Comments (1)
Additional context
deepspeed==0.14.2
CSAN detected a possible data race on tensor with data pointer 140108946735104
Access by stream 0 during kernel:
aten::slice.Tensor(Tensor(a) self, int dim=0, SymInt? start=None, SymInt? end=None, SymInt step=1) -> Tensor(a)
writing to argument(s) self, and to the output
With stack trace:
File "/usr/local/lib/python3.9/dist-packages/deepspeed/runtime/zero/stage_1_and_2.py", line 903, in reduce_partition_and_remove_grads
self.reduce_ready_partitions_and_remove_grads(param, i)
File "/usr/local/lib/python3.9/dist-packages/deepspeed/runtime/zero/stage_1_and_2.py", line 1416, in reduce_ready_partitions_and_remove_grads
self.reduce_independent_p_g_buckets_and_remove_grads(param, i)
File "/usr/local/lib/python3.9/dist-packages/deepspeed/runtime/zero/stage_1_and_2.py", line 949, in reduce_independent_p_g_buckets_and_remove_grads
new_grad_tensor = self.ipg_buffer[self.ipg_index].narrow(0, self.elements_in_ipg_bucket, param.numel())
File "/usr/local/lib/python3.9/dist-packages/torch/cuda/_sanitizer.py", line 570, in __torch_dispatch__
errors = self.event_handler._handle_kernel_launch(
File "/usr/local/lib/python3.9/dist-packages/torch/cuda/_sanitizer.py", line 371, in _handle_kernel_launch
stack_trace = traceback.StackSummary.extract(
Previous access by stream 152815408 during kernel:
aten::view(Tensor(a) self, SymInt[] size) -> Tensor(a)
writing to argument(s) self, and to the output
With stack trace:
File "/usr/local/lib/python3.9/dist-packages/deepspeed/runtime/zero/stage_1_and_2.py", line 903, in reduce_partition_and_remove_grads
self.reduce_ready_partitions_and_remove_grads(param, i)
File "/usr/local/lib/python3.9/dist-packages/deepspeed/runtime/zero/stage_1_and_2.py", line 1416, in reduce_ready_partitions_and_remove_grads
self.reduce_independent_p_g_buckets_and_remove_grads(param, i)
File "/usr/local/lib/python3.9/dist-packages/deepspeed/runtime/zero/stage_1_and_2.py", line 932, in reduce_independent_p_g_buckets_and_remove_grads
self.reduce_ipg_grads()
File "/usr/local/lib/python3.9/dist-packages/deepspeed/runtime/zero/stage_1_and_2.py", line 1367, in reduce_ipg_grads
self.average_tensor(self.ipg_buffer[self.ipg_index].narrow(0, 0, self.elements_in_ipg_bucket))
File "/usr/local/lib/python3.9/dist-packages/deepspeed/runtime/zero/stage_1_and_2.py", line 1127, in average_tensor
self.allreduce_and_scatter(buckets[bucket_key],
File "/usr/local/lib/python3.9/dist-packages/deepspeed/runtime/zero/stage_1_and_2.py", line 1031, in allreduce_and_scatter
self.allreduce_and_copy_with_multiple_ranks(small_bucket,
File "/usr/local/lib/python3.9/dist-packages/deepspeed/runtime/zero/stage_1_and_2.py", line 1005, in allreduce_and_copy_with_multiple_ranks
for buf, synced, bucket_rank in zip(small_bucket, self.unflatten(allreduced, small_bucket), bucket_ranks):
File "/usr/local/lib/python3.9/dist-packages/torch/_utils.py", line 534, in _unflatten_dense_tensors
return torch._C._nn.unflatten_dense_tensors(flat, tensors)
File "/usr/local/lib/python3.9/dist-packages/torch/cuda/_sanitizer.py", line 570, in __torch_dispatch__
errors = self.event_handler._handle_kernel_launch(
File "/usr/local/lib/python3.9/dist-packages/torch/cuda/_sanitizer.py", line 371, in _handle_kernel_launch
stack_trace = traceback.StackSummary.extract(
from deepspeed.
Related Issues (20)
- different setting for same (num_gpus * batch_size * grad_accum_steps) output different loss and gradient norm HOT 1
- [BUG] Stage 3 in WSL2 throws RuntimeError: Expected all tensors to be on the same device, but found at least two devices, cuda:0 and cpu! HOT 3
- [BUG] DeepSpeed is loads the whole model to every GPUs instead of partitioning HOT 1
- [BUG] RepeatingLoader may be invalid in the pipe stages neither the fist nor last
- [REQUEST] Supporting custom generation loop (outlines, LMQL, guidance) in DeepSpeedHybridEngine
- [BUG] M1 Mac has an issue with `hostname -I` not being a valid command HOT 2
- [BUG] CUDA OOM error when Hugging Face `ignore_mismatched_sizes` is enabled
- [BUG] Zero3 causes AttributeError: 'NoneType' object has no attribute 'numel' in continual training HOT 3
- [BUG] cannot import name '_get_socket_with_port' from 'torch.distributed.elastic.agent.server.api' HOT 3
- # [REQUEST] Upstream modifications of PaRO
- Reset Optimizer HOT 1
- nv-ds-chat CI test failure
- [HELP] ZeRO3 partition parameters after fully load to each GPU! HOT 3
- [BUG] ZeRO optimizer with MoE Expert Parallelism HOT 1
- [BUG] Pipeline Dataloader Samler: `shuffle=False`
- [REQUEST] Moving a trainable model with an optimiser between GPU and CPU
- [BUG] RuntimeError: Error building extension 'fused_adam' Loading extension module fused_adam
- [BUG] 1: error: must run as root and 2: raise RuntimeError("Ninja is required to load C++ extensions")
- [BUG] RuntimeError encountered when generating tokens from a DeepSpeedHybridEngine initialized with 4-bit quantization. HOT 2
- [BUG] is_zero_init_model is always False when I'm using zero_init! HOT 4
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from deepspeed.