barbarabhb/nl2sh-qwen25-coder-1.5b-tpu

TEXT GENERATIONPricing:Input $0.04 / Cached $0.008 / Output $0.08Concurrent Unit Cost:1Model Size:1.5BQuant:BF16Context Size:32kTool Calling:SupportedPublished:Aug 24, 2026License:apache-2.0Architecture:Transformer Open Weights Featherless Exclusive Cold

The barbarabhb/nl2sh-qwen25-coder-1.5b-tpu model is a 1.5 billion parameter Qwen2.5-Coder-1.5B-Instruct LoRA fine-tune, specifically optimized for natural language to shell command translation. Trained on the nl2sh-v3f dataset using a Kaggle TPU v3-8, it excels at converting natural language queries into accurate shell commands. This model features a 32768 token context length and demonstrates a 0.6367 pass rate on the InterCode-ALFA benchmark for its Q4_K_M variant, making it suitable for command-line automation tasks.

Loading preview...

Model Overview

The barbarabhb/nl2sh-qwen25-coder-1.5b-tpu is a specialized 1.5 billion parameter language model, fine-tuned from Qwen/Qwen2.5-Coder-1.5B-Instruct. Its primary function is to translate natural language instructions into executable shell commands.

Key Capabilities

  • Natural Language to Shell Command Translation: Optimized for converting user queries into precise shell commands, addressing a common need for command-line automation.
  • Robustness: Specifically designed to handle bare greetings by responding with simple echo hello commands, preventing unintended network interactions and addressing known issues in similar models.
  • Performance: Achieves a 0.6367 pass rate on the InterCode-ALFA benchmark (300 tasks, temperature 0) for its Q4_K_M variant, indicating strong performance in generating correct shell commands.
  • Training: Fine-tuned on the nl2sh-v3f dataset (228k pairs) using a Kaggle TPU v3-8, employing a hand-rolled PyTorch/XLA loop with specific training parameters (micro-batch 8 x grad-accum 4, seq 224, assistant-only loss, lr 2e-4 cosine, 1 epoch / 7130 optimizer steps).

Good For

  • Command-line Automation: Ideal for applications requiring the automatic generation of shell commands from natural language input.
  • Developer Tools: Can be integrated into IDEs or command-line interfaces to assist users in formulating complex commands.
  • Educational Purposes: Useful for teaching shell scripting by demonstrating how natural language translates to commands.

Finetunes
1 models