richarderkhov/arise-sustech_-_llm4decompile-1.3b-gguf overview

{
  "metadata": {},
  "card_data": {
    "frontmatter": {},
    "hero_image_url": "",
    "summary": "Quantization made by Richard Erkhov. Github Discord Request more models llm4decompile-1.3b - GGUF | Name | Quant method | Size | | ---- | ---- | ---- | | llm4decompile-1.3b.Q2_K.gguf | Q2_K | 0.52GB | | llm4decompile-1.3b.IQ3_XS.gguf | IQ3_XS | 0.57GB | | llm4decompile-1.3b.IQ3_S.gguf | IQ3_S | 0.6GB | | llm4decompile-1.3b.Q3_K_S.gguf | Q3_K_S | 0.6GB | | llm4decompile-1.3b.IQ3_M.gguf | IQ3_M | 0.63GB | | llm4decompile-1.3b.Q3_K.gguf | Q3_K | 0.66GB | | llm4decompile-1.3b.Q3_K_M.gguf | Q3_K_M | 0.66GB | | llm4decompile-1.3b.Q3_K_L.gguf | Q3_K_L | 0.69GB | | llm4decompile-1.3b.IQ4_XS.gguf | IQ4_XS | 0.7GB | | llm4decompile-1.3b.Q4_0.gguf | Q4_0 | 0.72GB | | llm4decompile-1.3b.IQ4_NL.gguf | IQ4_NL | 0.73GB | | llm4decompile-1.3b.Q4_K_S.gguf | Q4_K_S | 0.76GB | | llm4decompile-1.3b.Q4_K.gguf | Q4_K | 0.81GB | | llm4decompile-1.3b.Q4_K_M.gguf | Q4_K_M | 0.81GB | | llm4decompile-1.3b.Q4_1.gguf | Q4_1 | 0.8GB | | llm4decompile-1.3b.Q5_0.gguf | Q5_0 | 0.87GB | | llm4decompile-1.3b.Q5_K_S.gguf | Q5_K_S | 0.89GB | | llm4decompile-1.3b.Q5_K.gguf | Q5_K | 0.93GB | | llm4decompile-1.3b.Q5_K_M.gguf | Q5_K_M | 0.93GB | | llm4decompile-1.3b.Q5_1.gguf | Q5_1 | 0.95GB | | llm4decompile-1.3b.Q6_K.gguf | Q6_K | 1.09GB | | llm4decompile-1.3b.Q8_0.gguf | Q8_0 | 1.33GB | Original model description: --- language: code        #  {output_file}.s'#disassemble the binary file into assembly instructions subprocess.run(compile_command, shell=True, check=True) input_asm = '' with open(output_file+'.s') as f:#asm file asm= f.read() asm = asm.split('Disassembly of section .text:')[-1].strip() for tmp in asm.split('\\n'): tmp_asm = tmp.split('\\t')[-1]#remove the binary code tmp_asm = tmp_asm.split('#')[0].strip()#remove the comments input_asm+=tmp_asm+'\\n' input_asm = re.sub(zeros_pattern, '', input_asm) before = f\"# This is the assembly code with {opt_state} optimization:\\n\"#prompt after = \"\\n# What is the source code?\\n\"#prompt input_asm_prompt = before+input_asm.strip()+after with open(fileName +'_' + opt_state +'.asm','w',encoding='utf-8') as f: f.write(input_asm_prompt) ` Then use LLM4Decompile to translate the assembly instructions into C: `python from transformers import AutoTokenizer, AutoModelForCausalLM import torch model_path = 'arise-sustech/llm4decompile-1.3b' tokenizer = AutoTokenizer.from_pretrained(model_path) model = AutoModelForCausalLM.from_pretrained(model_path,torch_dtype=torch.bfloat16).cuda() with open(fileName +'_' + opt_state +'.asm','r') as f:#original file asm_func = f.read() inputs = tokenizer(asm_func, return_tensors=\"pt\").to(model.device) with torch.no_grad(): outputs = model.generate(**inputs, max_new_tokens=512) c_func_decompile = tokenizer.decode(outputs[0][len(inputs[0]):-1]) ` ### 4. License This code repository is licensed under the DeepSeek License. ### 5. Contact If you have any questions, please raise an issue. ### 6. Citation ` @misc{tan2024llm4decompile, title={LLM4Decompile: Decompiling Binary Code with Large Language Models}, author={Hanzhuo Tan and Qi Luo and Jing Li and Yuqun Zhang}, year={2024}, eprint={2403.05286}, archivePrefix={arXiv}, primaryClass={cs.PL} } ``",
    "quick_links": [],
    "benchmark_table_html": "",
    "readme_markdown": "Quantization made by Richard Erkhov.\n\n[Github](https://github.com/RichardErkhov)\n\n[Discord](https://discord.gg/pvy7H8DZMG)\n\n[Request more models](https://github.com/RichardErkhov/quant_request)\n\n\nllm4decompile-1.3b - GGUF\n- Model creator: https://huggingface.co/arise-sustech/\n- Original model: https://huggingface.co/arise-sustech/llm4decompile-1.3b/\n\n\n| Name | Quant method | Size |\n| ---- | ---- | ---- |\n| [llm4decompile-1.3b.Q2_K.gguf](https://huggingface.co/RichardErkhov/arise-sustech_-_llm4decompile-1.3b-gguf/blob/main/llm4decompile-1.3b.Q2_K.gguf) | Q2_K | 0.52GB |\n| [llm4decompile-1.3b.IQ3_XS.gguf](https://huggingface.co/RichardErkhov/arise-sustech_-_llm4decompile-1.3b-gguf/blob/main/llm4decompile-1.3b.IQ3_XS.gguf) | IQ3_XS | 0.57GB |\n| [llm4decompile-1.3b.IQ3_S.gguf](https://huggingface.co/RichardErkhov/arise-sustech_-_llm4decompile-1.3b-gguf/blob/main/llm4decompile-1.3b.IQ3_S.gguf) | IQ3_S | 0.6GB |\n| [llm4decompile-1.3b.Q3_K_S.gguf](https://huggingface.co/RichardErkhov/arise-sustech_-_llm4decompile-1.3b-gguf/blob/main/llm4decompile-1.3b.Q3_K_S.gguf) | Q3_K_S | 0.6GB |\n| [llm4decompile-1.3b.IQ3_M.gguf](https://huggingface.co/RichardErkhov/arise-sustech_-_llm4decompile-1.3b-gguf/blob/main/llm4decompile-1.3b.IQ3_M.gguf) | IQ3_M | 0.63GB |\n| [llm4decompile-1.3b.Q3_K.gguf](https://huggingface.co/RichardErkhov/arise-sustech_-_llm4decompile-1.3b-gguf/blob/main/llm4decompile-1.3b.Q3_K.gguf) | Q3_K | 0.66GB |\n| [llm4decompile-1.3b.Q3_K_M.gguf](https://huggingface.co/RichardErkhov/arise-sustech_-_llm4decompile-1.3b-gguf/blob/main/llm4decompile-1.3b.Q3_K_M.gguf) | Q3_K_M | 0.66GB |\n| [llm4decompile-1.3b.Q3_K_L.gguf](https://huggingface.co/RichardErkhov/arise-sustech_-_llm4decompile-1.3b-gguf/blob/main/llm4decompile-1.3b.Q3_K_L.gguf) | Q3_K_L | 0.69GB |\n| [llm4decompile-1.3b.IQ4_XS.gguf](https://huggingface.co/RichardErkhov/arise-sustech_-_llm4decompile-1.3b-gguf/blob/main/llm4decompile-1.3b.IQ4_XS.gguf) | IQ4_XS | 0.7GB |\n| [llm4decompile-1.3b.Q4_0.gguf](https://huggingface.co/RichardErkhov/arise-sustech_-_llm4decompile-1.3b-gguf/blob/main/llm4decompile-1.3b.Q4_0.gguf) | Q4_0 | 0.72GB |\n| [llm4decompile-1.3b.IQ4_NL.gguf](https://huggingface.co/RichardErkhov/arise-sustech_-_llm4decompile-1.3b-gguf/blob/main/llm4decompile-1.3b.IQ4_NL.gguf) | IQ4_NL | 0.73GB |\n| [llm4decompile-1.3b.Q4_K_S.gguf](https://huggingface.co/RichardErkhov/arise-sustech_-_llm4decompile-1.3b-gguf/blob/main/llm4decompile-1.3b.Q4_K_S.gguf) | Q4_K_S | 0.76GB |\n| [llm4decompile-1.3b.Q4_K.gguf](https://huggingface.co/RichardErkhov/arise-sustech_-_llm4decompile-1.3b-gguf/blob/main/llm4decompile-1.3b.Q4_K.gguf) | Q4_K | 0.81GB |\n| [llm4decompile-1.3b.Q4_K_M.gguf](https://huggingface.co/RichardErkhov/arise-sustech_-_llm4decompile-1.3b-gguf/blob/main/llm4decompile-1.3b.Q4_K_M.gguf) | Q4_K_M | 0.81GB |\n| [llm4decompile-1.3b.Q4_1.gguf](https://huggingface.co/RichardErkhov/arise-sustech_-_llm4decompile-1.3b-gguf/blob/main/llm4decompile-1.3b.Q4_1.gguf) | Q4_1 | 0.8GB |\n| [llm4decompile-1.3b.Q5_0.gguf](https://huggingface.co/RichardErkhov/arise-sustech_-_llm4decompile-1.3b-gguf/blob/main/llm4decompile-1.3b.Q5_0.gguf) | Q5_0 | 0.87GB |\n| [llm4decompile-1.3b.Q5_K_S.gguf](https://huggingface.co/RichardErkhov/arise-sustech_-_llm4decompile-1.3b-gguf/blob/main/llm4decompile-1.3b.Q5_K_S.gguf) | Q5_K_S | 0.89GB |\n| [llm4decompile-1.3b.Q5_K.gguf](https://huggingface.co/RichardErkhov/arise-sustech_-_llm4decompile-1.3b-gguf/blob/main/llm4decompile-1.3b.Q5_K.gguf) | Q5_K | 0.93GB |\n| [llm4decompile-1.3b.Q5_K_M.gguf](https://huggingface.co/RichardErkhov/arise-sustech_-_llm4decompile-1.3b-gguf/blob/main/llm4decompile-1.3b.Q5_K_M.gguf) | Q5_K_M | 0.93GB |\n| [llm4decompile-1.3b.Q5_1.gguf](https://huggingface.co/RichardErkhov/arise-sustech_-_llm4decompile-1.3b-gguf/blob/main/llm4decompile-1.3b.Q5_1.gguf) | Q5_1 | 0.95GB |\n| [llm4decompile-1.3b.Q6_K.gguf](https://huggingface.co/RichardErkhov/arise-sustech_-_llm4decompile-1.3b-gguf/blob/main/llm4decompile-1.3b.Q6_K.gguf) | Q6_K | 1.09GB |\n| [llm4decompile-1.3b.Q8_0.gguf](https://huggingface.co/RichardErkhov/arise-sustech_-_llm4decompile-1.3b-gguf/blob/main/llm4decompile-1.3b.Q8_0.gguf) | Q8_0 | 1.33GB |\n\n\n\n\nOriginal model description:\n---\nlanguage: code        # <-- my language\nwidget:\n - text: \"# This is the assembly code with O0 optimization:\\n<func0>:\\nendbr64\\nlea    (%rdi,%rsi,1),%eax\\nretq\\n# What is the source code?\\n\"\nlicense: other\ntags:\n- decompile\n- binary\n---\n\n### 1. Introduction of LLM4Decompile\n\nLLM4Decompile aims to decompile x86 assembly instructions into C. It is finetuned from Deepseek-Coder on 4B tokens of assembly-C pairs compiled from AnghaBench.\n\n- **Github Repository:** [LLM4Decompile](https://github.com/albertan017/LLM4Decompile)\n\n- **Paper link:** For more details check out the [paper](https://arxiv.org/abs/2403.05286).\n\n### 2. Evaluation Results\n| Model              | Re-compilability |           |           |           |           | Re-executability |           |           |           |           |\n|--------------------|:----------------:|:---------:|:---------:|:---------:|:---------:|:----------------:|-----------|-----------|-----------|:---------:|\n| Optimization-level          | O0               | O1        | O2        | O3        | Avg.      | O0               | O1        | O2        | O3        | Avg.      |\n| GPT4               | 0.92             | 0.94      | 0.88      | 0.84      | 0.895     | 0.1341           | 0.1890    | 0.1524    | 0.0854    | 0.1402    |\n| DeepSeek-Coder-33B |   0.0659         |   0.0866  |   0.1500  |   0.1463  |   0.1122  |   0.0000         |   0.0000  |   0.0000  |   0.0000  |   0.0000  |\n| LLM4Decompile-1b   |   0.8780         |   0.8732  |   0.8683  |   0.8378  |   0.8643  |   0.1573         |   0.0768  |   0.1000  |   0.0878  |   0.1055  |\n| LLM4Decompile-6b   |   0.8817         |   0.8951  |   0.8671  |   0.8476  |   0.8729  |   0.3000         |   0.1732  |   0.1988  |   0.1841  |   0.2140  |\n| LLM4Decompile-33b  |   0.8134         |   0.8195  |   0.8183  |   0.8305  |   0.8204  |   0.3049         |   0.1902  |   0.1817  |   0.1817  |   0.2146  |\n\n\n\n### 3. How to Use\nHere give an example of how to use our model.