---
license: apache-2.0
base_model: Qwen/Qwen2.5-Coder-1.5B-Instruct
tags:
  - microtensor
  - ctx_compress
  - llama
---

# Llama Context Compress SLM

Built with Llama.

Specialised small model for the Microtensor `ctx_compress` arena: compress a
coding-agent session into a short JSON summary.

**Base:** `Qwen/Qwen2.5-Coder-1.5B-Instruct`

Outputs follow the Llama 3.1 licence, so a distributed fine tune must carry the Llama name and show Built with Llama.

Dataset: [nebius/SWE-agent-trajectories](https://huggingface.co/datasets/nebius/SWE-agent-trajectories)

## Intended use

Raw completion against the arena `slm_prompt`. Greedy decode, stop at
`<|im_end|>`, `<|endoftext|>`, `<|end|>`, max 3072 new tokens.