esm

Name: esm
Author: K-Dense-AI

31kstars3.1kforks

How to install esm?

k-dense-ai/scientific-agent-skills/esm

$npx -y skills add K-Dense-AI/scientific-agent-skills --skill esm

Installs into the current project.

Summary

Use when working directly with the esm Python SDK, ESM3 or ESMC model IDs, Forge/Biohub inference clients, or ESMFold2 folding workflows.

Files · 1

View on GitHub

SKILL.md

1	# ESM: Evolutionary Scale Modeling
2
3	## Overview
4
5	ESM provides protein language models for understanding, generating, and designing proteins. Use this skill for current EvolutionaryScale/Biohub workflows: ESM3 for generative design, ESMC for representation learning and embeddings, hosted Forge/Biohub inference, and ESMFold2 all-atom structure prediction.
6
7	## Core Capabilities
8
9	### 1. Protein Sequence Generation with ESM3
10
11	Generate novel protein sequences with desired properties using multimodal generative modeling.
12
13	When to use:
14	- Designing proteins with specific functional properties
15	- Completing partial protein sequences
16	- Generating variants of existing proteins
17	- Creating proteins with desired structural characteristics
18
19	Basic usage:
20
21	```python
22	from esm.models.esm3 import ESM3
23	from esm.sdk.api import ESM3InferenceClient, ESMProtein, GenerationConfig
24
25	# Load local open weights after accepting the license on Hugging Face.
26	model: ESM3InferenceClient = ESM3.from_pretrained("esm3-open").to("cuda")
27
28	# Create protein prompt
29	protein = ESMProtein(sequence="MPRT___KEND") # '_' represents masked positions
30
31	# Generate completion
32	protein = model.generate(protein, GenerationConfig(track="sequence", num_steps=8))
33	print(protein.sequence)
34	```
35
36	For remote/cloud usage via Forge API:
37
38	```python
39	import os
40	import esm
41	from esm.sdk.api import ESMProtein, GenerationConfig
42
43	# Same interface as local ESM3; token from ESM_API_KEY (see Authentication)
44	model = esm.sdk.client("esm3-medium-2024-08", token=os.environ["ESM_API_KEY"])
45
46	# Generate
47	protein = model.generate(protein, GenerationConfig(track="sequence", num_steps=8))
48	```
49
50	See `references/esm3-api.md` for detailed ESM3 model specifications, advanced generation configurations, and multimodal prompting examples.
51
52	### 2. Structure Prediction and Inverse Folding
53
54	Use ESM3's structure track for structure prediction from sequence or inverse folding (sequence design from structure).
55
56	Structure prediction:
57
58	```python
59	from esm.sdk.api import ESM3InferenceClient, ESMProtein, GenerationConfig
60
61	# Predict structure from sequence
62	protein = ESMProtein(sequence="MPRTKEINDAGLIVHSP...")
63	protein_with_structure = model.generate(
64	protein,
65	GenerationConfig(track="structure", num_steps=protein.sequence.count("_"))
66	)
67
68	# Access predicted structure
69	coordinates = protein_with_structure.coordinates # 3D coordinates
70	pdb_string = protein_with_structure.to_pdb()
71	```
72
73	Inverse folding (sequence from structure):
74
75	```python
76	# Design sequence for a target structure
77	protein_with_structure = ESMProtein.from_pdb("target_structure.pdb")
78	protein_with_structure.sequence = None # Remove sequence
79
80	# Generate sequence that folds to this structure
81	designed_protein = model.generate(
82	protein_with_structure,
83	GenerationConfig(track="sequence", num_steps=50, temperature=0.7)
84	)
85	```
86
87	### 3. Protein Embeddings with ESM C
88
89	Generate high-quality embeddings for downstream tasks like function prediction, classification, or similarity analysis.
90
91	When to use:
92	- Extracting protein representations for machine learning
93	- Computing sequence similarities
94	- Feature extraction for protein classification
95	- Transfer learning for protein-related tasks
96
97	Basic usage:
98
99	```python
100	from esm.models.esmc import ESMC
101	from esm.sdk.api import ESMProtein, LogitsConfig
102
103	# Load ESM C model
104	model = ESMC.from_pretrained("esmc_300m").to("cuda")
105
106	# Get embeddings
107	protein = ESMProtein(sequence="MPRTKEINDAGLIVHSP...")
108	protein_tensor = model.encode(protein)
109	logits_output = model.logits(
110	protein_tensor,
111	LogitsConfig(sequence=True, return_embeddings=True),
112	)
113	embeddings = logits_output.embeddings
114	```
115
116	Batch processing:
117
118	```python
119	# Encode multiple proteins
120	proteins = [
121	ESMProtein(sequence="MPRTKEIND..."),
122	ESMProtein(sequence="AGLIVHSPQ..."),
123	ESMProtein(sequence="KTEFLNDGR...")
124	]
125
126	embeddings_list = [
127	model.logits(
128	model.encode(p),
129	LogitsConfig(sequence=True, return_embeddings=True),
130	).embeddings
131	for p in proteins
132	]
133	```
134
135	See `references/esm-c-api.md` for ESM C model details, efficiency comparisons, and advanced embedding strategies.
136
137	### 4. Function Conditioning and Annotation
138
139	Use ESM3's function track to generate proteins with specific functional annotations or predict function from sequence.
140
141	Function-conditioned generation:
142
143	```python
144	from esm.sdk.api import ESMProtein, FunctionAnnotation, GenerationConfig
145
146	# Create protein with desired function
147	protein = ESMProtein(
148	sequence="_" * 200, # Generate 200 residue protein
149	function_annotations=[
150	FunctionAnnotation(label="fluorescent_protein", start=50, end=150)
151	]
152	)
153
154	# Generate sequence with specified function
155	functional_protein = model.generate(
156	prot

