From: fairydreaming <redacted>
Date: Fri, 31 Jul 2026 07:03:30 +0000 (+0200)
Subject: llama : enforce the same K and V cache types for DeepSeek V4; enable FA if V cache... 
X-Git-Tag: upstream/0.0.10438~232
X-Git-Url: https://git.djapps.eu/?a=commitdiff_plain;h=69e62fc77c911da169cc8726b490028d53bb90fe;p=pkg%2Fggml%2Fsources%2Fllama.cpp

llama : enforce the same K and V cache types for DeepSeek V4; enable FA if V cache is quantized (#25871)

* llama : enforce the same K and V cache types for DeepSeek V4; enable FA if V cache is quantized

* llama : enforce the same K and V cache types for MLA models

---------

Co-authored-by: Stanisław Szymczyk <redacted>
---

diff --git a/src/llama-context.cpp b/src/llama-context.cpp
index 3e0af3f31..5ef7becf6 100644
--- a/src/llama-context.cpp
+++ b/src/llama-context.cpp
@@ -3553,6 +3553,22 @@ llama_context * llama_init_from_model(
         }
     }
 
+    if ((model->hparams.is_mla() || model->arch == LLM_ARCH_DEEPSEEK4) && params.type_k != params.type_v) {
+        LLAMA_LOG_ERROR("%s: model does not support different K (%s) and V (%s) cache types\n", __func__, ggml_type_name(params.type_k), ggml_type_name(params.type_v));
+        return nullptr;
+    }
+
+    if (ggml_is_quantized(params.type_v) && params.flash_attn_type != LLAMA_FLASH_ATTN_TYPE_ENABLED) {
+        if (params.flash_attn_type == LLAMA_FLASH_ATTN_TYPE_AUTO) {
+            LLAMA_LOG_INFO("%s: enabling flash_attn since it is required for quantized V cache\n", __func__);
+            params.flash_attn_type = LLAMA_FLASH_ATTN_TYPE_ENABLED;
+        }
+        if (params.flash_attn_type == LLAMA_FLASH_ATTN_TYPE_DISABLED) {
+            LLAMA_LOG_ERROR("%s: quantized V cache requires flash_attn to be enabled\n", __func__);
+            return nullptr;
+        }
+    }
+
     if (params.flash_attn_type != LLAMA_FLASH_ATTN_TYPE_DISABLED && ggml_is_quantized(params.type_k)) {
         const uint32_t blck_size = ggml_blck_size(params.type_k);
         for (uint32_t il = 0; il < model->hparams.n_layer(); ++il) {
@@ -3575,11 +3591,6 @@ llama_context * llama_init_from_model(
         }
     }
 
-    if (ggml_is_quantized(params.type_v) && params.flash_attn_type == LLAMA_FLASH_ATTN_TYPE_DISABLED) {
-        LLAMA_LOG_ERROR("%s: V cache quantization requires flash_attn\n", __func__);
-        return nullptr;
-    }
-
     if (params.pooling_type != LLAMA_POOLING_TYPE_UNSPECIFIED &&
         params.pooling_type != model->hparams.pooling_type) {
         //user-specified pooling-type is different from the model default