ahmedheakl/asm2asm-deepseek1.3b-xtokenizer-arm

CISC-to-RISC

A fine-tuned version of deepseek-ai/deepseek-coder-1.3b-instruct specialized in converting x86 assembly code to ARM assembly.

Model Overview

asm2asm-deepseek1.3b-xtokenizer-arm is designed to assist developers in converting x86 assembly instructions to ARM assembly. Leveraging the capabilities of the base model, this fine-tuned variant enhances accuracy and efficiency in assembly code transpilation tasks.

Intended Use

This model is intended for:

• Assembly Code Conversion: Assisting developers in translating x86 assembly instructions to ARM architecture.
• Educational Purposes: Helping learners understand the differences and translation mechanisms between x86 and ARM assembly.
• Code Optimization: Facilitating optimization processes by converting and refining assembly code across architectures.

Limitations

• Dataset Specificity: The model is fine-tuned on a specific dataset, which may limit its performance on assembly instructions outside the training distribution.
• Complex Instructions: May struggle with highly complex or unconventional assembly instructions not well-represented in the training data.
• Error Propagation: Inaccuracies in the generated ARM code can lead to functional discrepancies or bugs if not reviewed.

Training Data

Detailed information about the training dataset is required.

Training Procedure

Training Hyperparameters

The model was trained with the following hyperparameters:

• Learning Rate: 0.0002
• Training Batch Size: 1
• Evaluation Batch Size: 8
• Seed: 42
• Gradient Accumulation Steps: 4
• Total Training Batch Size: 4
• Optimizer: Adam (betas=(0.9, 0.999), epsilon=1e-08)
• Learning Rate Scheduler: Linear
• Number of Epochs: 2

Usage

All models and datasets are available on Hugging Face. Below is an example of how to use the best model for converting x86 assembly to ARM.

Inference Code

import torch
from transformers import AutoModelForCausalLM, AutoTokenizer
from tqdm import tqdm

# Replace 'hf_token' with your Hugging Face token
hf_token = "your_hf_token_here"

model_name = "ahmedheakl/asm2asm-deepseek1.3b-xtokenizer-arm"

instruction = """<｜begin▁of▁sentence｜>You are a helpful coding assistant specialized in converting from x86 to ARM assembly.
### Instruction:
Convert this x86 assembly into ARM
```asm
{asm_x86}
"```"
### Response:
```asm
{asm_arm}
"""

# Load the model
model = AutoModelForCausalLM.from_pretrained(
    model_name,
    token=hf_token,
    device_map="auto",
    torch_dtype=torch.bfloat16,
)

model.config.use_cache = True

# Load the tokenizer
tokenizer = AutoTokenizer.from_pretrained(
    model_name,
    trust_remote_code=True,
    token=hf_token,
)

def inference(asm_x86: str) -> str:
    prompt = instruction.format(asm_x86=asm_x86, asm_arm="")
    inputs = tokenizer(prompt, return_tensors="pt").to("cuda")
    generated_ids = model.generate(
        **inputs,
        use_cache=True,
        num_return_sequences=1,
        max_new_tokens=8000,
        do_sample=False,
        num_beams=4,
        eos_token_id=tokenizer.eos_token_id,
        pad_token_id=tokenizer.pad_token_id,
    )
    outputs = tokenizer.batch_decode(generated_ids)[0]
    torch.cuda.empty_cache()
    torch.cuda.synchronize()
    return outputs.split("```asm\n")[-1].split(f"```{tokenizer.eos_token}")[0]

# Example usage
x86 = "DWORD PTR -248[rbp] movsx rdx"
converted_arm = inference(x86)
print(converted_arm)

Experiments and Results

Model	Average Edit Distance (↓)	Exact Match (↑)	Test Accuracy (↑)
GPT4o	1296	0%	8.18%
DeepSeekCoder2-16B	1633	0%	7.36%
Yi-Coder-9B	1653	0%	6.33%
Yi-Coder-1.5B	275	16.98%	49.69%
DeepSeekCoder-1.3B	107	45.91%	77.23%
DeepSeekCoder-1.3B-xTokenizer-int4	119	46.54%	72.96%
DeepSeekCoder-1.3B-xTokenizer-int8	96	49.69%	75.47%
DeepSeekCoder-1.3B-xTokenizer	165	50.32%	79.25%

Table: Comparison of models' performance on the x86 to ARM transpilation task, measured by Edit Distance (lower is better), Exact Match (higher is better), and Test Accuracy (higher is better). The top section lists pre-existing models, while the bottom section lists models trained by us. The best results in each metric are highlighted in bold.

Citations

If you use this model in your research, please cite it as follows: