lazos/lfm2.5-350m-promql
The lazos/lfm2.5-350m-promql is a small language model, fine-tuned with QLoRA SFT on the LiquidAI/LFM2.5-350M-Base architecture. This model specializes in PromQL optimization, designed to rewrite PromQL expressions based on given instructions. It functions as an instruction-to-rewrite task model, outputting only the improved PromQL expression without additional explanation. A Q4_K_M GGUF quantization is also available for use with llama.cpp/Ollama.
Loading preview...
Model Overview
The lazos/lfm2.5-350m-promql is a specialized small language model developed by lazos, fine-tuned using QLoRA SFT on the LiquidAI/LFM2.5-350M-Base architecture. Its primary function is to act as a PromQL optimization expert, taking an instruction and a PromQL expression, then rewriting the expression accordingly. The model is designed to output only the optimized PromQL expression, without any additional explanations.
Key Capabilities
- PromQL Optimization: Rewrites PromQL expressions based on user instructions.
- Instruction-to-Rewrite Task: Specifically trained for transforming input PromQL queries.
- Compact Size: Built upon a 350M parameter base model, making it efficient for deployment.
- GGUF Quantization: Includes a Q4_K_M GGUF quantization for compatibility with
llama.cppand Ollama, enabling efficient local inference.
Usage and Licensing
The model utilizes a specific prompt format where a system prompt defines its role as a PromQL optimization expert. It expects user input containing an instruction and the PromQL query to be optimized. The training code, data, and benchmarks are publicly available on GitHub.
This model operates under the LFM Open License v1.0, which permits free use for entities with annual revenues under $10M USD. Commercial licenses are required for larger entities.