stanfordnlp/pyvene
Stanford NLP Python Library for Understanding and Improving PyTorch Models via Interventions
PythonApache-2.0
Issues
- 0
[Feature Request / Suggestion]: Is it possible to extend this to text embeddings?
#181 opened by fblissjr - 0
- 0
[Suggestion]: Support Causal Tracing for LLaMA model
#174 opened by aryopg - 2
[P1] Supporting Mistral with Unit Tests (#46)
#62 opened by frankaging - 0
- 0
- 1
[P2] Having a simple tutorial on how to use intervenable model inside another torch module, or pipeline (e.g., HF pipeline)
#45 opened by frankaging - 0
Make docs website
#168 opened by aryamanarora - 3
[P2] Sparse autoencoders
#77 opened by aryamanarora - 0
[Bug]: RoBERTa giving error with Pyvene 0.1.2. for ReFT. "forward() got an unexpected keyword argument 'use_cache'".
#163 opened by m-dev12 - 0
[External]: Attention head intervention doesn't work for models other than GPT-2
#158 opened by Bakser - 0
[Suggestion]: Docs: explain the first example
#162 opened by ocramz - 0
- 2
- 5
- 0
[P1] Tuned lens support
#148 opened by aryamanarora - 1
- 1
[P2] Mac chip MPS mode support
#67 opened by frankaging - 3
- 0
[P0] Adding back GPT2 and other model supports
#138 opened by frankaging - 2
- 0
[P0] Make interface compatible with HF trainer
#121 opened by aryamanarora - 0
- 0
[P1] Dynamic Intervention Scheduler
#111 opened by frankaging - 0
- 0
- 1
- 1
[P1] Tutorial of Inference-time Intervention
#68 opened by frankaging - 1
- 1
[P0] Upgrade tutorials to new API
#86 opened by aryamanarora - 0
- 1
Bug in BoundlessRotatedSpaceIntervention
#73 opened by jiudingsun01 - 0
[P1] Supports simple use case of intervention (fixed location intervention for all examples)
#52 opened by frankaging - 0
[Bug Fix] `MLPConfig` contains conflict field names with HF that cannot be set to other values when initializing
#51 opened by frankaging - 0
[P2] Quantized model support
#53 opened by frankaging - 0
[P2] Support interchange intervention training by unfreezing model weights with vanilla intervention
#24 opened by frankaging - 2
[P2] Probe training with interventions
#26 opened by frankaging - 0
- 0
[P0] Library Class Renaming Effort
#41 opened by frankaging - 0
- 0
[P1] Unit tests for main class objects
#38 opened by frankaging - 1
[External] Save Interventions
#37 opened by jiudingsun01 - 1
- 2
[P0] enable `use_fast` option in the alignable to hyper boost training speed in case intervention locations (for position+subspace) are fixed in a batch
#33 opened by frankaging - 1
[P1] Add unit test for vanilla intervention at different positions and streams
#32 opened by frankaging - 0
- 0
- 0
[P2] Refactor utils in models to smaller files
#23 opened by frankaging - 4
- 0
[P1] Introduce grouped interventions where each group is associated with the same source input
#22 opened by frankaging