ArtificialGuyBR

Home / Projects / Qwen2 1.5B Synthia I — Instruction-Tuned LLM

LLM · Text Generation

Qwen2 1.5B Synthia I — Instruction-Tuned LLM

Fine-tuned Qwen2-1.5B LLM trained on Synthia v1.5-I data for instruction following and conversational text generation.

About this model

QWEN 2 1.5B Synthia I is a causal language model fine-tuned from Qwen/Qwen2-1.5B on the Synthia v1.5-I dataset, which contains over 20.7k instruction-following examples. The result is a compact 1.5B-parameter model whose instruction-following behavior is enhanced over the base release while keeping Qwen2's strengths in language understanding, structured data processing, multilingual support, and long context handling. It is built for text generation and conversational AI applications.

This is a full fine-tune, not a LoRA, so no trigger word or instance prompt is required — the model responds to instructions directly. The training run used Axolotl 0.4.1 with Transformers 4.45.0.dev0, a learning rate of 1e-05, 3 epochs over a sequence length of 4096 with sample packing, cosine LR scheduling with 100 warmup steps, and a total effective batch size of 40.

Use it like any other Qwen2 text-generation model: load it with the Transformers pipeline or run it through text-generation-inference (the repository is endpoints-compatible). GPU memory needs are modest for a 1.5B model, but no official VRAM figures are published. The model is released under the Apache 2.0 license and is English-language, tagged as instruction-tuning and conversational.

Learn more

Read the in-depth guide about Qwen2 1.5B Synthia I — Instruction-Tuned LLM

Project signals

  • 105 Hugging Face downloads
  • 2 Hugging Face likes

Topics

transformers · pytorch · safetensors · qwen2 · text-generation · instruction-tuning · conversational · base_model:qwen/qwen2-1.5b · base_model:finetune:qwen/qwen2-1.5b · text-generation-inference

Explore the source