mba ibrs ibpb stibp vmmcall fsgsbase bmi1 avx2 smep bmi2 erms invpcid cqm rdt_a rdseed adx smap clflushopt clwb sha_ni xsaveopt xsavec xgetbv1 xsaves cqm_llc cqm_occup_llc cqm_mbm_total cqm_mbm_local clzero irperf xsaveerptr rdpru wbnoinvd amd_ppin brs arat npt lbrv svm_lock nrip_save tsc_scale vmcb_clean flushbyasid decodeassists pausefilter pfthreshold v_vmsave_vmload vgif v_spec_ctrl umip pku ospke vaes vpclmulqdq rdpid overflow_recov succor smca fsrm
Virtualization:                     AMD-V
L1d cache:                          4 MiB (128 instances)
L1i cache:                          4 MiB (128 instances)
L2 cache:                           64 MiB (128 instances)
L3 cache:                           1.5 GiB (16 instances)
NUMA node(s):                       2
NUMA node0 CPU(s):                  0-63,128-191
NUMA node1 CPU(s):                  64-127,192-255
Vulnerability Gather data sampling: Not affected
Vulnerability Itlb multihit:        Not affected
Vulnerability L1tf:                 Not affected
Vulnerability Mds:                  Not affected
Vulnerability Meltdown:             Not affected
Vulnerability Mmio stale data:      Not affected
Vulnerability Retbleed:             Not affected
Vulnerability Spec rstack overflow: Mitigation; safe RET
Vulnerability Spec store bypass:    Mitigation; Speculative Store Bypass disabled via prctl
Vulnerability Spectre v1:           Mitigation; usercopy/swapgs barriers and __user pointer sanitization
Vulnerability Spectre v2:           Mitigation; Retpolines, IBPB conditional, IBRS_FW, STIBP always-on, RSB filling, PBRSB-eIBRS Not affected
Vulnerability Srbds:                Not affected
Vulnerability Tsx async abort:      Not affected

Versions of relevant libraries:
[pip3] galore-torch==1.0
[pip3] numpy==2.0.2
[pip3] torch==2.5.1
[pip3] triton==3.1.0
[conda] galore-torch                 1.0              pypi_0           pypi
[conda] numpy                        2.0.2            pypi_0           pypi
[conda] torch                        2.5.1            pypi_0           pypi
[conda] triton                       3.1.0            pypi_0           pypi

[Truong5724]

@sdjasj Hi, I’d like to work on this issue. I would really appreciate it if you could let me know whether there are any additional requirements I should follow. Thank you!