From 6e08281e588bbba1a5d180290a94a43f167f3a1a Mon Sep 17 00:00:00 2001
From: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
Date: Sun, 29 Oct 2023 11:31:40 -0600
Subject: Extend llama_kv_cache_seq_rm to allow matching any sequence (#3843)

* Extend llama_kv_cache_seq_rm to allow matichng any sequence

* Replace llama_kv_cache_tokens_rm with llama_kv_cache_clear

Use llama_kv_cache_clear for cache clearing

Change calls to llama_kv_cache_tokens_rm that want to delete by position to use llama_kv_cache_seq_rm functionality
---
 llama.h | 15 ++++++---------
 1 file changed, 6 insertions(+), 9 deletions(-)

(limited to 'llama.h')

diff --git a/llama.h b/llama.h
index 6927bd60..d727dbd9 100644
--- a/llama.h
+++ b/llama.h
@@ -334,17 +334,14 @@ extern "C" {
     LLAMA_API DEPRECATED(int llama_get_kv_cache_token_count(const struct llama_context * ctx),
             "avoid using this, it will be removed in the future, instead - count the tokens in user code");
 
-    // Remove all tokens data of cells in [c0, c1)
-    // c0 < 0 : [0,  c1]
-    // c1 < 0 : [c0, inf)
-    LLAMA_API void llama_kv_cache_tokens_rm(
-            struct llama_context * ctx,
-                         int32_t   c0,
-                         int32_t   c1);
+    // Clear the KV cache
+    LLAMA_API void llama_kv_cache_clear(
+            struct llama_context * ctx);
 
     // Removes all tokens that belong to the specified sequence and have positions in [p0, p1)
-    // p0 < 0 : [0,  p1]
-    // p1 < 0 : [p0, inf)
+    // seq_id < 0 : match any sequence
+    // p0 < 0     : [0,  p1]
+    // p1 < 0     : [p0, inf)
     LLAMA_API void llama_kv_cache_seq_rm(
             struct llama_context * ctx,
                     llama_seq_id   seq_id,
-- 
cgit v1.2.3