
Causal Interventions with pyvene
FreePerform causal interventions on PyTorch models effortlessly.
Free · Opens the source repo
What Causal Interventions with pyvene does
The pyvene skill provides a framework for conducting causal interventions on PyTorch models, leveraging the capabilities of the pyvene library developed by Stanford NLP. It allows users to perform various types of interventions, such as activation patching and causal tracing, which are essential for understanding and improving model behavior. By utilizing a declarative, dictionary-based configuration, users can easily set up and execute experiments that are reproducible and shareable, facilitating collaboration and validation of findings.
This skill is particularly useful for researchers and developers working with neural networks who need to test causal hypotheses or analyze the impact of specific components within their models. With support for any PyTorch architecture, not limited to transformers, pyvene opens up a range of possibilities for model interpretability and experimentation. Users can define intervention configurations that target specific layers and components, allowing for precise control over the model's behavior during experiments.
The skill includes detailed workflows for causal tracing and activation patching, providing step-by-step guidance on how to implement these techniques effectively. By following these workflows, users can identify causal relationships within their models and determine which components are critical for achieving desired behaviors. This capability is crucial for advancing understanding in areas such as model robustness and explainability, making pyvene a valuable tool in the AI research toolkit.
When to use it
Use pyvene when you need to conduct causal tracing, activation patching, or interchange intervention training on PyTorch models.
When not to use it
Consider alternatives if you require exploratory activation analysis or need lower-level control over model execution.
What you can build with it
Causal Tracing for Model Analysis
Utilize pyvene to identify where factual associations are stored in a model by corrupting inputs and restoring activations.
Activation Patching for Behavior Testing
Conduct experiments to determine which model components are essential for specific behaviors through activation patching.
Interchange Intervention Training
Explore the effects of different training strategies on model performance by using interchange intervention training capabilities.
How to install Causal Interventions with pyvene
View source1. Install with the skills CLI
npx skills add davila7/claude-code-templates/mechanistic-interpretability-pyvene --agent claude-code2. Or install it manually
Download the skill folder and drop it into ~/.claude/skills/ for all projects, or .claude/skills/ to scope it to one repo. Restart Claude Code so it picks up the new skill.
Anthropic's agentic coding CLI, and the reference implementation of Agent Skills. Drop a skill folder into ~/.claude/skills and Claude Code loads it automatically whenever a task matches the skill's description. Claude Code docs
Inside SKILL.md
Written by davila7pyvene: Causal Interventions for Neural Networks
pyvene is Stanford NLP's library for performing causal interventions on PyTorch models. It provides a declarative, dict-based framework for activation patching, causal tracing, and interchange intervention training - making intervention experiments reproducible and shareable.
GitHub: stanfordnlp/pyvene (840+ stars) Paper: pyvene: A Library for Understanding and Improving PyTorch Models via Interventions (NAACL 2024)
When to Use pyvene
Use pyvene when you need to:
- Perform causal tracing (ROME-style localization)
- Run activation patching experiments
- Conduct interchange intervention training (IIT)
- Test causal hypotheses about model components
- Share/reproduce intervention experiments via HuggingFace
- Work with any PyTorch architecture (not just transformers)
Consider alternatives when:
- You need exploratory activation analysis → Use TransformerLens
- You want to train/analyze SAEs → Use SAELens
- You need remote execution on massive models → Use nnsight
- You want lower-level control → Use nnsight
Installation
pip install pyvene
Standard import:
import pyvene as pv
Core Concepts
IntervenableModel
The main class that wraps any PyTorch model with intervention capabilities:
import pyvene as pv
from transformers import AutoModelForCausalLM, AutoTokenizer
# Load base model
model = AutoModelForCausalLM.from_pretrained("gpt2")
tokenizer = AutoTokenizer.from_pretrained("gpt2")
# Define intervention configuration
config = pv.IntervenableConfig(
representations=[
pv.RepresentationConfig(
layer=8,
component="block_output",
intervention_type=pv.VanillaIntervention,
)
]
)
# Create intervenable model
intervenable = pv.IntervenableModel(config, model)
Intervention Types
| Type | Description | Use Case |
|---|---|---|
VanillaIntervention | Swap activations between runs | Activation patching |
AdditionIntervention | Add activations to base run | Steering, ablation |
SubtractionIntervention | Subtract activations | Ablation |
ZeroIntervention | Zero out activations | Component knockout |
RotatedSpaceIntervention | DAS trainable intervention | Causal discovery |
CollectIntervention | Collect activations | Probing, analysis |
Component Targets
# Available components to intervene on
components = [
"block_input", # Input to transformer block
"block_output", # Output of transformer block
"mlp_input", # Input to MLP
"mlp_output", # Output of MLP
"mlp_activation", # MLP hidden activations
"attention_input", # Input to attention
"attention_output", # Output of attention
"attention_value_output", # Attention value vectors
"query_output", # Query vectors
"key_output", # Key vectors
"value_output", # Value vectors
"head_attention_value_output", # Per-head values
]
Workflow 1: Causal Tracing (ROME-style)
Locate where factual associations are stored by corrupting inputs and restoring activations.
