From fadef5f31ef6acd9b39b72103931b5eb62f98585 Mon Sep 17 00:00:00 2001
From: Sam Tsao <sam89881@gmail.com>
Date: Wed, 13 Dec 2023 13:24:48 +0800
Subject: [PATCH] Make ollama support additional_kwargs correctly (#9455)
make ollama_additional_kwargs working
---
llama_index/embeddings/ollama_embedding.py | 2 +-
llama_index/llms/ollama.py | 4 ++--
2 files changed, 3 insertions(+), 3 deletions(-)
diff --git a/llama_index/embeddings/ollama_embedding.py b/llama_index/embeddings/ollama_embedding.py
index d35f2d9f7e..de26ca1c63 100644
--- a/llama_index/embeddings/ollama_embedding.py
+++ b/llama_index/embeddings/ollama_embedding.py
@@ -83,7 +83,7 @@ class OllamaEmbedding(BaseEmbedding):
ollama_request_body = {
"prompt": prompt,
"model": self.model_name,
- **self.ollama_additional_kwargs,
+ "options": self.ollama_additional_kwargs,
}
response = requests.post(
diff --git a/llama_index/llms/ollama.py b/llama_index/llms/ollama.py
index 078e2f7a3c..3cf5af31c5 100644
--- a/llama_index/llms/ollama.py
+++ b/llama_index/llms/ollama.py
@@ -63,7 +63,7 @@ class Ollama(CustomLLM):
def _model_kwargs(self) -> Dict[str, Any]:
base_kwargs = {
"temperature": self.temperature,
- "max_length": self.context_window,
+ "num_ctx": self.context_window,
}
return {
**base_kwargs,
@@ -117,7 +117,7 @@ class Ollama(CustomLLM):
response = requests.post(
url=f"{self.base_url}/api/generate/",
headers={"Content-Type": "application/json"},
- json={"prompt": prompt, "model": self.model, **all_kwargs},
+ json={"prompt": prompt, "model": self.model, "options": all_kwargs},
stream=True,
)
response.encoding = "utf-8"
--
GitLab