Neuronpedia Python Library
Authentication
Some APIs on Neuronpedia require an API key. For example, if you want to bookmark something in your account, or upload a new vector, you'll need to identify yourself with a Neuronpedia API key.
Setting the API Key
- Sign up for free at
neuronpedia.org. - Get your Neuronpedia API key from
neuronpedia.org/account. - Set the environment variable
NEURONPEDIA_API_KEYto your API key. You can do this through a.envfile or other similar methods.
Example: Upload a Vector, then Steer With It
from neuronpedia.sample_data import GEMMA2_2B_IT_DINOSAURS_VECTOR
from neuronpedia.np_vector import NPVector
import os, json
# from neuronpedia.org/account
os.environ["NEURONPEDIA_API_KEY"] = "YOUR_NP_API_KEY"
# upload the custom vector
np_vector = NPVector.new(
label="dinosaurs",
model_id="gemma-2-2b-it",
layer_num=20,
hook_type="hook_resid_pre",
vector=GEMMA2_2B_IT_DINOSAURS_VECTOR,
default_steer_strength=44,
)
# steer with it
responseJson = np_vector.steer_chat(
steered_chat_messages=[{"role": "user", "content": "Write a one sentence story."}]
)
print(json.dumps(responseJson, indent=2))
print("UI Steering at: " + responseJson["shareUrl"])
The output of the above will be similar to:
{
"STEERED": {
"chat_template": [
{
"content": "Write a one sentence story.",
"role": "user"
},
{
"content": "The last dinosaur roared, its breath a smoke-filled mirror of the dying sun.",
"role": "model"
}
],
"raw": "<bos><start_of_turn>user\nWrite a one sentence story.<end_of_turn>\n<start_of_turn>model\nThe last dinosaur roared, its breath a smoke-filled mirror of the dying sun. \n<end_of_turn><eos>"
},
[...]
}
See the examples folder for detailed notebooks and usage.
Important: Uniquely Identifying a Vector/Feature
On Neuronpedia, a vector or feature has a unique identifier comprised of three parts:
- Model ID: The ID of the model that this vector belongs to. For example,
gpt2-small. - Source: A "group name" that starts with the layer number, and usually contains some other identifying information depending on what it is, like number of features in the group. Some examples of source:
6-res-jb:6= Layer 6,res= residual stream,jb= Joseph Bloom3-gemmascope-att-16k:3= Layer 3,gemmascope= Gemma Scope,att= Attention,16k= 16k-width SAE
- Index: The index in the model + source. This is a string, and currently the string is always an integer. However, this may change in the future.
Example:
gemma-2-2b/3-gemmascope-att-16k/4232- Model: gemma-2-2b
- Source: 3-gemmascope-att-16k
- Index: 4232
To make API calls that involve an existing vector or feature on Neuronpedia, you pass the model, source, and index as required.
Example API Call: Get SAE Feature (including activations, explanations, etc)
from neuronpedia.np_sae_feature import SAEFeature
sae_feature = SAEFeature.get("gemma-2-2b", "3-gemmascope-att-16k", "4232")
print(sae_feature)
1 maintainer
Metadata
1 maintainer
Release files for neuronpedia 1.2.1
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | Provenance | |
|---|---|---|---|---|
| neuronpedia-1.2.1.tar.gz | 45.7 kB | Without provenance | Details |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Reset |
|---|---|---|---|---|
| neuronpedia-1.2.1-py3-none-any.whl | Python 3 | none | any | Details |
Total release size: 111.3 kB
Release files / neuronpedia-1.2.1.tar.gz
| Download URL | neuronpedia-1.2.1.tar.gz |
|---|---|
| Size | 45.7 kB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
33941faea683d0d0b1a4b910c33248200aed4a711da7e4c074b8567899e2cb29
|
|
BLAKE2b-256 checksum How to use checksums |
40b1907e117d0b04eabf24f0a405e179a923dcf63e4525aaa2bfd867c42abd7a
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
uv/0.11.7 {"installer":{"name":"uv","version":"0.11.7","subcommand":["publish"]},"python":null,"implementation":{"name":null,"version":null},"distro":{"name":"Ubuntu","version":"26.04","id":"resolute","libc":null},"system":{"name":null,"release":null},"cpu":null,"openssl_version":null,"setuptools_version":null,"rustc_version":null,"ci":null}
|
Release files / neuronpedia-1.2.1-py3-none-any.whl
| Download URL | neuronpedia-1.2.1-py3-none-any.whl |
|---|---|
| Size | 65.7 kB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
c439a0a08c22b61a7b8fe3638191ebae9113c39b7cdbf7a37fccc4589c106af6
|
|
BLAKE2b-256 checksum How to use checksums |
fd235a4f0cbe3d718db9b2d41cd0eca5a6849b5f72346396c0695abb87ec2169
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
uv/0.11.7 {"installer":{"name":"uv","version":"0.11.7","subcommand":["publish"]},"python":null,"implementation":{"name":null,"version":null},"distro":{"name":"Ubuntu","version":"26.04","id":"resolute","libc":null},"system":{"name":null,"release":null},"cpu":null,"openssl_version":null,"setuptools_version":null,"rustc_version":null,"ci":null}
|
Security for neuronpedia 1.2.1
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
No provenance found
None of the files in this release were uploaded with provenance metadata.
A note on package security
PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.
