Going through the doc, accelerate currently does not seem to support loading weights straight to the GPU if the model is to be used for training. If this is correct, are there any plans to support this in the future?
Related topics
Topic | Replies | Views | Activity | |
---|---|---|---|---|
Does one need to load the model to GPU before calling train when using accelerate? | 2 | 1030 | October 31, 2023 | |
Accelerate! I have a query, no actual problem to be solved! | 2 | 284 | August 8, 2023 | |
[Nov 16th Event] Sylvain Gugger: Supercharge your PyTorch training loop with 🤗 Accelerate | 5 | 398 | November 16, 2021 | |
Does accelerate.prepare() destroy model weights even if --model_name_or_path is specified and model is loaded? | 1 | 726 | June 23, 2023 | |
Loading a HF Model in Multiple GPUs and Run Inferences in those GPUs | 10 | 9723 | October 16, 2024 |