lfm25-encoder-goemotions-smoke

LiquidAI/LFM2.5-Encoder-350M fine-tuned for multi-label text classification on google-research-datasets/go_emotions.

• Labels (28): admiration, amusement, anger, annoyance, approval, caring, confusion, curiosity, desire, disappointment, disapproval, disgust, embarrassment, excitement, fear, gratitude, grief, joy, love, nervousness, optimism, pride, realization, relief, remorse, sadness, surprise, neutral
• Date: 2026-07-28 16:39 UTC

This model uses a custom classification head (mean pooling over a backbone without a native sequence-classification class), so loading requires trust_remote_code=True. vLLM serving requires a standard architecture.

Evaluation

Metric	Value
f1_micro @ 0.5	0.0000
f1_macro @ 0.5	0.0000
f1_micro @ tuned	0.2436
f1_macro @ tuned	0.0786

Per-label decision thresholds tuned on the eval split are stored in config.classifier_thresholds.

Usage

import torch
from transformers import AutoModelForSequenceClassification, AutoTokenizer

model = AutoModelForSequenceClassification.from_pretrained("davanstrien/lfm25-encoder-goemotions-smoke", trust_remote_code=True)
tokenizer = AutoTokenizer.from_pretrained("davanstrien/lfm25-encoder-goemotions-smoke", trust_remote_code=True)

inputs = tokenizer("your text here", return_tensors="pt", truncation=True)
probs = torch.sigmoid(model(**inputs).logits)[0]
thresholds = torch.tensor(model.config.classifier_thresholds)  # tuned on validation
labels = [model.config.id2label[i] for i in (probs >= thresholds).nonzero().flatten().tolist()]
print(labels)

Reproduction

Produced on Hugging Face Jobs (gpu) with the train-classifier.py recipe from uv-scripts. Run it yourself:

hf jobs uv run --flavor gpu --secrets HF_TOKEN \
    https://huggingface.co/datasets/uv-scripts/classification/raw/main/train-classifier.py \
    google-research-datasets/go_emotions davanstrien/lfm25-encoder-goemotions-smoke --label-column labels