created a seperate list of models to test for public PRs (#920)

* created a seperate list of models to test for public PRs

* ran format

Author: bryce13950

tests/acceptance/test_evals.py

@@ -6,7 +6,7 @@
 
 @pytest.fixture(scope="module")
 def model():
-    return HookedTransformer.from_pretrained("gpt2-small")
+    return HookedTransformer.from_pretrained("gpt2-small", device="cpu")
 
 
 def test_basic_ioi_eval(model):

tests/acceptance/test_hooked_transformer.py

@@ -21,7 +21,19 @@
 
 PYTHIA_MODEL_NAMES = [name for name in OFFICIAL_MODEL_NAMES if name.startswith("EleutherAI/pythia")]
 
-model_names = [
+# Small models for basic testing
+PUBLIC_MODEL_NAMES = [
+    "attn-only-demo",
+    "gpt2-small",
+    "opt-125m",
+    "pythia-70m",
+    "tiny-stories-33M",
+    "microsoft/phi-1",
+    "google/gemma-2b",
+]
+
+# Full set of models to test
+FULL_MODEL_NAMES = [
     "attn-only-demo",
     "gpt2-small",
     "opt-125m",
@@ -42,6 +54,10 @@
     "google/gemma-2b",
     "google/gemma-7b",
 ]
+
+# Use full model list if HF_TOKEN is available, otherwise use public models only
+model_names = FULL_MODEL_NAMES if os.environ.get("HF_TOKEN", "") else PUBLIC_MODEL_NAMES
+
 text = "Hello world!"
 """ 
 # Code to regenerate loss store

tests/acceptance/test_tokenizer_special_tokens.py

@@ -1,10 +1,15 @@
+import os
+
 from transformers import AutoTokenizer
 
 import transformer_lens.loading_from_pretrained as loading
 from transformer_lens import HookedTransformer, HookedTransformerConfig
 
-# Get's tedious typing these out everytime I want to sweep over all the distinct small models
-MODEL_TESTING_LIST = [
+# Small models for basic testing
+PUBLIC_MODEL_TESTING_LIST = ["gpt2-small", "opt-125m", "pythia-70m"]
+
+# Full set of models to test when HF_TOKEN is available
+FULL_MODEL_TESTING_LIST = [
     "solu-1l",
     "gpt2-small",
     "gpt-neo-125M",
@@ -14,6 +19,11 @@
     "pythia-70m",
 ]
 
+# Use full model list if HF_TOKEN is available, otherwise use public models only
+MODEL_TESTING_LIST = (
+    FULL_MODEL_TESTING_LIST if os.environ.get("HF_TOKEN", "") else PUBLIC_MODEL_TESTING_LIST
+)
+
 
 def test_d_vocab_from_tokenizer():
     cfg = HookedTransformerConfig(