FritzStack/QWEmotioN-4bit

Usage

4-bit-quantized Qwen 0.6B fine-tuned on the english version of `brighter-dataset/BRIGHTER-emotion-categories'.

To use the model:

from transformers import AutoModelForCausalLM, AutoTokenizer
import torch


tokenizer = AutoTokenizer.from_pretrained(
    "FritzStack/QWEmotioN-4bit", 
    trust_remote_code=True
)

model = AutoModelForCausalLM.from_pretrained(
    "FritzStack/QWEmotioN-4bit",
    torch_dtype=torch.float16,
    device_map="auto",
    trust_remote_code=True
)

def predict_emotions(text, max_new_tokens=50):
    """
    Predict emotions for a given text
    """
    prompt = f"{text}. "
    inputs = tokenizer(prompt, return_tensors="pt").to(model.device)

    with torch.no_grad():
        outputs = model.generate(
            **inputs,
            max_new_tokens=max_new_tokens,
            do_sample=False,
            #temperature=0.8,
            top_k = 10,
            #repetition_penalty=35.,
            pad_token_id=tokenizer.eos_token_id
        )

    generated_text = tokenizer.decode(
        outputs[0][len(inputs.input_ids[0]):],
        skip_special_tokens=False
    ).strip()

    return generated_text

print(predict_emotions("I miss you"))
### Output
Emotion Output: sadness <|im_end|>

Uploaded model

• Developed by: FritzStack
• License: apache-2.0
• Finetuned from model : unsloth/qwen3-0.6b-bnb-4bit

This qwen3 model was trained 2x faster with Unsloth and Huggingface's TRL library.