\nFirst compile the C code into binary, disassemble the binary into assembly instructions:\n```python\nimport subprocess\nimport os\nimport re\n\ndigit_pattern = r'\\b0x[a-fA-F0-9]+\\b'# binary codes in Hexadecimal\nzeros_pattern = r'^0+\\s'#0s\nOPT = [\"O0\", \"O1\", \"O2\", \"O3\"]\nfileName = 'path/to/file'\nwith open(fileName+'.c','r') as f:#original file\n    c_func = f.read()\nfor opt_state in OPT:\n    output_file = fileName +'_' + opt_state\n    input_file = fileName+'.c'\n    compile_command = f'gcc -c -o {output_file}.o {input_file} -{opt_state} -lm'#compile the code with GCC on Linux\n    subprocess.run(compile_command, shell=True, check=True)\n    compile_command = f'objdump -d {output_file}.o > {output_file}.s'#disassemble the binary file into assembly instructions\n    subprocess.run(compile_command, shell=True, check=True)\n    \n    input_asm = ''\n    with open(output_file+'.s') as f:#asm file\n        asm= f.read()\n    asm = asm.split('Disassembly of section .text:')[-1].strip()\n    for tmp in asm.split('\\n'):\n        tmp_asm = tmp.split('\\t')[-1]#remove the binary code\n        tmp_asm = tmp_asm.split('#')[0].strip()#remove the comments\n        input_asm+=tmp_asm+'\\n'\n    input_asm = re.sub(zeros_pattern, '', input_asm)\n    before = f\"# This is the assembly code with {opt_state} optimization:\\n\"#prompt\n    after = \"\\n# What is the source code?\\n\"#prompt\n    input_asm_prompt = before+input_asm.strip()+after\n    with open(fileName +'_' + opt_state +'.asm','w',encoding='utf-8') as f:\n        f.write(input_asm_prompt)\n```\n\nThen use LLM4Decompile to translate the assembly instructions into C:\n```python\nfrom transformers import AutoTokenizer, AutoModelForCausalLM\nimport torch\n\nmodel_path = 'arise-sustech/llm4decompile-1.3b'\ntokenizer = AutoTokenizer.from_pretrained(model_path)\nmodel = AutoModelForCausalLM.from_pretrained(model_path,torch_dtype=torch.bfloat16).cuda()\n\nwith open(fileName +'_' + opt_state +'.asm','r') as f:#original file\n    asm_func = f.read()\ninputs = tokenizer(asm_func, return_tensors=\"pt\").to(model.device)\nwith torch.no_grad():\n    outputs = model.generate(**inputs, max_new_tokens=512)\nc_func_decompile = tokenizer.decode(outputs[0][len(inputs[0]):-1])\n```\n\n### 4. License\nThis code repository is licensed under the DeepSeek License.\n\n### 5. Contact\n\nIf you have any questions, please raise an issue.\n\n### 6. Citation\n```\n@misc{tan2024llm4decompile,\n      title={LLM4Decompile: Decompiling Binary Code with Large Language Models}, \n      author={Hanzhuo Tan and Qi Luo and Jing Li and Yuqun Zhang},\n      year={2024},\n      eprint={2403.05286},\n      archivePrefix={arXiv},\n      primaryClass={cs.PL}\n}\n```\n\n",
    "related_quantizations": []
  },
  "tags": [
    "gguf",
    "arxiv:2403.05286",
    "endpoints_compatible",
    "region:us"
  ],
  "likes": 0,
  "downloads": 111,
  "gated": false,
  "private": false,
  "last_modified": "2024-06-30T00:27:37.000Z",
  "created_at": "2024-06-29T23:42:23.000Z",
  "pipeline_tag": "",
  "library_name": ""
}

{
  "_id": "66809bdf81998a6a089fa2ec",
  "id": "RichardErkhov/arise-sustech_-_llm4decompile-1.3b-gguf",
  "modelId": "RichardErkhov/arise-sustech_-_llm4decompile-1.3b-gguf",
  "sha": "66a91c63356d1c04f551c7098b8920c3ac630c49",
  "createdAt": "2024-06-29T23:42:23.000Z",
  "lastModified": "2024-06-30T00:27:37.000Z",
  "author": "RichardErkhov",
  "downloads": 111,
  "likes": 0,
  "gated": false,
  "private": false,
  "pipeline_tag": "",
  "library_name": "",
  "siblings_count": 24
}