Instructions to use AbacusResearch/Jallabi-34B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use AbacusResearch/Jallabi-34B with Transformers:

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

pipe = pipeline("text-generation", model="AbacusResearch/Jallabi-34B")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("AbacusResearch/Jallabi-34B")
model = AutoModelForCausalLM.from_pretrained("AbacusResearch/Jallabi-34B")
messages = [
    {"role": "user", "content": "Who are you?"},
]
inputs = tokenizer.apply_chat_template(
	messages,
	add_generation_prompt=True,
	tokenize=True,
	return_dict=True,
	return_tensors="pt",
).to(model.device)

outputs = model.generate(**inputs, max_new_tokens=40)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use AbacusResearch/Jallabi-34B with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "AbacusResearch/Jallabi-34B"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "AbacusResearch/Jallabi-34B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/AbacusResearch/Jallabi-34B

SGLang

How to use AbacusResearch/Jallabi-34B 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/Jallabi-34B" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "AbacusResearch/Jallabi-34B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

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/Jallabi-34B" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "AbacusResearch/Jallabi-34B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use AbacusResearch/Jallabi-34B with Docker Model Runner:
```
docker model run hf.co/AbacusResearch/Jallabi-34B
```
Browse Quantizations to use this model in llama.cpp, Ollama, LM Studio, or any compatible app.

These are llama only weights of https://huggingface.co/liuhaotian/llava-v1.6-34b . The Clip encoder part is removed and and this model is llama weights only that can be loaded using LlamaForCausalLM. Which indirectly is a https://huggingface.co/NousResearch/Nous-Hermes-2-Yi-34B licence.

Open LLM Leaderboard Evaluation Results

Detailed results can be found here

Metric	Value
Avg.	70.73
AI2 Reasoning Challenge (25-Shot)	66.04
HellaSwag (10-Shot)	83.81
MMLU (5-Shot)	76.40
TruthfulQA (0-shot)	51.46
Winogrande (5-shot)	81.45
GSM8k (5-shot)	65.20

Open LLM Leaderboard Evaluation Results

Detailed results can be found here

Metric	Value
Avg.	25.97
IFEval (0-Shot)	35.29
BBH (3-Shot)	43.62
MATH Lvl 5 (4-Shot)	3.93
GPQA (0-shot)	11.86
MuSR (0-shot)	20.24
MMLU-PRO (5-shot)	40.91

Downloads last month: 226

Safetensors

Model size

34B params

Tensor type

F32

Model tree for AbacusResearch/Jallabi-34B

Quantizations

1 model

Evaluation results

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

66.040
normalized accuracy on HellaSwag (10-Shot)
validation set Open LLM Leaderboard

83.810
accuracy on MMLU (5-Shot)
test set Open LLM Leaderboard

76.400
mc2 on TruthfulQA (0-shot)
validation set Open LLM Leaderboard

51.460
accuracy on Winogrande (5-shot)
validation set Open LLM Leaderboard

81.450
accuracy on GSM8k (5-shot)
test set Open LLM Leaderboard

65.200
strict accuracy on IFEval (0-Shot)
Open LLM Leaderboard

35.290
normalized accuracy on BBH (3-Shot)
Open LLM Leaderboard

43.620
exact match on MATH Lvl 5 (4-Shot)
Open LLM Leaderboard

3.930
acc_norm on GPQA (0-shot)
Open LLM Leaderboard

11.860
acc_norm on MuSR (0-shot)
Open LLM Leaderboard

20.240
accuracy on MMLU-PRO (5-shot)
test set Open LLM Leaderboard

40.910