Skip to content

[Draft] Update README with recent highlights, benchmarks, and support matrix … #1508

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
trvachov wants to merge 1 commit into
base: main
Choose a base branch
from
+28 −8
Open

[Draft] Update README with recent highlights, benchmarks, and support matrix … #1508

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
34 changes: 28 additions & 6 deletions README.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -21,6 +21,18 @@ NVIDIA BioNeMo Framework is a comprehensive suite of programming tools, librarie
<em> Training benchmarks for ESM-2, a well known protein sequence model using the BERT architecture.</em>
</p>

<p align="center">
Copy link
Collaborator

@zoeyz101 zoeyz101 Mar 11, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

It would be great to add a bit of context here as to what BioNeMo features these benchmarks relate to for a new user reading this. Are we planning to add more than 3 benchmark figures here as well? If there is more than 3, it might be nice to add a subheading for them.

<img src="docs/docs/assets/images/esm2/esm2_low_precision/esm2_8gpu_tflops.png" width="600">
<br>
<em>ESM-2 low-precision training benchmarks (BF16 vs MXFP8 vs NVFP4) on NVIDIA B300 SXM GPUs with FSDP2.</em>
</p>

<p align="center">
<img src="docs/docs/assets/images/recipes/70b-cp-benchmarks.png" width="600">
<br>
<em>Llama 3 70B context parallelism scaling on 36x NVIDIA GB300 NVL36, from 8K to 144K context length.</em>
</p>

