added new block for recent diagram, and colab compatibility notebook (#…

…758)

* added new block for recent diagram, and colab compatability notebook

* updated check

* ran format

* added better checking for prop

* fixed format

docs/source/index.md

@@ -18,6 +18,8 @@ I used to work for the [Anthropic interpretability team](https://transformer-cir
 
 The core features were heavily inspired by the interface to [Anthropic's excellent Garcon tool](https://transformer-circuits.pub/2021/garcon/index.html). Credit to Nelson Elhage and Chris Olah for building Garcon and showing me the value of good infrastructure for enabling exploratory research!
 
+A great place to start is to take a look at a helpful diagram of [all weight matrices and activation tensors with TransformerLens notation](_static/TransformerLens_Diagram.svg) courtesy of [Austin Kozlowski](https://github.com/akozlo). Another helpful tool to help you get going as quickly as possible is our [Colab Compatability Demo](https://github.com/TransformerLensOrg/TransformerLens/tree/main/demos/Colab_Compatibility.ipynb), which will give you a good idea of what you can do in various Colab environments.
+
 ```{toctree}
 :hidden:
 :caption: Introduction

transformer_lens/loading_from_pretrained.py

@@ -1075,7 +1075,9 @@ def convert_hf_model_config(model_name: str, **kwargs):
         use_local_attn = True if hf_config.sliding_window else False
         cfg_dict = {
             "d_model": hf_config.hidden_size,
-            "d_head": hf_config.head_dim or hf_config.hidden_size // hf_config.num_attention_heads,
+            "d_head": hf_config.head_dim
+            if hasattr(hf_config, "head_dim") and hf_config.head_dim > 0
+            else hf_config.hidden_size // hf_config.num_attention_heads,
             "n_heads": hf_config.num_attention_heads,
             "d_mlp": hf_config.intermediate_size,
             "n_layers": hf_config.num_hidden_layers,