Model Overview

This model, laion/nl2bash-swesmith-stack-bugsseq, is an 8 billion parameter language model derived from the Qwen/Qwen3-8B architecture. It has been specifically fine-tuned by laion to excel in tasks related to natural language to bash command translation and reasoning.

Key Capabilities

• Natural Language to Bash Translation: Optimized for converting natural language instructions into executable bash commands.
• Reasoning Tasks: Enhanced for reasoning, particularly within the context of code and command generation.
• Specialized Dataset Training: Fine-tuned on a unique combination of datasets:
  • penfever/GLM-4.6-nl2bash-verified-32ep-32k-reasoning
  • penfever/GLM-4.6-swesmith-32ep-131k-nosumm-reasoning
  • penfever/GLM-4.6-stackexchange-overflow-sandboxes-32eps-65k-reasoning
  • penfever/GLM-4.6-inferredbugs-32ep-65k-reasoning

Training Details

The model was trained with a learning rate of 4e-05, using a cosine learning rate scheduler with a 0.1 warmup ratio over 7 epochs. It utilized a total batch size of 16 across 8 GPUs, employing the AdamW_Torch_Fused optimizer. The training environment included Transformers 4.56.1 and Pytorch 2.9.1+cu128.

Intended Use Cases

This model is particularly well-suited for applications requiring the generation of bash commands from natural language queries, as well as tasks involving code-related reasoning and bug inference, leveraging its specialized training on relevant datasets.