tel1980/qwen-python-slm

TEXT GENERATIONPricing:Input $0.04 / Cached $0.008 / Output $0.08Concurrent Unit Cost:1Model Size:1.5BQuant:BF16Context Size:32kTool Calling:SupportedPublished:Aug 19, 2026License:apache-2.0Architecture:Transformer Open Weights Featherless Exclusive Cold

tel1980/qwen-python-slm is a 1.5 billion parameter Small Language Model (SLM) fine-tuned from Qwen/Qwen2.5-Coder-1.5B-Instruct, specialized in generating code for Data Engineering, Data Science, Machine Learning, and AI Agents. It excels at producing Python, SQL, and PySpark code, with a context length of 32768 tokens. This model is optimized for technical instructions in Portuguese, focusing on data and AI ecosystems.

Loading preview...

Overview

tel1980/qwen-python-slm is a 1.5 billion parameter Small Language Model (SLM) developed by tel1980, specifically fine-tuned for code generation within the Data Engineering, Data Science, Machine Learning, Deep Learning, LLMs, AI Agents, DataOps, and MLOps ecosystems. It is based on the Qwen/Qwen2.5-Coder-1.5B-Instruct architecture and utilizes QLoRA for supervised fine-tuning, adding specialized adapters for Python, SQL, and PySpark.

Key Capabilities

  • Specialized Code Generation: Generates Python scripts for ETL/ELT, data analysis, and ML; SQL queries for SQLite/PostgreSQL; and PySpark code for distributed processing.
  • Domain-Specific Instruction: Recognizes prefixes like /python, /sql, /pyspark, /ml, /llm, /agent, and /dataops to guide code generation towards specific domains.
  • Multilingual Code Output: Supports instructions in Portuguese while generating code in Python, SQL, and PySpark.
  • Efficient Training: Achieved a final loss of 0.508 after 500 steps on an NVIDIA GeForce RTX 5050, with only 1.18% of total parameters being trainable.

Use Cases

  • Prototyping AI agents with frameworks like LangChain, LlamaIndex, and CrewAI.
  • Automating MLOps/DataOps pipelines using tools like Airflow, dbt, and MLflow.
  • Educational purposes and enhancing productivity for data developers.

Limitations

  • Trained with limited resources, potentially leading to hallucinations or non-executable code.
  • SQL validation is syntactic; semantic correctness depends on the provided schema.
  • Performance outside the data/AI domain (e.g., frontend, mobile) may be lower quality.

Finetunes
4 models