|
About the 🤗Accelerate category
|
|
1
|
2387
|
February 20, 2022
|
|
Troubles with features in .prepare()
|
|
1
|
19
|
November 30, 2024
|
|
How to run inference on multigpus
|
|
0
|
9
|
November 29, 2024
|
|
Errors when using gradient accumulation with FSDP + PEFT LoRA + SFTTrainer
|
|
1
|
223
|
September 6, 2024
|
|
General question about large model loading
|
|
2
|
821
|
November 28, 2024
|
|
Slurm Issues running accelerate
|
|
1
|
519
|
November 28, 2024
|
|
Proposal to Enhance `get_state_dict` and Introduce `load_from_state_dict` for Greater Flexibility
|
|
0
|
17
|
November 23, 2024
|
|
Request for Clarification and Possible Refinement of `Plugin` and `KwargsHandler` Design
|
|
1
|
40
|
November 23, 2024
|
|
Proposal to Rename `notebook_launcher` for Broader Accessibility and Clarity
|
|
1
|
46
|
November 23, 2024
|
|
How to correctly use model weights outside of forward in distributed training set-up with Accelerate?
|
|
0
|
66
|
November 12, 2024
|
|
Saving bf16 Model Weights When Using Accelerate+DeepSpeed
|
|
1
|
117
|
November 11, 2024
|
|
Inconsistent Training Time with Accelerate
|
|
0
|
14
|
November 8, 2024
|
|
Bug with model.generate if max_length or max_new_tokens are set, with accelerate deepspeed zero level 3
|
|
4
|
1051
|
November 7, 2024
|
|
Issue with LoRA Adapter Loading on Multiple GPUs during Fine-Tuning with Accelerate and SFTTrainer
|
|
3
|
266
|
September 18, 2024
|
|
What is the correct way to compute metrics while training using Accelerate?
|
|
0
|
13
|
October 29, 2024
|
|
Evaluation Metrics are not matching with Shuffle = False
|
|
0
|
15
|
October 19, 2024
|
|
The used dataset had no length, returning gathered tensors. You should drop the remainder yourself
|
|
3
|
47
|
October 19, 2024
|
|
How to specify FSDP config without launching via Accelerate
|
|
3
|
30
|
October 18, 2024
|
|
Loading a HF Model in Multiple GPUs and Run Inferences in those GPUs
|
|
10
|
8144
|
October 16, 2024
|
|
Distributed inference: how to store results in a global variable
|
|
2
|
19
|
October 16, 2024
|
|
Cuda Out of Memory with Multi-GPU Accelerate for gemma-2b
|
|
0
|
56
|
October 13, 2024
|
|
Asymmetric Loss Function has no effect in Accelerate
|
|
0
|
12
|
October 13, 2024
|
|
Restoring the state of the DataLoader using skip_first_batches() after first epoch
|
|
0
|
15
|
October 11, 2024
|
|
HuggingFacePipeline Llama2 load_in_4bit from_model_id the model has been loaded with `accelerate` and therefore cannot be moved to a specific device
|
|
2
|
6639
|
October 9, 2024
|
|
Which (and how) Multi GPU strategy to use to train model with longer max_length (Phi-2 fits in Single GPU but qLoRa gives OOM with 512)?
|
|
3
|
1099
|
September 20, 2024
|
|
Why does Transformer (LLaMa 3.1-8B) give different logits during inference for the same sample when used with single versus multi gpu prediction?
|
|
0
|
44
|
September 20, 2024
|
|
Accelerate doesn't seem to use my GPU?
|
|
7
|
3735
|
September 18, 2024
|
|
Accelerator load_state for LM head with tied weights
|
|
0
|
26
|
September 16, 2024
|
|
Accelerate Distributed Randomly Hangs
|
|
0
|
20
|
September 11, 2024
|
|
FSDP Auto Wrap does not work using `accelerate` in Multi-GPU Setup
|
|
0
|
71
|
September 6, 2024
|