Jade72b is a Brazilian Portuguese conversational finetune of Qwen2.5 72B built to express a strong, persistent persona. This model is designed for PT-BR chat, chatbot use cases, and character-style interaction, with colloquial language, abbreviations, slang, and a WhatsApp-like tone.
Jade72b is a persona-first model. It was intentionally finetuned so the model speaks like Jade even without a strong system prompt. Because of that, the model often answers in PT-BR with informal phrasing such as vc, slang, and a friendly conversational tone from the very first turn.
Madras1qwen/qwen2.5-72bpt-BR)apache-2.0This model was trained to:
Typical behavior includes:
vctmj, mano, tlgdIf Jade already sounds like a recurring character during inference, that is expected behavior, not an error.
The finetune objective was to make the persona live in the weights, not only in prompting.
High-level training approach:
system persona instructions during SFT so the model directly internalizes the Jade styleThis is why the model can already answer with personality, abbreviations, and slang even with a simple user-only prompt.
High-level setup used for this finetune:
26,000 examples4 epochsBest fit:
Less ideal for:
For the strongest Jade behavior:
Example prompts:
oi jade, tudo bem?jade, me explica isso de um jeito simplesvc acha que vale a pena estudar python hoje?from transformers import AutoModelForCausalLM, AutoTokenizer
import torch
model_id = "Madras1/Jade72b"
tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(
model_id,
torch_dtype=torch.bfloat16,
device_map="auto",
)
messages = [
{"role": "user", "content": "oi jade, tudo bem?"}
]
text = tokenizer.apply_chat_template(
messages,
tokenize=False,
add_generation_prompt=True,
)
inputs = tokenizer(text, return_tensors="pt").to(model.device)
outputs = model.generate(
**inputs,
max_new_tokens=256,
temperature=0.7,
top_p=0.9,
)
print(tokenizer.decode(outputs[0], skip_special_tokens=True))
Because this is a persona-oriented finetune:
model-00001-of-00021.safetensors 2.5 GBmodel-00020-of-00021.safetensors 2.5 GBmodel-00017-of-00021.safetensors 2.0 GBmodel-00005-of-00021.safetensors 2.0 GBmodel-00008-of-00021.safetensors 2.0 GBmodel-00014-of-00021.safetensors 2.0 GBmodel-00011-of-00021.safetensors 2.0 GBmodel-00018-of-00021.safetensors 2.0 GBmodel-00009-of-00021.safetensors 2.0 GBmodel-00006-of-00021.safetensors 2.0 GBmodel-00012-of-00021.safetensors 2.0 GBmodel-00015-of-00021.safetensors 2.0 GBmodel-00010-of-00021.safetensors 1.9 GBmodel-00007-of-00021.safetensors 1.9 GBmodel-00013-of-00021.safetensors 1.9 GBmodel-00019-of-00021.safetensors 1.9 GBmodel-00016-of-00021.safetensors 1.9 GBmodel-00004-of-00021.safetensors 1.9 GBmodel-00003-of-00021.safetensors 1.9 GBmodel-00002-of-00021.safetensors 1.9 GBmodel-00021-of-00021.safetensors 1.0 GBadapter_model.safetensors 870 MBHave the files? Submit a torrent.
Submit a torrentHave the complete files and permission to share them? Replace the example folder path. The script checks the files, makes a torrent and submits its magnet and checksums. Keep your torrent client seeding. Get your community key from your settings, and read the script before running it.
curl -fsSL https://pirateface.co/package.sh | bash -s -- --repo Madras1/Jade72b ./model-folderQwen/Qwen2.5-72B