Step-by-Step
import pyvene as pv
from transformers import AutoModelForCausalLM, AutoTokenizer
import torch
model = AutoModelForCausalLM.from_pretrained("gpt2-xl")
tokenizer = AutoTokenizer.from_pretrained("gpt2-xl")
# 1. Define clean and corrupted inputs
clean_prompt = "The Space Needle is in downtown"
corrupted_prompt = "The ##### ###### ## ## ########" # Noise
clean_tokens = tokenizer(clean_prompt, return_tensors="pt")
corrupted_tokens = tokenizer(corrupted_prompt, return_tensors="pt")
# 2. Get clean activations (source)
with torch.no_grad():
clean_outputs = model(**clean_tokens, output_hidden_states=True)
clean_states = clean_outputs.hidden_states
# 3. Define restoration intervention
def run_causal_trace(layer, position):
"""Restore clean activation at specific layer and position."""
config = pv.IntervenableConfig(
representations=[
pv.RepresentationConfig(
layer=layer,
component="block_output",
intervention_type=pv.VanillaIntervention,
unit="pos",
max_number_of_units=1,
)
]
)
intervenable = pv.IntervenableModel(config, model)
# Run with intervention
_, patched_outputs = intervenable(
base=corrupted_tokens,
sources=[clean_tokens],
unit_locations={"sources->base": ([[[position]]], [[[position]]])},
output_original_output=True,
)
# Return probability of correct token
probs = torch.softmax(patched_outputs.logits[0, -1], dim=-1)
seattle_token = tokenizer.encode(" Seattle")[0]
return probs[seattle_token].item()
# 4. Sweep over layers and positions
n_layers = model.config.n_layer
seq_len = clean_tokens["input_ids"].shape[1]
results = torch.zeros(n_layers, seq_len)
for layer in range(n_layers):
for pos in range(seq_len):
results[layer, pos] = run_causal_trace(layer, pos)
# 5. Visualize (layer x position heatmap)
# High values indicate causal importance
Checklist
- Prepare clean prompt with target factual association
- Create corrupted version (noise or counterfactual)
- Define intervention config for each (layer, position)
- Run patching sweep
- Identify causal hotspots in heatmap
Workflow 2: Activation Patching for Circuit Analysis
Test which components are necessary for a specific behavior.
Step-by-Step
import pyvene as pv
from transformers import AutoModelForCausalLM, AutoTokenizer
import torch
model = AutoModelForCausalLM.from_pretrained("gpt2")
tokenizer = AutoTokenizer.from_pretrained("gpt2")
# IOI task setup
clean_prompt = "When John and Mary went to the store, Mary gave a bottle to"
corrupted_prompt = "When John and Mary went to the store, John gave a bottle to"
clean_tokens = tokenizer(clean_prompt, return_tensors="pt")
corrupted_tokens = tokenizer(corrupted_prompt, return_tensors="pt")
john_token = tokenizer.encode(" John")[0]
mary_token = tokenizer.encode(" Mary")[0]
def logit_diff(logits):
"""IO - S logit difference."""
return logits[0, -1, john_token] - logits[0, -1, mary_token]
# Patch attention output at each layer
def patch_attention(layer):
config = pv.IntervenableConfig(
representations=[
pv.RepresentationConfig(
layer=layer,
component="attention_output",
intervention_type=pv.VanillaIntervention,
)
]
)
intervenable = pv.IntervenableModel(config, model)
_, patched_outputs = intervenable(
base=corrupted_tokens,
sources=[clean_tokens],
)
return logit_diff(patched_outputs.logits).item()
# Find which layers matter
results = []
for layer in range(model.config.n_layer):
diff = patch_attention(layer)
results.append(diff)
print(f"Layer {layer}: logit diff = {diff:.3f}")
Workflow 3: Interchange Intervention Training (IIT)
Train interventions to discover causal structure.
Step-by-Step
import pyvene as pv
from transformers import AutoModelForCausalLM
import torch
model = AutoModelForCausalLM.from_pretrained("gpt2")
# 1. Define trainable intervention
config = pv.IntervenableConfig(
representations=[
pv.RepresentationConfig(
layer=6,
component="block_output",
intervention_type=pv.RotatedSpaceIntervention, # Trainable
low_rank_dimension=64, # Learn 64-dim subspace
)
]
)
intervenable = pv.IntervenableModel(config, model)
# 2. Set up training
optimizer = torch.optim.Adam(
intervenable.get_trainable_parameters(),
lr=1e-4
)
# 3. Training loop (simplified)
for base_input, source_input, target_output in dataloader:
optimizer.zero_grad()
_, outputs = intervenable(
base=base_input,
sources=[source_input],
)
loss = criterion(outputs.logits, target_output)
loss.backward()
optimizer.step()
# 4. Analyze learned intervention
# The rotation matrix reveals causal subspace
rotation = intervenable.interventions["layer.6.block_output"][0].rotate_layer
DAS (Distributed Alignment Search)
# Low-rank rotation finds interpretable subspaces
config = pv.IntervenableConfig(
representations=[
pv.RepresentationConfig(
layer=8,
component="block_output",
intervention_type=pv.LowRankRotatedSpaceIntervention,
low_rank_dimension=1, # Find 1D causal direction
)
]
)
Workflow 4: Model Steering (Honest LLaMA)
Steer model behavior during generation.
