-
https://github.com/JasonGross/guarantees-based-mechanistic-interpretability/blob/d0cf2510be626fa9ebc022009d28f715646762cf/notebooks_jason/max_of_2_grokking.py#L126-L128
re-downloads the artifacts fro…
-
When running the same `flake8` commands that are specified for the "Lint with flake8" task in `.github\workflows\test-models.yml` I get the following error, on both my PR branch and main branch:
``…
-
### Describe the bug
I see
```
wandb: WARNING A graphql request initiated by the public wandb API timed out (timeout=19 sec). Create a new API with an
integer timeout larger than 19, e.g., `api = …
-
This is great work! I was running it and noticed that the testing accuracy computation code
```py
logits = model(all_pairs)[:, -1]
model_labels = np.zeros((p, p))
for x in range(p):
for y in …
-
### Question
My gpu can't download checkpoints directly from huggingface.co, I write a function to load models from local directory, then passing to HookedTransformer.
![image](https://github.com/…
-
**Describe the bug**
Pythia model outputs don't exactly match the Huggingface Transformers implementation.
**Code example**
```
def check_similarity_with_hf_model(tl_model, hf_model, atol, promp…
-
Hi,
Thanks for contributing to this great list by compiling so many resources! I just want to (shamelessly) self-promote some of my own contributions to the interpretability field:
1. Add a pap…
-
**Describe the bug**
In the repo README, there's a line "Feel free to join the [Open Source Mechanistic Interpretability Slack](https://join.slack.com/t/opensourcemechanistic/shared_invite/zt-1qosyh8…
-
Can you talk a little bit about the pros and cons of each and in which kinds of situations/applications one would be more useful than the other? Like why pick one over the other if you had to pick?
-
### Question
I am attempting to run a model in an offline environment using the following code:
```python
import os
import transformers
from transformer_lens import HookedTransformer
base_…