NisargUpadhyay/Qwen3.5-2B-Filter-Extractor

Qwen3.5-2B Filter Extractor (merged)

Model: Qwen3.5-2B fine-tuned adapter merged into base weights.

Files included:

• —model.safetensors — merged weights
• —config.json — model config
• —tokenizer.json and tokenizer_config.json — tokenizer files
• —generation_config.json — generation defaults
• —chat_template.jinja — chat formatting template

Notes:

• —Architecture: Qwen3.5-2B (hybrid linear + full-attention)
• —dtype: bfloat16
• —This model may not be directly compatible with Ollama/llama.cpp due to hybrid attention layers.

Usage (Transformers):

from transformers import AutoModelForCausalLM, AutoTokenizer
model = AutoModelForCausalLM.from_pretrained("NisargUpadhyay/Qwen3.5-2B-Filter-Extractor")
tokenizer = AutoTokenizer.from_pretrained("NisargUpadhyay/Qwen3.5-2B-Filter-Extractor")
print(tokenizer.encode("Hello\n"))

License: Add license details here.