From aeb42dc9004cd650fe348ffa78f6f3a0a8a6f15e Mon Sep 17 00:00:00 2001
From: Teruaki Ishizaki <tell.ishi@gmail.com>
Date: Tue, 4 Feb 2025 11:40:33 +0900
Subject: [PATCH] partners: Fixed the procedure of initializing pad_token_id
 (#29500)

- **Description:** Add to check pad_token_id and eos_token_id of model
config. It seems that this is the same bug as the HuggingFace TGI bug.
It's same bug as #29434
- **Issue:** #29431
- **Dependencies:** none
- **Twitter handle:** tell14

Example code is followings:
```python
from langchain_huggingface.llms import HuggingFacePipeline

hf = HuggingFacePipeline.from_model_id(
    model_id="meta-llama/Llama-3.2-3B-Instruct",
    task="text-generation",
    pipeline_kwargs={"max_new_tokens": 10},
)

from langchain_core.prompts import PromptTemplate

template = """Question: {question}

Answer: Let's think step by step."""
prompt = PromptTemplate.from_template(template)

chain = prompt | hf

question = "What is electroencephalography?"

print(chain.invoke({"question": question}))
```
---
 .../llms/huggingface_pipeline.py                      | 11 ++++++++++-
 .../tests/integration_tests/test_standard.py          | 10 ++++++----
 .../huggingface/tests/unit_tests/test_chat_models.py  | 11 +++++++----
 3 files changed, 23 insertions(+), 9 deletions(-)

diff --git a/libs/partners/huggingface/langchain_huggingface/llms/huggingface_pipeline.py b/libs/partners/huggingface/langchain_huggingface/llms/huggingface_pipeline.py
index daecda0d87..4314b906ab 100644
--- a/libs/partners/huggingface/langchain_huggingface/llms/huggingface_pipeline.py
+++ b/libs/partners/huggingface/langchain_huggingface/llms/huggingface_pipeline.py
@@ -202,7 +202,16 @@ class HuggingFacePipeline(BaseLLM):
         model = model_cls.from_pretrained(model_id, **_model_kwargs)
 
         if tokenizer.pad_token is None:
-            tokenizer.pad_token_id = model.config.eos_token_id
+            if model.config.pad_token_id is not None:
+                tokenizer.pad_token_id = model.config.pad_token_id
+            elif model.config.eos_token_id is not None and isinstance(
+                model.config.eos_token_id, int
+            ):
+                tokenizer.pad_token_id = model.config.eos_token_id
+            elif tokenizer.eos_token_id is not None:
+                tokenizer.pad_token_id = tokenizer.eos_token_id
+            else:
+                tokenizer.add_special_tokens({"pad_token": "[PAD]"})
 
         if (
             (
diff --git a/libs/partners/huggingface/tests/integration_tests/test_standard.py b/libs/partners/huggingface/tests/integration_tests/test_standard.py
index ce9b5b6b44..ca2e930c93 100644
--- a/libs/partners/huggingface/tests/integration_tests/test_standard.py
+++ b/libs/partners/huggingface/tests/integration_tests/test_standard.py
@@ -67,12 +67,14 @@ class TestHuggingFaceEndpoint(ChatModelIntegrationTests):
         super().test_bind_runnables_as_tools(model)
 
     @pytest.mark.xfail(reason=("Not implemented"))
-    def test_structured_output(self, model: BaseChatModel) -> None:
-        super().test_structured_output(model)
+    def test_structured_output(self, model: BaseChatModel, schema_type: str) -> None:
+        super().test_structured_output(model, schema_type)
 
     @pytest.mark.xfail(reason=("Not implemented"))
-    def test_structured_output_async(self, model: BaseChatModel) -> None:  # type: ignore[override]
-        super().test_structured_output(model)
+    async def test_structured_output_async(
+        self, model: BaseChatModel, schema_type: str
+    ) -> None:  # type: ignore[override]
+        super().test_structured_output(model, schema_type)
 
     @pytest.mark.xfail(reason=("Not implemented"))
     def test_structured_output_pydantic_2_v1(self, model: BaseChatModel) -> None:
diff --git a/libs/partners/huggingface/tests/unit_tests/test_chat_models.py b/libs/partners/huggingface/tests/unit_tests/test_chat_models.py
index fb9d141d02..10b67c93e8 100644
--- a/libs/partners/huggingface/tests/unit_tests/test_chat_models.py
+++ b/libs/partners/huggingface/tests/unit_tests/test_chat_models.py
@@ -248,10 +248,13 @@ def test_bind_tools_errors(
 
 def test_bind_tools(chat_hugging_face: Any) -> None:
     tools = [MagicMock(spec=BaseTool)]
-    with patch(
-        "langchain_huggingface.chat_models.huggingface.convert_to_openai_tool",
-        side_effect=lambda x: x,
-    ), patch("langchain_core.runnables.base.Runnable.bind") as mock_super_bind:
+    with (
+        patch(
+            "langchain_huggingface.chat_models.huggingface.convert_to_openai_tool",
+            side_effect=lambda x: x,
+        ),
+        patch("langchain_core.runnables.base.Runnable.bind") as mock_super_bind,
+    ):
         chat_hugging_face.bind_tools(tools, tool_choice="auto")
         mock_super_bind.assert_called_once()
         _, kwargs = mock_super_bind.call_args