BOINC AI Optimum Habana
Last updated
Last updated
π€ Optimum Habana is the interface between the π€ Transformers and π€ Diffusers libraries and . It provides a set of tools that enable easy model loading, training and inference on single- and multi-HPU settings for various downstream tasks as shown in the table below.
HPUs offer fast model training and inference as well as a great price-performance ratio. Check out and for concrete examples. If you are not familiar with HPUs, we recommend you take a look at .
The following model architectures, tasks and device distributions have been validated for π€ Optimum Habana:
In the tables below, β means single-card, multi-card and DeepSpeed have all been validated.
Transformers
BERT
β
β
RoBERTa
β
β
ALBERT
β
β
DistilBERT
β
β
GPT2
β
β
BLOOM(Z)
β
DeepSpeed
StarCoder
β
Single card
GPT-J
DeepSpeed
Single card
DeepSpeed
GPT-NeoX
DeepSpeed
DeepSpeed
OPT
β
DeepSpeed
Llama 2 / CodeLlama
DeepSpeed
LoRA
DeepSpeed
LoRA
StableLM
β
Single card
Falcon
β
Single card
CodeGen
β
Single card
MPT
β
Single card
T5
β
β
BART
β
Single card
ViT
β
β
Swin
β
β
Wav2Vec2
β
β
CLIP
β
β
BridgeTower
β
β
ESMFold
β
Single card
Diffusers
Stable Diffusion
β
Single card
LDM3D
β
Single card
Other models and tasks supported by the π Transformers and π Diffusers library may also work. You can refer to this for using them with π Optimum Habana. Besides, explains how to modify any from the π Transformers library to make it work with π Optimum Habana.