Cwen-7B-Instruct

Introduction

Cwen-7B-Instruct is a fine-tuned version of Qwen2-7B-Instruct, optimized using the cData coding dataset to enhance its coding capabilities across various languages, with a primary focus on low-level ones.

Quickstart

Here provides a code snippet with apply_chat_template to show you how to load the tokenizer and model and how to generate contents.

from transformers import AutoModelForCausalLM, AutoTokenizer
device = "cuda" # the device to load the model onto

model = AutoModelForCausalLM.from_pretrained(
    "motexture/Cwen-7B-Instruct",
    torch_dtype="auto",
    device_map="auto"
)
tokenizer = AutoTokenizer.from_pretrained("motexture/Cwen-7B-Instruct")

prompt = "Write a C++ program that demonstrates the concept of separate compilation and linkage using namespaces and header files. The program should consist of multiple source files, each containing a portion of the program's code, and a header file that contains the interface information for the program.\n\nThe program should define a namespace my_namespace that contains a class  MyClass with a member function print() that takes an integer as an argument. The program should also define a function main() that uses an object of the MyClass class to print a message.\n\nThe program should be compiled and linked separately, with each source file being compiled individually and then linked together to form the final executable."
messages = [
    {"role": "system", "content": "You are a helpful assistant."},
    {"role": "user", "content": prompt}
]
text = tokenizer.apply_chat_template(
    messages,
    tokenize=False,
    add_generation_prompt=True
)
model_inputs = tokenizer([text], return_tensors="pt").to(device)

generated_ids = model.generate(
        model_inputs.input_ids,
        max_new_tokens=4096,
        do_sample=True,
        temperature=0.3
)
generated_ids = [
    output_ids[len(input_ids):] for input_ids, output_ids in zip(model_inputs.input_ids, generated_ids)
]

response = tokenizer.batch_decode(generated_ids, skip_special_tokens=True)[0]

Citation

@article{qwen2,
  title={Qwen2 Technical Report},
  year={2024}
}