aipoch/open-science / resources/skills/esmfold2
esmfold2
VerifiedBiohub ESMFold2 / ESMFold2-Fast all-atom co-folding (Candido et al. 2026, github.com/Biohub/esm). Single-sequence and MSA modes; protein, DNA, RNA, ligand (CCD/SMILES), modified residues. FoldBench Ab-Ag 50-55%, PPI 70-77% DockQ-pass. Also covers the ESMC-{300M,600M,6B} protein language models from the same release: masked-LM logits, hidden states, mutation scoring, contact prediction, and the SAE interpretability head. MIT-licensed weights on HuggingFace org `biohub`. Use this skill when: (1) Predicting complex structures with single-sequence input, (2) Validating designed binders with ESMFold2-Fast, (3) Running ESMFold2 with MSA input, (4) Getting ESMC embeddings or per-residue mutation scores, (5) Choosing kernel backend and sampling-step settings for paper-faithful throughput.
| File | Kind | Size |
|---|---|---|
| .catalog_stamp | other | 13 B |
| SKILL.md | skill md | 10.7 KB |
| references/design-hook.md | markdown | 1.4 KB |
| references/esmc.md | markdown | 3.0 KB |
references/design-hook.md
View on GitHub# ESMFold2-Experimental* — design hook
The four `biohub/ESMFold2-Experimental*` repos load as `ESMFold2ExperimentalModel`
(separate class, `modeling_esmfold2_experimental.py`). Unlike `ESMFold2Model`,
its `forward()` is **not** `@torch.inference_mode()`-decorated and exposes a
`res_type_soft` kwarg ("Used in design to provide a soft sequence input"):
```python
out = model(**features, res_type_soft=soft_oh, # [1,L,NUM_RES_TYPES], requires_grad
num_loops=1, num_sampling_steps=1, num_diffusion_samples=1)
D = out['distogram_logits'] # carries grad; trunk runs under
# torch.set_grad_enabled(res_type_soft is not None)
```
Diffusion + confidence-head run under `torch.no_grad()`, so only `distogram_logits`
carries the graph. **Design (Alg 11) must use an Experimental variant** — the
release `ESMFold2`/`ESMFold2-Fast` `forward()` is `@torch.inference_mode()` and
cannot be back-propagated through.
## Gotcha: `set_kernel_backend("fused")` on Experimental models
**Do not call `set_kernel_backend` on Experimental models.** The confidence head
runs outside the bf16 autocast block, so its fused matmul sees fp32 inputs and throws:
```
RuntimeError: self and mat2 must have the same dtype, but got BFloat16 and Float
```
The reference path is correct (the paper's design loop never calls it). The fused
backend is only validated for release `ESMFold2Model` / `ESMFold2-Fast` via
`builder.fold()`.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| brainstormingobra/superpowersYou MUST use this before any creative work - creating features, building components, adding functionality, modifying behavior, or planning anything new, in software or out of it (a talk, a business, a renovation). | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| brand-guidelinesanthropics/skillsApplies Anthropic's official brand colors and typography to any sort of artifact that may benefit from having Anthropic's look-and-feel. Use it when brand colors or style guidelines, visual formatting, or company design standards apply. | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |
| frontend-designanthropics/skillsGuidance for distinctive, intentional visual design when building new UI or reshaping an existing one. Helps with aesthetic direction, typography, and making choices that don't read as templated defaults. | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |
| mcp-builderanthropics/skillsGuide for creating high-quality MCP (Model Context Protocol) servers that enable LLMs to interact with external services through well-designed tools. Use when building MCP servers to integrate external APIs or services, whether in Python (FastMCP) or Node/TypeScript (MCP SDK). | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |