tensortalk

TensorTalk

Merged fine-tuned TensorTalk model for Universiti Malaya handbook question answering.

This repository contains the merged model weights, tokenizer files, and generation configuration needed to load the model with transformers.

Files

• model.safetensors: merged fine-tuned model weights
• config.json: Qwen3 causal language model configuration
• tokenizer.json, tokenizer_config.json, chat_template.jinja: tokenizer and chat formatting files
• generation_config.json: generation defaults

Loading

from transformers import AutoModelForCausalLM, AutoTokenizer

model_id = "nfdlh/tensortalk"

tokenizer = AutoTokenizer.from_pretrained(model_id, trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained(
    model_id,
    torch_dtype="auto",
    device_map="auto",
    trust_remote_code=True,
)