huggingface accelerate issues

huggingface / accelerate

🚀 A simple way to launch, train, and use PyTorch models on almost any device and distributed configuration, automatic mixed precision (including fp8), and easy-to-configure FSDP and DeepSpeed support

https://huggingface.co/docs/accelerate

Apache License 2.0

7.32k stars 872 forks source link

issues

Newest

Newest Most commented Recently updated Oldest Least commented Least recently updated

gather objects in TPU is not supported

#2858 carlesoctav opened 2 weeks ago
4
Fix get_backend bug and add clear_device_cache function

#2857 NurmaU opened 2 weeks ago
1
Drop torch re-imports in npu and mlu paths

#2856 dvrogozh closed 2 weeks ago
1
Refactor logging to use logger in `dispatch_model`

#2855 panjd123 closed 2 weeks ago
1
Auto create dir when merging FSDP weights

#2854 helloworld1 closed 2 weeks ago
1
Default FSDP weights merge to safetensors

#2853 helloworld1 closed 2 weeks ago
1
Better error when a bad directory is given for weight merging

#2852 muellerzr opened 2 weeks ago
1
Remove underlines between badges

#2851 novialriptide closed 2 weeks ago
1
Revert "Slight rename"

#2850 SunMarc closed 2 weeks ago
1
Incorrect output when using accelerate in a pytorch Unet model

#2849 cporrasn opened 2 weeks ago
1
Unable to merge checkpoint use `accelerate merge-weights`

#2848 helloworld1 opened 3 weeks ago
4
device_map="auto" leads to `Expected all tensors to be on the same device` error on generate call

#2847 ryan-caesar-ramos opened 3 weeks ago
1
Training is slower after using generate on unwrapped model

#2846 ylacombe opened 3 weeks ago
1
Speed up imports and add a CI

#2845 muellerzr closed 1 day ago
5
doc: fix link

#2844 imba-tjd closed 3 weeks ago
1
remove warning hook addede during dispatch_model

#2843 SunMarc closed 2 weeks ago
1
fix(ci): remove unnecessary permissions

#2842 McPatate closed 3 weeks ago
2
Add DDP Communication Hooks

#2841 yhna940 closed 2 weeks ago
6
Big Models, move model to CPU after dispatching to multiple devices

#2840 balaabhijit closed 3 weeks ago
3
slurs accelerate Deepspeed launching rises a lot of "WARNING: Skipping ... as it contains forbidden characters or missing values." and the training does not utilize multi node

#2839 ermu2001 closed 3 weeks ago
1
Does DeepSpeed + Accelerate Support Pipeline Parallelism

#2838 sam-h-bean opened 3 weeks ago
1
wait_for_everyone() did not work

#2837 ByungKwanLee closed 2 weeks ago
1
feat(ci): add trufflehog secrets detection

#2836 McPatate closed 3 weeks ago
1
A Question about finetuning models using adapters with FSDP with accelerate

#2835 Abhrant closed 3 weeks ago
3
To reduce Python overhead and maximize the performance of your TPU, pass in the steps_per_execution argument to Keras Model.compile. In this example, it increases throughput by about 50%:

#2834 imrankh46 opened 3 weeks ago
2
monitor-interval, take 2

#2833 muellerzr closed 3 weeks ago
1
Cannot train quantized model with both model and data parallelism

#2832 JubilantJerry opened 3 weeks ago
2
GradientState _add_dataloader never called

#2831 RuiningLi closed 3 weeks ago
4
Improve test speeds by up to 30% in multi-gpu settings

#2830 muellerzr closed 3 weeks ago
6
RuntimeError: Expected all tensors to be on the same device, but found at least two devices, cuda:0 and cpu! I am on a single T4 GPU

#2829 kevalshah90 opened 3 weeks ago
1
optimizer.step_was_skipped not correct in accelerator.accumulate

#2828 Fadelis98 opened 3 weeks ago
1
Accelerate logging in_order=True does not work properly

#2827 zhcm opened 4 weeks ago
7
Remove out-dated xpu device check code in `get_balanced_memory`

#2826 faaany closed 3 weeks ago
4
xpu: support xpu backend from stock pytorch (>=2.4)

#2825 dvrogozh closed 2 weeks ago
6
Added Multicpu SLURM example .sh script in examples/slurm folder

#2824 okhleif-IL closed 4 weeks ago
0
Dataloader yields wrong sequence when resuming training

#2823 lolalebreton opened 4 weeks ago
4
Optimize the megatron plugin

#2822 zhangsheng377 closed 3 weeks ago
7
Accelerate + DeepSpeed

#2821 ByungKwanLee opened 4 weeks ago
2
add cuda dep for a test

#2820 SunMarc closed 4 weeks ago
1
ValueError: Attempting to unscale FP16 gradients.

#2819 NimbusLongfei opened 4 weeks ago
2
[Multi-node] num_processes is configured wrongly by accelerate config

#2818 jubueche opened 4 weeks ago
0
cpu_offload with diffusers save_pretrained occurs the error: NotImplementedError: Cannot copy out of meta tensor; no data!

#2817 zengziru opened 1 month ago
1
[DeepSpeed + Slurm + Accelerate] Rendez vous timeout error, questions about correct setup

#2816 jubueche closed 4 weeks ago
4
[DRAFT] More efficient multi-gpu tests

#2815 muellerzr closed 3 weeks ago
1
Fix DeepSpeed config validation error by changing `stage3_prefetch_bucket_size` value to an integer

#2814 adk9 closed 3 weeks ago
1
"Only Tensors of floating point and complex dtype can require gradients", on FSDP, Accelerate, quatization

#2813 artkpv opened 1 month ago
2
Text generation task otuputs nonsense when using transformers.pipeline with device_map="auto"

#2812 cristi-zz opened 1 month ago
12
accelerate Inappropriate convert all inputs to specific dtype

#2811 rangehow opened 1 month ago
5
fix fstr format

#2810 Jintao-Huang closed 3 weeks ago
3
NotImplementedError: Cannot copy out of meta tensor; no data!

#2809 CHNRyan closed 3 weeks ago
3

Previous Next