import pyvene as pv
from transformers import AutoModelForCausalLM, AutoTokenizer
model = AutoModelForCausalLM.from_pretrained("meta-llama/Llama-2-7b-hf")
tokenizer = AutoTokenizer.from_pretrained("meta-llama/Llama-2-7b-hf")
# Load pre-trained steering intervention
intervenable = pv.IntervenableModel.load(
"zhengxuanzenwu/intervenable_honest_llama2_chat_7B",
model=model,
)
# Generate with steering
prompt = "Is the earth flat?"
inputs = tokenizer(prompt, return_tensors="pt")
# Intervention applied during generation
outputs = intervenable.generate(
inputs,
max_new_tokens=100,
do_sample=False,
)
print(tokenizer.decode(outputs[0]))
Saving and Sharing Interventions
# Save locally
intervenable.save("./my_intervention")
# Load from local
intervenable = pv.IntervenableModel.load(
"./my_intervention",
model=model,
)
# Share on HuggingFace
intervenable.save_intervention("username/my-intervention")
# Load from HuggingFace
intervenable = pv.IntervenableModel.load(
"username/my-intervention",
model=model,
)
Common Issues & Solutions
Issue: Wrong intervention location
# WRONG: Incorrect component name
config = pv.RepresentationConfig(
component="mlp", # Not valid!
)
# RIGHT: Use exact component name
config = pv.RepresentationConfig(
component="mlp_output", # Valid
)
Issue: Dimension mismatch
# Ensure source and base have compatible shapes
# For position-specific interventions:
config = pv.RepresentationConfig(
unit="pos",
max_number_of_units=1, # Intervene on single position
)
# Specify locations explicitly
intervenable(
base=base_tokens,
sources=[source_tokens],
unit_locations={"sources->base": ([[[5]]], [[[5]]])}, # Position 5
)
Issue: Memory with large models
# Use gradient checkpointing
model.gradient_checkpointing_enable()
# Or intervene on fewer components
config = pv.IntervenableConfig(
representations=[
pv.RepresentationConfig(
layer=8, # Single layer instead of all
component="block_output",
)
]
)
Issue: LoRA integration
# pyvene v0.1.8+ supports LoRAs as interventions
config = pv.RepresentationConfig(
intervention_type=pv.LoRAIntervention,
low_rank_dimension=16,
)
Key Classes Reference
| Class | Purpose |
|---|---|
IntervenableModel | Main wrapper for interventions |
IntervenableConfig | Configuration container |
RepresentationConfig | Single intervention specification |
VanillaIntervention | Activation swapping |
RotatedSpaceIntervention | Trainable DAS intervention |
CollectIntervention | Activation collection |
Supported Models
pyvene works with any PyTorch model. Tested on:
- GPT-2 (all sizes)
- LLaMA / LLaMA-2
- Pythia
- Mistral / Mixtral
- OPT
- BLIP (vision-language)
- ESM (protein models)
- Mamba (state space)
Reference Documentation
For detailed API documentation, tutorials, and advanced usage, see the references/ folder:
| File | Contents |
|---|---|
| references/README.md | Overview and quick start guide |
| references/api.md | Complete API reference for IntervenableModel, intervention types, configurations |
| references/tutorials.md | Step-by-step tutorials for causal tracing, activation patching, DAS |
External Resources
Tutorials
Papers
- Locating and Editing Factual Associations in GPT - Meng et al. (2022)
- Inference-Time Intervention - Li et al. (2023)
- Interpretability in the Wild - Wang et al. (2022)
Official Documentation
Comparison with Other Tools
| Feature | pyvene | TransformerLens | nnsight |
|---|---|---|---|
| Declarative config | Yes | No | No |
| HuggingFace sharing | Yes | No | No |
| Trainable interventions | Yes | Limited | Yes |
| Any PyTorch model | Yes | Transformers only | Yes |
| Remote execution | No | No | Yes (NDIF) |
Frequently asked questions about Causal Interventions with pyvene
Similar skills
Spring Boot Testing
Master testing techniques for Spring Boot 4 applications.
GitHub Issues
Manage GitHub issues efficiently with MCP tools.
Geofeed Tuner
Optimize your IP geolocation feeds in CSV format.
Batch Files
Master Windows batch scripting for automation and task management.
Adobe Illustrator Scripting
Automate your Illustrator workflows with ExtendScript.
Plugin Structure
Create and organize Claude Code plugins effectively.
