Skip to content

Issues: huggingface/accelerate

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

105 Open 1,522 Closed
105 Open 1,522 Closed
Author
Filter by author
Loading
Label
Filter by label
Loading
Use alt + click/return to exclude labels
or ⇧ + click/return for logical OR
Projects
Filter by project
Loading
Milestones
Filter by milestone
Loading
Assignee
Filter by who’s assigned
Sort
Sort by
Newest Oldest Most commented Least commented Recently updated Least recently updated Best match
Most reactions
👍 👎 😄 🎉 😕 ❤️ 🚀 👀

Issues list

Error while fine tuning with peft, lora, accelerate, SFTConfig and SFTTrainer
#3230 opened Nov 8, 2024 by Isdriai
2 of 4 tasks
2
torch.cuda.is_available() false when running multi-gpu inference with accelerate launch
#3225 opened Nov 6, 2024 by paulgekeler
2 of 4 tasks
"mat2 must be a matrix" error when finetuning Dreambooth flux with FSDP
#3224 opened Nov 5, 2024 by weixiong-ur
2 of 4 tasks
Add case-insensitive parsing of bool environment variables
#3222 opened Nov 5, 2024 by wizeng23
Incorrect type in output of utils.pad_across_processes when input is torch.bool
#3218 opened Nov 4, 2024 by mariusarvinte
2 of 4 tasks
1
1
PyPI published Accelerate==1.1.0 is missing Source Distributions
#3216 opened Nov 4, 2024 by helloworld1
4 tasks
2
ConnectionError: Tried to launch distributed communication on port 29401, but another process is utilizing it. Please specify a different port (such as using the --main_process_port flag or specifying a different main_process_port in your config file) and rerun your script. To automatically use the next open port (on a single node), you can set this to 0.
#3214 opened Nov 4, 2024 by qinchangchang
1 of 4 tasks
How could I convert ZeRO-0 deepspeed weights into fp32 model checkpoint?
#3210 opened Nov 1, 2024 by liming-ai
The optimizer is not receiving the FSDP model parameters.
#3209 opened Nov 1, 2024 by eljandoubi
2 of 4 tasks
1
2
Multiple node inference
#3208 opened Nov 1, 2024 by DLCM-wrz
Multinode, multigpu example fails
#3206 opened Oct 31, 2024 by ffrancesco94
2 of 4 tasks
1
9
Command line arguments related to deepspeed for accelerate launch do not override those of default_config.yaml
#3203 opened Oct 29, 2024 by JdbermeoUZH
2 of 4 tasks
Problem with metrics calculation and dataloader
#3202 opened Oct 28, 2024 by gssriram
2 of 4 tasks
Cuda OOM when accelerator.prepare
#3200 opened Oct 25, 2024 by antoinedelplace
2 of 4 tasks
3
using deepspeed original json config, when using bf16, get the error RuntimeError: expected mat1 and mat2 to have the same dtype, but got: c10::BFloat16 != c10::Half.
#3197 opened Oct 25, 2024 by PMPBinZhang
2 of 4 tasks
Possible issue in Accelerate FSDP Documentation
#3195 opened Oct 24, 2024 by Quicksilver466
1
Unable to access model gradients with DeepSpeed and Accelerate
#3184 opened Oct 22, 2024 by shouyezhe
2 of 4 tasks
3
accelerator.prepare() get OOM,but available in single GPU
#3182 opened Oct 21, 2024 by lqf0624
2 of 4 tasks
1
[Bug] The clip_grad_norm of xla fsdp is not right
#3180 opened Oct 21, 2024 by hanwen-sun
4 tasks
Why don't ML engineers use shampoo ?🧴
#3178 opened Oct 18, 2024 by G-structure
2
Split_batches argument in Accelerator.__init__ is available, but not used
#3177 opened Oct 17, 2024 by yaraksen
MPI on CPU-only: "no support for _allgather_base"
#3176 opened Oct 17, 2024 by tikhu
2 of 4 tasks
Support --standalone for concurrent single node multi-GPU jobs
#3175 opened Oct 16, 2024 by Olive-Z
Can I load model once or dataset once and copy to subprocess?
#3172 opened Oct 16, 2024 by Hans-digit
1
[Bug] The Transformer Engine plugin seems to be incompatible with LayerNorm that has no weights.
#3171 opened Oct 16, 2024 by IDKiro
2 of 4 tasks
ProTip! Exclude everything labeled bug with -label:bug.