Llama-2-7b-evol-code-alpaca-pruned_70

Llama-2-7b-pruned70-retrained-evolcodealpaca

This repo contains a 70% sparse Llama 2 7B finetuned for code generation tasks using the Evolved CodeAlpaca dataset.

Official model weights from Enabling High-Sparsity Foundational Llama Models with Efficient Pretraining and Deployment.

Authors: Neural Magic, Cerebras

Usage

Below we share some code snippets on how to get quickly started with running the model.

Sparse Transfer

By leveraging a pre-sparsified model's structure, you can efficiently fine-tune on new data, leading to reduced hyperparameter tuning, training times, and computational costs. Learn about this process here.

Running the model

This model may be run with the transformers library. For accelerated inference with sparsity, deploy with nm-vllm or deepsparse.

# pip install transformers accelerate
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("neuralmagic/Llama-2-7b-pruned70-retrained-evolcodealpaca")
model = AutoModelForCausalLM.from_pretrained("neuralmagic/Llama-2-7b-pruned70-retrained-evolcodealpaca", device_map="auto")

input_text = "def fibonacci(n):\n"
input_ids = tokenizer(input_text, return_tensors="pt").to("cuda")

outputs = model.generate(**input_ids)
print(tokenizer.decode(outputs[0]))

Evaluation Benchmark Results

Model evaluation metrics and results.

Model Training Details

This model was obtained by gradual sparse-transfer of the sparse foundational model Llama-2-7b-pruned50-retrained on 60% of the evolcodealpaca dataset. The 50% sparse foundational model was finetuned for 2 epochs and then pruned to 70% sparsity using SparseGPT. Then, the model was finetuned for 1 more epoch with the SquareHead knowledge distillation with Llama-2-7b-evolcodealpaca as teacher.

Help

For further support, and discussions on these models and AI in general, join Neural Magic's Slack Community