Feature llama 33 (#826)

* added llama 3.3 config

* fixed key

* added debug point

* updated model compatibility notebook

* ran format

* removed log point

Author: bryce13950

demos/Colab_Compatibility.ipynb

@@ -58,14 +58,14 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 2,
+   "execution_count": 1,
    "metadata": {},
    "outputs": [
     {
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "TransformerLens currently supports 205 models out of the box.\n"
+      "TransformerLens currently supports 206 models out of the box.\n"
      ]
     }
    ],
@@ -429,6 +429,7 @@
     "    \"meta-llama/Llama-2-70b-chat-hf\",\n",
     "    \"meta-llama/Llama-3.1-70B\",\n",
     "    \"meta-llama/Llama-3.1-70B-Instruct\",\n",
+    "    \"meta-llama/Llama-3.3-70B-Instruct\",\n",
     "    \"meta-llama/Meta-Llama-3-70B\",\n",
     "    \"meta-llama/Meta-Llama-3-70B-Instruct\",\n",
     "    \"mistralai/Mixtral-8x7B-Instruct-v0.1\",\n",

transformer_lens/loading_from_pretrained.py

@@ -151,14 +151,15 @@
     "meta-llama/Meta-Llama-3-8B-Instruct",
     "meta-llama/Meta-Llama-3-70B",
     "meta-llama/Meta-Llama-3-70B-Instruct",
-    "meta-llama/Llama-3.2-1B",
-    "meta-llama/Llama-3.2-3B",
-    "meta-llama/Llama-3.2-1B-Instruct",
-    "meta-llama/Llama-3.2-3B-Instruct",
     "meta-llama/Llama-3.1-70B",
     "meta-llama/Llama-3.1-8B",
     "meta-llama/Llama-3.1-8B-Instruct",
     "meta-llama/Llama-3.1-70B-Instruct",
+    "meta-llama/Llama-3.2-1B",
+    "meta-llama/Llama-3.2-3B",
+    "meta-llama/Llama-3.2-1B-Instruct",
+    "meta-llama/Llama-3.2-3B-Instruct",
+    "meta-llama/Llama-3.3-70B-Instruct",
     "Baidicoot/Othello-GPT-Transformer-Lens",
     "bert-base-cased",
     "roneneldan/TinyStories-1M",
@@ -960,6 +961,30 @@ def convert_hf_model_config(model_name: str, **kwargs):
             "NTK_by_parts_high_freq_factor": 4.0,
             "NTK_by_parts_factor": 32.0,
         }
+    elif "Llama-3.3-70B" in official_model_name:
+        cfg_dict = {
+            "d_model": 8192,
+            "d_head": 128,
+            "n_heads": 64,
+            "d_mlp": 28672,
+            "n_layers": 80,
+            "n_ctx": 2048,  # capped due to memory issues
+            "eps": 1e-5,
+            "d_vocab": 128256,
+            "act_fn": "silu",
+            "n_key_value_heads": 8,
+            "normalization_type": "RMS",
+            "positional_embedding_type": "rotary",
+            "rotary_adjacent_pairs": False,
+            "rotary_dim": 32,
+            "final_rms": True,
+            "gated_mlp": True,
+            "rotary_base": 500000.0,
+            "use_NTK_by_parts_rope": True,
+            "NTK_by_parts_low_freq_factor": 1.0,
+            "NTK_by_parts_high_freq_factor": 4.0,
+            "NTK_by_parts_factor": 8.0,
+        }
     elif "Llama-3.1-8B" in official_model_name:
         cfg_dict = {
             "d_model": 4096,