People who liked this also viewed

6 picks

Type

find-skillsHelps users discover and install agent skills when they ask questions like "how do I do X", "find a skill for X", "is there a skill that can...", or express interest in extending capabilities. This skSkillsJul 20262.6M27k
grill-meA relentless interview to sharpen a plan or design.SkillsJul 2026633k182k
vercel-react-best-practicesVercel's official collection of agent skillsSkillsJul 2026573k29k
improve-codebase-architectureScan a codebase for deepening opportunities, present them as a visual HTML report, then grill through whichever one you pick.SkillsJul 2026522k182k
setup-matt-pocock-skillsConfigure this repo for the engineering skills — set up its issue tracker, triage label vocabulary, and domain doc layout. Run once before first use of the other engineering skills.SkillsJul 2026454k182k
triageMove issues and external PRs through a state machine of triage roles — categorise, verify, grill if needed, and write agent-ready briefs.SkillsJul 2026424k182k

esm

31kstars3.1kforks

View on GitHub

How to install esm?

k-dense-ai/scientific-agent-skills/esm

$npx -y skills add K-Dense-AI/scientific-agent-skills --skill esm

Installs into the current project.

Summary

Use when working directly with the esm Python SDK, ESM3 or ESMC model IDs, Forge/Biohub inference clients, or ESMFold2 folding workflows.

