From d6829284c4912d5186469c4a624127c24b231d34 Mon Sep 17 00:00:00 2001
From: askerlee <shaohua@gmail.com>
Date: Tue, 13 Jan 2026 22:20:22 +0800
Subject: [PATCH] Allow local install and model loading

---
 .gitignore           | 1 +
 pyproject.toml       | 3 +++
 scripts/base_eval.py | 7 +++++--
 3 files changed, 9 insertions(+), 2 deletions(-)

diff --git a/.gitignore b/.gitignore
index d82809a1..740d38b6 100644
--- a/.gitignore
+++ b/.gitignore
@@ -12,3 +12,4 @@ eval_bundle/
 .claude
 CLAUDE.md
 wandb/
+*.egg-info/
diff --git a/pyproject.toml b/pyproject.toml
index 87a967f7..3f0c1e82 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -71,3 +71,6 @@ conflicts = [
         { extra = "gpu" },
     ],
 ]
+
+[tool.setuptools]
+packages = ["nanochat"]
diff --git a/scripts/base_eval.py b/scripts/base_eval.py
index bd83ff36..672faec0 100644
--- a/scripts/base_eval.py
+++ b/scripts/base_eval.py
@@ -134,13 +134,16 @@ def load_hf_model(hf_path: str, device):
     print0(f"Loading model from: {hf_path}")
     # Load the model
     from transformers import AutoModelForCausalLM
-    model = AutoModelForCausalLM.from_pretrained(hf_path)
+    model = AutoModelForCausalLM.from_pretrained(hf_path, trust_remote_code=True)
     model.to(device)
     model.eval()
     max_seq_len = 1024 if "openai-community/gpt2" in hf_path else None
     model = ModelWrapper(model, max_seq_len=max_seq_len)
     # Load the tokenizer
-    tokenizer = HuggingFaceTokenizer.from_pretrained(hf_path)
+    if os.path.exists(hf_path):
+        tokenizer = HuggingFaceTokenizer.from_directory(hf_path)
+    else:
+        tokenizer = HuggingFaceTokenizer.from_pretrained(hf_path)
     return model, tokenizer
 
 # -----------------------------------------------------------------------------