Instructions to use ai-eldorado/LlamaEld-3.1-8B-Instruct with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use ai-eldorado/LlamaEld-3.1-8B-Instruct with Transformers:

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

pipe = pipeline("text-generation", model="ai-eldorado/LlamaEld-3.1-8B-Instruct")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForMultimodalLM

tokenizer = AutoTokenizer.from_pretrained("ai-eldorado/LlamaEld-3.1-8B-Instruct")
model = AutoModelForMultimodalLM.from_pretrained("ai-eldorado/LlamaEld-3.1-8B-Instruct")
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 Settings

vLLM

How to use ai-eldorado/LlamaEld-3.1-8B-Instruct with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "ai-eldorado/LlamaEld-3.1-8B-Instruct"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "ai-eldorado/LlamaEld-3.1-8B-Instruct",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/ai-eldorado/LlamaEld-3.1-8B-Instruct

SGLang

How to use ai-eldorado/LlamaEld-3.1-8B-Instruct 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 "ai-eldorado/LlamaEld-3.1-8B-Instruct" \
    --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": "ai-eldorado/LlamaEld-3.1-8B-Instruct",
		"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 "ai-eldorado/LlamaEld-3.1-8B-Instruct" \
        --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": "ai-eldorado/LlamaEld-3.1-8B-Instruct",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use ai-eldorado/LlamaEld-3.1-8B-Instruct with Docker Model Runner:
```
docker model run hf.co/ai-eldorado/LlamaEld-3.1-8B-Instruct
```

LlamaEld-3.1-8B-Instruct

Model Overview

LlamaEld-3.1-8B-Instruct is a Portuguese-focused instruction-tuned large language model derived from Llama-3.1-8B-Instruct. It has been further fine-tuned using curated Portuguese-language text extracted and filtered from the ClueWeb22 dataset.

This model aims to improve performance on Portuguese understanding, instruction following, and generation tasks, particularly in real-world web text domains.

Model Details

Base Model: meta-llama/Llama-3.1-8B-Instruct
Architecture: Transformer (decoder-only)
Parameters: ~8 billion
Fine-tuning Type: Supervised fine-tuning (SFT)
Primary Language: Portuguese
License: Same as base model (Llama 3.1 license—verify compliance before use)

Training Data

The model was fine-tuned using:

Source: ClueWeb22
Subset: Portuguese-language documents
Processing:
- Language filtering (Portuguese only)
- Quality filtering and cleaning
- Removal of noisy or low-quality web content

Data Characteristics

ClueWeb22 is a large-scale web crawl dataset, meaning:

It contains diverse and heterogeneous content
It may include biases, inaccuracies, and outdated information
Despite filtering, some noise may persist

Intended Use

Suitable for:

Portuguese text generation and completion
Instruction following in Portuguese
Chatbots and conversational AI
Summarization and rewriting tasks
Question answering (general domain)

Not Recommended for:

High-stakes decision-making (legal, medical, financial)
Tasks requiring guaranteed factual accuracy
Sensitive or regulated applications without additional safeguards

Usage

Example (Transformers)

from transformers import AutoTokenizer, AutoModelForCausalLM
import torch

model_id = "ai-eldorado/LlamaEld-3.1-8B-Instruct"

tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(
    model_id,
    torch_dtype=torch.float16,
    device_map="auto"
)

prompt = "Explique o impacto da inteligência artificial na educação."
inputs = tokenizer(prompt, return_tensors="pt")

outputs = model.generate(
    **inputs,
    max_new_tokens=256,
    temperature=0.7,
    top_p=0.9
)

print(tokenizer.decode(outputs[0], skip_special_tokens=True))

Evaluation

No standardized benchmark results are currently provided.

However, expected improvements include:

Better fluency in Portuguese
Stronger adherence to instructions in Portuguese
Improved handling of web-style content and vocabulary

For production use, we recommend performing task-specific evaluation.

Limitations

May generate incorrect or hallucinated information
Sensitive to prompt phrasing
Bias inherited from web data (ClueWeb22)
Performance in non-Portuguese languages may be degraded
Not optimized for reasoning-heavy tasks compared to larger models

Safety and Bias

Because the model is trained on web data:

It may reproduce harmful stereotypes or biases
It may generate inappropriate or misleading content

Recommended Mitigations:

Add moderation layers
Use prompt engineering for safer outputs
Apply human-in-the-loop validation in critical systems

Downloads last month: 17

Safetensors

Model size

8B params

Tensor type

BF16

Model tree for ai-eldorado/LlamaEld-3.1-8B-Instruct

Base model

meta-llama/Llama-3.1-8B

Finetuned

meta-llama/Llama-3.1-8B-Instruct

Finetuned

(2787)

this model