Files · 1

View on GitHub

SKILL.md

1	# ESM: Evolutionary Scale Modeling
2
3	## Overview
4
5	ESM provides protein language models for understanding, generating, and designing proteins. Use this skill for current EvolutionaryScale/Biohub workflows: ESM3 for generative design, ESMC for representation learning and embeddings, hosted Forge/Biohub inference, and ESMFold2 all-atom structure prediction.
6
7	## Core Capabilities
8
9	### 1. Protein Sequence Generation with ESM3
10
11	Generate novel protein sequences with desired properties using multimodal generative modeling.
12
13	When to use:
14	- Designing proteins with specific functional properties
15	- Completing partial protein sequences
16	- Generating variants of existing proteins
17	- Creating proteins with desired structural characteristics
18
19	Basic usage:
20
21	```python
22	from esm.models.esm3 import ESM3
23	from esm.sdk.api import ESM3InferenceClient, ESMProtein, GenerationConfig
24
25	# Load local open weights after accepting the license on Hugging Face.
26	model: ESM3InferenceClient = ESM3.from_pretrained("esm3-open").to("cuda")
27
28	# Create protein prompt
29	protein = ESMProtein(sequence="MPRT___KEND") # '_' represents masked positions
30
31	# Generate completion
32	protein = model.generate(protein, GenerationConfig(track="sequence", num_steps=8))
33	print(protein.sequence)
34	```
35
36	For remote/cloud usage via Forge API:
37
38	```python
39	import os
40	import esm
41	from esm.sdk.api import ESMProtein, GenerationConfig
42
43	# Same interface as local ESM3; token from ESM_API_KEY (see Authentication)
44	model = esm.sdk.client("esm3-medium-2024-08", token=os.environ["ESM_API_KEY"])
45
46	# Generate
47	protein = model.generate(protein, GenerationConfig(track="sequence", num_steps=8))
48	```
49
50	See `references/esm3-api.md` for detailed ESM3 model specifications, advanced generation configurations, and multimodal prompting examples.
51
52	### 2. Structure Prediction and Inverse Folding
53
54	Use ESM3's structure track for structure prediction from sequence or inverse folding (sequence design from structure).
55
56	Structure prediction:
57
58	```python
59	from esm.sdk.api import ESM3InferenceClient, ESMProtein, GenerationConfig
60
61	# Predict structure from sequence
62	protein = ESMProtein(sequence="MPRTKEINDAGLIVHSP...")
63	protein_with_structure = model.generate(
64	protein,
65	GenerationConfig(track="structure", num_steps=protein.sequence.count("_"))
66	)
67
68	# Access predicted structure
69	coordinates = protein_with_structure.coordinates # 3D coordinates
70	pdb_string = protein_with_structure.to_pdb()
71	```
72
73	Inverse folding (sequence from structure):
74
75	```python
76	# Design sequence for a target structure
77	protein_with_structure = ESMProtein.from_pdb("target_structure.pdb")
78	protein_with_structure.sequence = None # Remove sequence
79
80	# Generate sequence that folds to this structure
81	designed_protein = model.generate(
82	protein_with_structure,
83	GenerationConfig(track="sequence", num_steps=50, temperature=0.7)
84	)
85	```
86
87	### 3. Protein Embeddings with ESM C
88
89	Generate high-quality embeddings for downstream tasks like function prediction, classification, or similarity analysis.
90
91	When to use:
92	- Extracting protein representations for machine learning
93	- Computing sequence similarities
94	- Feature extraction for protein classification
95	- Transfer learning for protein-related tasks
96
97	Basic usage:
98
99	```python
100	from esm.models.esmc import ESMC
101	from esm.sdk.api import ESMProtein, LogitsConfig
102
103	# Load ESM C model
104	model = ESMC.from_pretrained("esmc_300m").to("cuda")
105
106	# Get embeddings
107	protein = ESMProtein(sequence="MPRTKEINDAGLIVHSP...")
108	protein_tensor = model.encode(protein)
109	logits_output = model.logits(
110	protein_tensor,
111	LogitsConfig(sequence=True, return_embeddings=True),
112	)
113	embeddings = logits_output.embeddings
114	```
115
116	Batch processing:
117
118	```python
119	# Encode multiple proteins
120	proteins = [
121	ESMProtein(sequence="MPRTKEIND..."),
122	ESMProtein(sequence="AGLIVHSPQ..."),
123	ESMProtein(sequence="KTEFLNDGR...")
124	]
125
126	embeddings_list = [
127	model.logits(
128	model.encode(p),
129	LogitsConfig(sequence=True, return_embeddings=True),
130	).embeddings
131	for p in proteins
132	]
133	```
134
135	See `references/esm-c-api.md` for ESM C model details, efficiency comparisons, and advanced embedding strategies.
136
137	### 4. Function Conditioning and Annotation
138
139	Use ESM3's function track to generate proteins with specific functional annotations or predict function from sequence.
140
141	Function-conditioned generation:
142
143	```python
144	from esm.sdk.api import ESMProtein, FunctionAnnotation, GenerationConfig
145
146	# Create protein with desired function
147	protein = ESMProtein(
148	sequence="_" * 200, # Generate 200 residue protein
149	function_annotations=[
150	FunctionAnnotation(label="fluorescent_protein", start=50, end=150)
151	]
152	)
153
154	# Generate sequence with specified function
155	functional_protein = model.generate(
156	prot

People who liked this also viewed

6 picks

Type