## ⚡ Quick Start

```bash
Expand All @@ -43,6 +55,11 @@ cd bionemo-framework/bionemo-recipes/recipes/esm2_native_te/

## Recent News

- 03/09/2026 [Qwen2.5 / Qwen3 model](bionemo-recipes/models/qwen/) with TE acceleration, FP8/MXFP8, KV-cache inference, and bidirectional HF checkpoint conversion.
- 03/05/2026 [ESM2 NVFP4 and MXFP8](bionemo-recipes/recipes/esm2_native_te/README.md#low-precision-performance-benchmarks) low-precision training — up to **2,367 TFLOPS/GPU** on NVIDIA B300 at 15B scale with per-layer precision control.
- 02/23/2026 [Mixtral MoE model](bionemo-recipes/models/mixtral/) with TE `GroupedLinear` for efficient parallel expert computation, FP8/FP4 support, and HF conversion.
- 02/13/2026 [ESM2 PEFT recipe](bionemo-recipes/recipes/esm2_peft_te/) for LoRA fine-tuning with sequence packing support.
- 01/14/2026 [Llama3 Context Parallelism](bionemo-recipes/recipes/llama3_native_te/README.md#performance-benchmarks) — scaling Llama 3 70B to 144K context on 36x GB300 NVL36 with ~65% MFU.
- 10/27/2025 [CodonFM recipe](https://github.com/NVIDIA/bionemo-framework/tree/main/bionemo-recipes/recipes/codonfm_ptl_te) released! This is an accelerated version of the original [research codebase](https://github.com/NVIDIA-Digital-Bio/CodonFM) with [scientific preprint](https://research.nvidia.com/labs/dbr/assets/data/manuscripts/nv-codonfm-preprint.pdf).
- 09/30/2025 Megatron/NeMo 5D parallel BioNeMo Framework image v2.7 [released on NGC](https://catalog.ngc.nvidia.com/orgs/nvidia/teams/clara/containers/bionemo-framework) for both x86 and ARM CPUs.
- 09/01/2025 [bionemo-recipes](https://github.com/NVIDIA/bionemo-framework/tree/main/bionemo-recipes) goes live! Lightweight and portable examples with state-of-the-art training performance you can riff on to meet your needs.
Expand All @@ -61,13 +78,18 @@ A core use-case of the BioNeMo Framework is to help digital biology scientists a
| ---------------------------------------------- | ----------------------------------------------------------------------------------------------------------------------------- | -------------- | ----------- | ------------- | ------ | ---------------- | ------ | ------------------- |
| `models/`<br>`amplify` | TE accelerated protein BERT, pushed to HuggingFace | ✅ Active | ❌ | ✅ | ✅ | 🚧 WIP | ✅ | 🚧 WIP |
| `models/`<br>`esm2` | TE accelerated protein BERT, pushed to HuggingFace | ✅ Active | ❌ | ✅ | ✅ | ✅ | ✅ | ✅ |
| `models/`<br>`llama3` | TE accelerated Llama 3 | ✅ Active | ❌ | 🚧 WIP | ✅ | ✅ | 🚧 WIP | 🚧 WIP |
| `models/`<br>`llama3` | TE accelerated Llama 3 | ✅ Active | ❌ | 🚧 WIP | ✅ | ✅ | ✅ | ✅ |
| `models/`<br>`geneformer` | TE accelerated single-cell BERT | 🚧 WIP | ❌ | ✅ | 🚧 WIP | 🚧 WIP | 🚧 WIP | 🚧 WIP |
| `recipes/`<br>`codonfm_ptl_te` | Recipe for [CodonFM](https://research.nvidia.com/labs/dbr/assets/data/manuscripts/nv-codonfm-preprint.pdf)'s Encodon using TE | ✅ Active | ❌ | 🚧 WIP | ✅ | ✅ | 🚧 WIP | 🚧 WIP |
| `recipes/`<br>`esm2_accelerate_te` | Recipe for ESM2 TE + HF Accelerate | ✅ Active | ❌ | 🚧 WIP | ✅ | ❌ | ✅ | 🚧 WIP |
| `recipes/`<br>`esm2_native_te` | Recipe for ESM2 TE + native PyTorch | ✅ Active | ❌ | ✅ | ✅ | ✅ | ✅ | 🚧 WIP |
| `recipes/`<br>`esm2_native_te` | Recipe for ESM2 TE + native PyTorch | ✅ Active | ❌ | ✅ | ✅ | ✅ | ✅ | ✅ |
| `recipes/`<br>`geneformer_native_te_mfsdp_fp8` | Recipe for Geneformer HF model | 🚧 WIP | ❌ | ✅ | ✅ | ❌ | ✅ | 🚧 WIP |
| `recipes/`<br>`llama3_native_te` | Recipe for Llama 3 TE + native PyTorch | ✅ Active | ❌ | 🚧 WIP | ✅ | ✅ | 🚧 WIP | 🚧 WIP |
| `recipes/`<br>`llama3_native_te` | Recipe for Llama 3 TE + native PyTorch | ✅ Active | ❌ | 🚧 WIP | ✅ | ✅ | ✅ | ✅ |
| `models/`<br>`mixtral` | TE accelerated MoE model | ✅ Active | ❌ | 🚧 WIP | ✅ | ✅ | ✅ | 🚧 WIP |
| `models/`<br>`qwen` | TE accelerated Qwen2.5/Qwen3 | ✅ Active | ❌ | 🚧 WIP | ✅ | ✅ | ✅ | 🚧 WIP |
| `recipes/`<br>`esm2_peft_te` | Recipe for ESM2 LoRA fine-tuning | ✅ Active | ❌ | ❌ | ✅ | ✅ | 🚧 WIP | ❌ |
| `recipes/`<br>`evo2_megatron` | Recipe for Evo2 via Megatron Bridge | 🚧 WIP | ❌ | ❌ | ✅ | ❌ | ✅ | ❌ |
| `recipes/`<br>`fp8_analysis` | FP8 training analyzer & heatmap tool | ✅ Active | N/A | N/A | N/A | N/A | N/A | N/A |
| `recipes/`<br>`vit` | Recipe for Vision Transformer | 🚧 WIP | ❌ | ✅ | ✅ | ❌ | ✅ | 🚧 WIP |

</small>
Expand Down Expand Up @@ -113,7 +135,7 @@ BioNeMo Framework is part of a larger ecosystem of NVIDIA Biopharma products. Ge

## Documentation Resources

- **Official Documentation:** Contents of `sub-packages` including user guides, API references, and troubleshooting, are documented on our [official documentation](https://docs.nvidia.com/bionemo-framework/latest/). Nightly builds of this documentation is available on [BioNeMo Framework GitHub Pages](https://nvidia.github.io/bionemo-framework/)
- **Official Documentation:** Documentation for sub-packages, including user guides, API references, and troubleshooting, is available on our [official documentation](https://docs.nvidia.com/bionemo-framework/latest/). Nightly builds of this documentation is available on [BioNeMo Framework GitHub Pages](https://nvidia.github.io/bionemo-framework/)

- **🚧 In-Progress Documentation 🚧:** `bionemo-recipes` documentation is currently work in progress, however the recipes are meant to be self-documented and easy to understand—we suggest you throw them into your favorite genai code assistant!

Expand All @@ -136,8 +158,8 @@ docker run --rm -it \

#### Initializing 3rd-party dependencies as git submodules

The NeMo and Megatron-LM dependencies are included as git submodules in bionemo2. The pinned commits for these submodules represent the "last-known-good" versions of these packages
that are confirmed to be working with bionemo2 (and those that are tested in CI).
The NeMo and Megatron-LM dependencies are included as git submodules in BioNeMo Framework. The pinned commits for these submodules represent the "last-known-good" versions of these packages
that are confirmed to be working with BioNeMo Framework (and those that are tested in CI).

To initialize these sub-modules when cloning the repo, add the `--recursive` flag to the git clone command:

Expand Down
2 changes: 0 additions & 2 deletions bionemo-recipes/models/amplify/README.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -117,5 +117,3 @@ Or, upload all models at once with:
```bash
for dir in *; do huggingface-cli upload nvidia/$(basename "$dir") "$dir/"; done
```

z
Loading