Instructions to use AbacusResearch/haLLAwa3 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use AbacusResearch/haLLAwa3 with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-generation", model="AbacusResearch/haLLAwa3")

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("AbacusResearch/haLLAwa3")
model = AutoModelForCausalLM.from_pretrained("AbacusResearch/haLLAwa3")

Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use AbacusResearch/haLLAwa3 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "AbacusResearch/haLLAwa3"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "AbacusResearch/haLLAwa3",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/AbacusResearch/haLLAwa3

SGLang

How to use AbacusResearch/haLLAwa3 with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "AbacusResearch/haLLAwa3" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "AbacusResearch/haLLAwa3",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "AbacusResearch/haLLAwa3" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "AbacusResearch/haLLAwa3",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use AbacusResearch/haLLAwa3 with Docker Model Runner:
```
docker model run hf.co/AbacusResearch/haLLAwa3
```
Browse Quantizations to use this model in llama.cpp, Ollama, LM Studio, or any compatible app.

Hallawa3: The Fusion of Expertise and Precision for 7B Models"

Unveiling 'Hallawa', an AI marvel that embodies the perfect blend of expert knowledge and cutting-edge technology, tailored for 7B models where direct answers are paramount. This AI powerhouse excels in delivering precise responses, ideal for use cases that demand accuracy and immediacy. Excelling in document understanding and prompts in its size. With 'Hallawa', you tap into a repository of intelligence that's been acknowledged by over 1400 downloads on the OpenLLM leaderboard, boasting a remarkable score of 71. This model isn't just about quantity but quality, setting new benchmarks in the realm of language models.

Whether you're looking to enhance customer service, drive research, or accelerate decision-making, 'Hallawa' stands as your go-to solution, engineered to exceed expectations in scenarios where only the most accurate and immediate answers will suffice. Join the ranks of those leveraging 'Hallawa' for their most critical applications and witness the transformation it brings to your operations. haLLAwa3 is a merge of the following models using mergekit:

🧩 Configuration

slices:
  - sources:
      - model: openchat/openchat-3.5-0106
        layer_range: [0, 32]
      - model: machinists/Mistral-7B-SQL
        layer_range: [0, 32]
merge_method: slerp
base_model: openchat/openchat-3.5-0106
parameters:
  t:
    - filter: self_attn
      value: [0, 0.5, 0.3, 0.7, 1]
    - filter: mlp
      value: [1, 0.5, 0.7, 0.3, 0]
    - value: 0.5
dtype: bfloat16

Open LLM Leaderboard Evaluation Results

Detailed results can be found here

Metric	Value
Avg.	71.34
AI2 Reasoning Challenge (25-Shot)	67.83
HellaSwag (10-Shot)	87.02
MMLU (5-Shot)	64.23
TruthfulQA (0-shot)	63.71
Winogrande (5-shot)	80.51
GSM8k (5-shot)	64.75

Downloads last month: 92

Safetensors

Model size

7B params

Tensor type

BF16

Model tree for AbacusResearch/haLLAwa3

Quantizations

2 models

Evaluation results

normalized accuracy on AI2 Reasoning Challenge (25-Shot)
test set Open LLM Leaderboard

67.830
normalized accuracy on HellaSwag (10-Shot)
validation set Open LLM Leaderboard

87.020
accuracy on MMLU (5-Shot)
test set Open LLM Leaderboard

64.230
mc2 on TruthfulQA (0-shot)
validation set Open LLM Leaderboard

63.710
accuracy on Winogrande (5-shot)
validation set Open LLM Leaderboard

80.510
accuracy on GSM8k (5-shot)
test set Open LLM Leaderboard

64.750