---
datasets:
- HuggingFaceH4/ultrachat_200k
base_model:
- EssentialAI/rnj-1-instruct
---
# rnj-1-instruct-nvfp4

**Format:** NVFP4 — weights & activations quantized to FP4 with dual scaling.  
**Base model:** `EssentialAI/rnj-1-instruct`  
**How it was made:** One-shot calibration with LLM Compressor (NVFP4 recipe), long-seq calibration with HuggingFaceH4/ultrachat_200k.  

> Notes: Keep `lm_head` in high precision; calibrate on long, domain-relevant sequences.

Check the original model card for information about this model.

# Running the model with VLLM in Docker

Note: I couldn't get this one to run in VLLM. I'm not sure if there's a trick to run Gemma 3 based models in VLLM. If anyone knows a trick I can update the model card with the updated command.

```sh
sudo docker run --runtime nvidia --gpus all -p 8000:8000 --ipc=host vllm/vllm-openai:nightly --model Firworks/rnj-1-instruct-nvfp4 --dtype auto --max-model-len 32768
```
This was tested on an RTX Pro 6000 Blackwell cloud instance.

If there are other models you're interested in seeing quantized to NVFP4 for use on the DGX Spark, or other modern Blackwell (or newer) cards let me know. I'm trying to make more NVFP4 models available to allow more people to try them out.