Downloads · 30 days
18.3K
11% of all-time downloads
poolside/Laguna-S-2.1-DFlash-NVFP4
Laguna-S-2.1-DFlash-NVFP4 is a machine learning model from poolside. Use it for the machine learning task on the model card, and read the license before you ship it in a product. It is set up for speculators.
<p align="center" <img alt="poolside-banner" src="https://poolside.ai/assets/laguna/laguna-s-2-1-banner.svg" width="800px" </p
Downloads · 30 days
18.3K
11% of all-time downloads
All-time downloads
172K
Public
Parameters
1.1B
8.9 GB on disk
Likes
32
Public
Click a slice to open those files.
.safetensors2.2 GB · 100%
From the Hugging Face model README
DFlash speculator for the NVFP4 target poolside/Laguna-S-2.1-NVFP4. The speculator is a 6-layer Laguna-style draft model (BF16); pair it with the NVFP4 base for lower-latency serving via speculative decoding.
Trained: e0630_rhiemann_baseline SFT, DFlash Stage-2, 15k steps. Recommended
serving setting: num_speculative_tokens=7.
DFlash upstream support is in progress (vLLM #46853, SGLang #29446, TRT-LLM #15666). Use
poolside/Laguna-S-2.1-NVFP4 as the target model.
Measured with TP=2, temperature=0, and num_speculative_tokens=15.
| Concurrency | GSM8K | MATH-500 | HumanEval | MBPP | MT-Bench |
|---|---|---|---|---|---|
| 1 | 3.324x | 2.893x | 3.692x | 2.426x | 2.338x |
| 4 | 2.498x | 2.174x | 2.742x | 1.848x | 1.831x |
| 8 | 2.279x | 1.948x | 2.634x | 1.719x | 1.772x |
| 16 | 2.302x | 1.965x | 2.626x | 1.731x | 1.672x |
| Concurrency | GSM8K | MATH-500 | HumanEval | MBPP | MT-Bench |
|---|---|---|---|---|---|
| 1 | 5.775 | 4.942 | 6.438 | 4.171 | 4.017 |
| 4 | 5.804 | 4.935 | 6.298 | 4.161 | 4.091 |
| 8 | 5.719 | 4.885 | 6.537 | 4.193 | 4.373 |
| 16 | 5.758 | 4.896 | 6.412 | 4.123 | 3.981 |