piiceetah-unmask-qwen2.5-0.5b-instruct

UnmaskingQwen3 for Token Classification

This model is a fine-tuned version of a custom UnmaskingQwen3ForTokenClassification model for token classification tasks.

Model Details

• Model Type: Custom UnmaskingQwen3ForTokenClassification
• Task: Token Classification (NER/POS/Chunking)
• Training Framework: Transformers + Accelerate

Usage

from transformers import AutoTokenizer, AutoModelForTokenClassification

# Load the model and tokenizer
tokenizer = AutoTokenizer.from_pretrained("your-username/your-model-name", trust_remote_code=True)
model = AutoModelForTokenClassification.from_pretrained("your-username/your-model-name", trust_remote_code=True)

# Use for inference
inputs = tokenizer(["Your text here"], return_tensors="pt", is_split_into_words=False)
outputs = model(**inputs)
predictions = outputs.logits.argmax(dim=-1)

Training Details

• Training Data: ['automated-analytics/ai4privacy-pii-masking-en-v1-ner-coarse', 'automated-analytics/gretel-pii-masking-en-v1-ner-coarse']
• Learning Rate: 5e-05
• Batch Size: 128
• Epochs: 3
• Max Length: 128

Important Note

This model uses a custom model class. Make sure to use trust_remote_code=True when loading the model.