Part of the AWS ML Certification

Fine-tuned(Transfer Learning) a large language model (LLM) on SageMaker. Utilized LoRa for efficient parameter reduction and transfer learning for fast inference with CloudFront. Leveraged a pre-trained model (databricks/dolly-v2-3b) from Hugging Face for optimal performance and resource conservation.

Requirements

The following table lists the required libraries and their versions that you need to install to run this project:

Package	Version
accelerate	>=0.20.3, <1
bitsandbytes	==0.39.0
click	>=8.0.4, <9
datasets	>=2.10.0, <3
deepspeed	>=0.8.3, <0.9
faiss-cpu	==1.7.4
ipykernel	==6.22.0
langchain	==0.0.161
torch	>=1.13.1, <2
transformers	==4.28.1
peft	==0.3.0
pytest	==7.3.2
numpy	>=1.25.2

You can install these packages using pip with the following command:

pip install accelerate>=0.20.3,<1 \
             bitsandbytes==0.39.0 \
             click>=8.0.4,<9 \
             datasets>=2.10.0,<3 \
             deepspeed>=0.8.3,<0.9 \
             faiss-cpu==1.7.4 \
             ipykernel==6.22.0 \
             langchain==0.0.161 \
             torch>=1.13.1,<2 \
             transformers==4.28.1 \
             peft==0.3.0 \
             pytest==7.3.2 \
             numpy>=1.25.2