Overview

This specialized model is engineered for high-autonomy software engineering tasks, specifically optimized for resolving complex GitHub-style issues. By leveraging a reasoning-heavy architecture and a critic-loop mechanism—utilizing a Qwen-14B backbone—it excels at navigating large codebases and performing precise bug fixes. Unlike general-purpose LLMs, this model is tuned for the SWE-bench and SWE-gym environments, focusing on reducing hallucinated API calls and improving the success rate of autonomous PR generation. For developers, this means a tool capable of handling multi-file dependencies and logical reasoning across an entire repository rather than just snippet completion. It integrates well into CI/CD pipelines where automated triage and initial patching are required.

Highlights

Optimized for autonomous software engineering and bug resolution
Integrated critic loop for higher code correctness
Capable of reasoning across large, multi-file repositories
Apache-2.0 licensed for flexible commercial integration
High performance on SWE-bench and SWE-gym benchmarks

Usage

Install

# Install Hugging Face transformers
pip install transformers torch

SDK Usage

# Load model with transformers
from transformers import AutoModel, AutoTokenizer

model = AutoModel.from_pretrained("secmlr/SWE-BENCH-generation_claude_reasoning_llm_correct_swe_gym_1500_plus_critic_qwen_code_14b")
tokenizer = AutoTokenizer.from_pretrained("secmlr/SWE-BENCH-generation_claude_reasoning_llm_correct_swe_gym_1500_plus_critic_qwen_code_14b")

Hugging Face Download

We recommend downloading the model via the Hugging Face CLI or Hub SDK.

Guidance：Before downloading, install huggingface_hub with:

Guidance

pip install -U huggingface_hub

CLI Download

Download the full repository

huggingface-cli download secmlr/SWE-BENCH-generation_claude_reasoning_llm_correct_swe_gym_1500_plus_critic_qwen_code_14b

Download a single file to a local folder (e.g. config.json into ./dir)

huggingface-cli download secmlr/SWE-BENCH-generation_claude_reasoning_llm_correct_swe_gym_1500_plus_critic_qwen_code_14b config.json --local-dir ./dir

See the official docs for more CLI options

SDK Download

# 模型下载
from huggingface_hub import snapshot_download
model_dir = snapshot_download('secmlr/SWE-BENCH-generation_claude_reasoning_llm_correct_swe_gym_1500_plus_critic_qwen_code_14b')

Git Download

Make sure git-lfs is installed first

Git Download

git lfs install
git clone https://huggingface.co/secmlr/SWE-BENCH-generation_claude_reasoning_llm_correct_swe_gym_1500_plus_critic_qwen_code_14b

To skip LFS large-file downloads, use:

Skip LFS

GIT_LFS_SKIP_SMUDGE=1 git clone https://huggingface.co/secmlr/SWE-BENCH-generation_claude_reasoning_llm_correct_swe_gym_1500_plus_critic_qwen_code_14b

Model files are hosted on the Hugging Face Hub — download directly via HF CLI / SDK / Git, not through this site.

PyTorch / Transformers Usage

Install Transformers

pip install -U transformers torch

Load the model and run inference

from transformers import AutoModelForCausalLM, AutoTokenizer

model = AutoModelForCausalLM.from_pretrained('secmlr/SWE-BENCH-generation_claude_reasoning_llm_correct_swe_gym_1500_plus_critic_qwen_code_14b')
tokenizer = AutoTokenizer.from_pretrained('secmlr/SWE-BENCH-generation_claude_reasoning_llm_correct_swe_gym_1500_plus_critic_qwen_code_14b')

Full Documentation

来源: HuggingFace

---
library_name: transformers
license: apache-2.0
base_model: Qwen/Qwen2.5-Coder-14B-Instruct
tags:

llama-factory

full

generated_from_trainer

model-index:

name: SWE-BENCH-generation_claude_reasoning_llm_correct_swe_gym_1500_plus_critic_qwen_code_14b

results: []
---