From 07aa1bbc56a0f0ecd0175aefaed98aed7d21ace3 Mon Sep 17 00:00:00 2001
From: Richard Edgar <riedgar@microsoft.com>
Date: Fri, 3 May 2024 10:45:52 -0400
Subject: [PATCH] [Models] Add Phi3 to tests (#776)

Phi3 is now out, so start adding it into our testing frameworks
---
 .github/workflows/action_gpu_unit_tests.yml   | 3 +++
 .github/workflows/action_plain_unit_tests.yml | 2 ++
 tests/conftest.py                             | 4 ++++
 tests/models/test_transformers.py             | 6 ++----
 4 files changed, 11 insertions(+), 4 deletions(-)

diff --git a/.github/workflows/action_gpu_unit_tests.yml b/.github/workflows/action_gpu_unit_tests.yml
index 4273db977..b4bcd7da3 100644
--- a/.github/workflows/action_gpu_unit_tests.yml
+++ b/.github/workflows/action_gpu_unit_tests.yml
@@ -50,6 +50,9 @@ jobs:
           pip install pytest
           pip install -e .[schemas,test]
           if [ -f requirements.txt ]; then pip install -r requirements.txt; fi
+      - name: Other dependencies
+        run: |
+          pip install sentencepiece
       - name: GPU pip installs
         run: |
           pip install accelerate
diff --git a/.github/workflows/action_plain_unit_tests.yml b/.github/workflows/action_plain_unit_tests.yml
index 58e3c6450..a675a88fe 100644
--- a/.github/workflows/action_plain_unit_tests.yml
+++ b/.github/workflows/action_plain_unit_tests.yml
@@ -22,6 +22,7 @@ jobs:
           # - "transformers_mistral_7b" See Issue 713
           - "hfllama7b"
           - "hfllama_mistral_7b"
+          - "phi3cpu-mini-4k-instruct"
 
     runs-on: ${{ inputs.os }}
     steps:
@@ -44,6 +45,7 @@ jobs:
           if [ -f requirements.txt ]; then pip install -r requirements.txt; fi
       - name: Install model-specific dependencies
         run: |
+          pip install sentencepiece
           pip install "llama-cpp-python!=0.2.58"
       - name: Run tests (except server)
         shell: bash
diff --git a/tests/conftest.py b/tests/conftest.py
index 30a5c60ef..40903a204 100644
--- a/tests/conftest.py
+++ b/tests/conftest.py
@@ -12,6 +12,10 @@
     "phi2cpu": dict(
         name="transformers:microsoft/phi-2", kwargs={"trust_remote_code": True}
     ),
+    "phi3cpu-mini-4k-instruct": dict(
+        name="transformers:microsoft/Phi-3-mini-4k-instruct",
+        kwargs={"trust_remote_code": True},
+    ),
     "hfllama7b": dict(
         name="huggingface_hubllama:TheBloke/Llama-2-7B-GGUF:llama-2-7b.Q5_K_M.gguf",
         kwargs={"verbose": True, "n_ctx": 4096},
diff --git a/tests/models/test_transformers.py b/tests/models/test_transformers.py
index 5deaf8874..3a73e3127 100644
--- a/tests/models/test_transformers.py
+++ b/tests/models/test_transformers.py
@@ -1,6 +1,6 @@
 import pytest
 
-from guidance import gen, select
+from guidance import gen, select, models, assistant, system, user
 
 from ..utils import get_model
 
@@ -59,7 +59,6 @@ def test_transformer_smoke_select(model_name, model_kwargs):
 
 @pytest.mark.skip("Don't overload the build machines")
 def test_phi3_loading():
-    from guidance import models
 
     lm = models.Transformers(
         r"microsoft/Phi-3-mini-4k-instruct", trust_remote_code=True
@@ -70,8 +69,7 @@ def test_phi3_loading():
 
 @pytest.mark.skip("Don't overload the build machines")
 def test_phi3_chat():
-    # TODO: we currently use the wrong chat template for this model, need to update to match: https://huggingface.co/microsoft/Phi-3-mini-4k-instruct
-    from guidance import models, system, user, assistant
+    # TODO: Double check chat format: https://huggingface.co/microsoft/Phi-3-mini-4k-instruct
 
     lm = models.TransformersChat(
         r"microsoft/Phi-3-mini-4k-instruct", trust_remote_code=True