reflection
Beam-1.0-501B-A23B
text generationapache-2.012.6K downloadsHugging Face
No server in stock right now has the 1120 GB this model needs for inference.
GPU memory it needs (estimate)
Rule of thumb from the model's size. Real needs depend on batch size, context length and framework.
Inference (full precision)
1120 GB
Inference (8-bit)
560 GB
Inference (4-bit)
280 GB
LoRA fine-tuning
1214 GB
QLoRA (8-bit base)
607 GB
Full fine-tuning
4482 GB