fix(embedding): set kv_unified=True when embedding=True to enable batch processing (abetlen#2217)

* fix(embedding): set kv_unified=True when embedding=True to enable batch processing

* chore: update changelog for batch embedding fix

---------

Co-authored-by: abetlen <abetlen@gmail.com>

Author: SanjanaB123

CHANGELOG.md

@@ -7,6 +7,8 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ## [Unreleased]
 
+- fix: Enable unified KV cache for embedding contexts to preserve full per-sequence context in batch embedding calls by @SanjanaB123 in #2217
+
 ## [0.3.23]
 
 - feat: Update llama.cpp to ggerganov/llama.cpp@7d442abf

llama_cpp/llama.py

@@ -402,7 +402,7 @@ def __init__(
                 self.n_batch,
                 llama_cpp.llama_max_parallel_sequences(),
             )
-
+            self.context_params.kv_unified = True
         self._ctx = self._stack.enter_context(
             contextlib.closing(
                 internals.LlamaContext(