Instructions to use daniel-eai/Llama-3.2-2B-40pct-Compressed-3B-EN-V1 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use daniel-eai/Llama-3.2-2B-40pct-Compressed-3B-EN-V1 with Transformers:

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

pipe = pipeline("text-generation", model="daniel-eai/Llama-3.2-2B-40pct-Compressed-3B-EN-V1", trust_remote_code=True)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("daniel-eai/Llama-3.2-2B-40pct-Compressed-3B-EN-V1", trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained("daniel-eai/Llama-3.2-2B-40pct-Compressed-3B-EN-V1", trust_remote_code=True)

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use daniel-eai/Llama-3.2-2B-40pct-Compressed-3B-EN-V1 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "daniel-eai/Llama-3.2-2B-40pct-Compressed-3B-EN-V1"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "daniel-eai/Llama-3.2-2B-40pct-Compressed-3B-EN-V1",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/daniel-eai/Llama-3.2-2B-40pct-Compressed-3B-EN-V1

SGLang

How to use daniel-eai/Llama-3.2-2B-40pct-Compressed-3B-EN-V1 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 "daniel-eai/Llama-3.2-2B-40pct-Compressed-3B-EN-V1" \
    --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": "daniel-eai/Llama-3.2-2B-40pct-Compressed-3B-EN-V1",
		"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 "daniel-eai/Llama-3.2-2B-40pct-Compressed-3B-EN-V1" \
        --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": "daniel-eai/Llama-3.2-2B-40pct-Compressed-3B-EN-V1",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use daniel-eai/Llama-3.2-2B-40pct-Compressed-3B-EN-V1 with Docker Model Runner:
```
docker model run hf.co/daniel-eai/Llama-3.2-2B-40pct-Compressed-3B-EN-V1
```

Llama-3.2-2B — 40% Compressed from Llama-3.2-3B (English)

Built with Llama. This repository is part of the Efficient and Robust AI System (E-AI) Project by Vincent-Daniel Yun. This model is a compressed edition of meta-llama/Llama-3.2-3B with 11 of 28 transformer layers removed (17 layers remain, ≈2.11B parameters), recovered training-free. It is a base (non–chat-tuned) model.

🔗 Project: https://www.worldwidedaniel.com/eai-project 📅 Release: 2026-06-28 · Version: V1

⚖️ License: governed by the Llama 3.2 Community License Agreement (https://huggingface.co/meta-llama/Llama-3.2-3B/blob/main/LICENSE.txt), and subject to the Llama 3.2 Acceptable Use Policy (https://www.llama.com/llama3_2/use-policy). By using this model you agree to those terms. "Llama" is a trademark of Meta Platforms, Inc. Built with Llama.

⚠️ Language: English-focused. Use as a discrimination / classification engine — open-ended long-form generation is degraded by compression (see PPL). For factual questions use retrieval (RAG).

About E-AI

Modern AI is powerful but heavy. The E-AI (Efficient-AI) project builds compact yet capable AI — making every model lightweight and fast — so AI can assist people in urgent, high-stakes moments.

Method

The pruning method and the training-free recovery method are proprietary, undisclosed methods created by Vincent-Daniel Yun and are not released. Only the resulting model is shared.

Results (measured, full test sets)

PPL on 2048-token context (↓ better); downstream & MMLU are 0-shot via lm-eval-harness (↑ better).

Metric	Llama-3.2-3B (base)	This model (40%)
PPL · WikiText2 ↓	7.81	209.11
PPL · C4 ↓	10.26	156.25
ARC-c ↑	0.4599	0.3208
ARC-e ↑	0.7163	0.4508
BoolQ ↑	0.7324	0.6229
COPA ↑	0.8600	0.7200
HellaSwag ↑	0.7363	0.4494
OpenBookQA ↑	0.4300	0.3100
RACE ↑	0.4010	0.3158
RTE ↑	0.5487	0.6354
WinoGrande ↑	0.6985	0.6219
Avg. downstream (9) ↑	0.6203	0.4941
MMLU ↑	0.5437	0.5120

Task-suitability (vs base Llama-3.2-3B)

Task	Llama-3.2-3B (base)	This model (40%)
Topic classification (AG News)	0.550	0.418
LLM-as-judge (RewardBench)	0.612	0.610
SafetyBench (MCQ)	0.743	0.629
MultiRC	0.572	0.572
WiC	0.497	0.500
MRPC	0.625	0.684
CB (NLI)	0.500	0.196
SST-2	0.751	0.611
MedQA	0.511	0.503
MedMCQA	0.491	0.464
PubMedQA	0.732	0.556
Belebele-en	0.654	0.590
Belebele-ko	0.527	0.446
XNLI-zh	0.411	0.347
ToxiGen	0.432	0.432
TruthfulQA	0.392	0.495

(Discrimination is largely preserved; generation/PPL degrades with compression. Yes/No safety-F1 benchmarks are omitted as the non-chat base is poorly calibrated for them.)

Efficiency (measured, fp16, single GPU)

	Llama-3.2-3B (base)	This model (40%)
Layers	28	17
Parameters	3.21B	2.11B
Peak inference memory (fp16)	9.53 GB	6.94 GB (−27%)
Forward latency (fp16)	571 ms	378 ms (−34%)

Usage — Transformers

import torch
from transformers import AutoModelForCausalLM, AutoTokenizer
m = AutoModelForCausalLM.from_pretrained("daniel-eai/Llama-3.2-2B-40pct-Compressed-3B-EN-V1", trust_remote_code=True, dtype=torch.float16, device_map="cuda")
tok = AutoTokenizer.from_pretrained("daniel-eai/Llama-3.2-2B-40pct-Compressed-3B-EN-V1", trust_remote_code=True)
ids = tok("The capital of France is", return_tensors="pt").to("cuda")
print(tok.decode(m.generate(**ids, max_new_tokens=20)[0]))

trust_remote_code=True is required (custom decoder layer in modeling_llama_recovered.py).

License & Acknowledgements

Derivative of Llama 3.2, distributed under the Llama 3.2 Community License. Built with Llama. Thanks to Prof. Sai Praneeth Karimireddy (USC) and Prof. Sunwoo Lee (Inha University) for guidance, and to Meta for releasing Llama 3.2.

Downloads last month: -

Safetensors

Model size

2B params

Tensor type

F16

Model tree for daniel-eai/Llama-3.2-2B-40pct-Compressed-3B-EN-V1

Base model

meta-llama/Llama-3.2-3B

Finetuned

(474)

this model