msr-fiddle/pipedream

Issues

• same train_loader but got different loader size

  #80 opened 7 months ago by Hyaloid
  2

• Some error about communication

  #43 opened 7 months ago by jglicat
  4

• optimizer got an empty parameter list when rank=1

  #79 opened 7 months ago by Hyaloid
  1

• what is the role of pre_hook_pytorch_latest.patch？

  #77 opened a year ago by matrix97317
  1

• When I was testing the pipedream code with version-updated torch, I encountered the following error (1.1.0 -> 1.11.0):

  #78 opened a year ago by lengien
  2

• Supporting T5

  #68 opened a year ago by gperrotta
  0

• Running in docker will give you an error that you can't find a physical address

  #76 opened a year ago by guanyonglai
  1

• AttributeError: module 'torch.distributed' has no attribute 'P2POp'

  #75 opened a year ago by guanyonglai
  1

• Is there any 2bw code that will run on the native GPU

  #74 opened 2 years ago by guanyonglai
  1

• AttributeError: module 'models.resnet50.resnet50' has no attribute 'model'

  #73 opened 2 years ago by guanyonglai
  1

• Translation demo: Division by zero

  #49 opened 4 years ago by grwlf
  4

• Planner for PipeDream-2BW

  #57 opened 4 years ago by nict-wisdom
  5

• Question about PipeDream's optimizer

  #71 opened 2 years ago by lllukehuang
  0

• Multi-machine distribution problem

  #34 opened 5 years ago by ADAM-CT
  6

• Question about time complexity of PipeDream-2BW's planner algorithm

  #70 opened 3 years ago by ConnollyLeon
  0

• What's the latest version of PyTorch supported?

  #52 opened 4 years ago by SimonZsx
  10

• The arguments of self.start_helper_thread() should be more flexible instead of fixed as int64.

  #69 opened 3 years ago by gouchangjiang
  0

• How is the Double-Buffered Weight Mechanism implemented?

  #66 opened 3 years ago by BinhangYuan
  0

• Is there AllReduce in data parallelism?

  #65 opened 3 years ago by Allen-Czyysx
  6

• All ranks are not trained. They are blocked all the time

  #39 opened 5 years ago by ADAM-CT
  3

• GPT2 355m model convergence with 2BW training

  #64 opened 3 years ago by nitikasaran68
  0

• GPU Peer2Peer communication via --num_ranks_in_server argument

  #60 opened 4 years ago by siddharth9820
  1

• Resource temporarily unavailable

  #61 opened 4 years ago by liulixinkerry
  0

• The BLEU score of translation model seems abnormal. The model doesn't seem to train effectively.

  #63 opened 4 years ago by njuyexiangyu
  0

• To run PipeDream_2BW branch without --recompute_step

  #62 opened 4 years ago by Shigangli
  0

• Handling uneven number of batches per replicated instance of a layer

  #59 opened 4 years ago by siddharth9820
  0

• Running a transformer module

  #58 opened 4 years ago by oranichu
  0

• Multi node training

  #36 opened 5 years ago by ADAM-CT
  6

• Communication error when training with Pipedream

  #56 opened 4 years ago by gudiandian
  6

• Hanging with [4,3,1] GPU assignment

  #55 opened 4 years ago by BestSonny
  0

• Can the profiler handle dynamic graphs?

  #54 opened 4 years ago by rahul003
  0

• Unexpected Error

  #31 opened 5 years ago by kanonjz
  8

• What is the meaning of `antichain` in optimizer_graph_hierarchical.py ?

  #53 opened 4 years ago by grwlf
  0

• How to expose the "register_pre_hook()" interface?

  #51 opened 4 years ago by letian-zhang
  1

• Actual results did not match the optimizer expectation

  #50 opened 4 years ago by nirandaperera
  1

• Translation demo: Installation instrutions issue. Missing CUDA kernels.

  #48 opened 4 years ago by grwlf
  3

• Batch size and optimizer

  #47 opened 4 years ago by nirandaperera
  6

• Infinite loop problem in convert_graph_to_model.py

  #45 opened 4 years ago by jayhpark530
  3

• Error occurred in profiling

  #44 opened 4 years ago by gudiandian
  3

• Is this the version used in SOSP paper?

  #46 opened 4 years ago by nirandaperera
  1

• whats the role of optimizer/inference_optimizer_graph.py

  #40 opened 4 years ago by ADAM-CT
  1

• docker pull error

  #42 opened 4 years ago by cnzhanj
  1

• How to determine replication factors

  #37 opened 5 years ago by ADAM-CT
  4

• "stage_to_depth_map" not found

  #38 opened 5 years ago by ADAM-CT
  2

• I can't get the same result as your model after segmentation

  #41 opened 5 years ago by ADAM-CT
  2

• RuntimeError: [enforce fail at ../third_party/gloo/gloo/transport/tcp/device.cc:127] rp != nullptr. Unable to find address for: dgx-1.ai

  #32 opened 5 years ago by ADAM-CT
  3

• bandwidth parameter

  #35 opened 5 years ago by ADAM-CT
  2

• I would like to know what the role of the following driver.py is？

  #33 opened 5 years ago by ADAM-CT
  2

• Gpu underutilization

  #29 opened 5 years ago by ADAM-CT
  7

• How to use two GPUs in two servers separately to run Pipedream

  #30 opened 5 years ago by kanonjz
  8