v1.92065_b5439_RM1.102

Nexesenex · Nexesenex · commit 27ed1f244980 · 2025-05-21T21:15:31.000+02:00
diff --git a/gpttype_adapter.cpp b/gpttype_adapter.cpp
@@ -3337,7 +3337,7 @@ ModelLoadResult gpttype_load_model(const load_model_inputs inputs, FileFormat in
 
         //determine mem per token
         std::vector<int> tmp = {1, 2, 3, 4};
-        llama_kv_cache_clear(llama_ctx_v4);
+        // llama_kv_cache_clear(llama_ctx_v4);
         auto er = llama_decode(llama_ctx_v4, llama_batch_get_one(tmp.data(), tmp.size()));
         if(er!=0)
         {
diff --git a/koboldcpp.py b/koboldcpp.py
@@ -64,8 +64,8 @@
 # dry_seq_break_max = 128
 
 # global vars
-KcppVersion = "1.92060"
-LcppVersion = "b5427"
+KcppVersion = "1.92065"
+LcppVersion = "b5439"
 EsoboldVersion = "RMv1.102"
 CudaSpecifics = "Cu128_Ar86_SMC2_DmmvX32Y1"
 ReleaseDate = "2025/05/21"

Original file line number	Diff line number	Diff line change
`@@ -3337,7 +3337,7 @@ ModelLoadResult gpttype_load_model(const load_model_inputs inputs, FileFormat in`
`3337`	`3337`
`3338`	`3338`	`//determine mem per token`
`3339`	`3339`	`std::vector<int> tmp = {1, 2, 3, 4};`
`3340`		`- llama_kv_cache_clear(llama_ctx_v4);`
	`3340`	`+ // llama_kv_cache_clear(llama_ctx_v4);`
`3341`	`3341`	`auto er = llama_decode(llama_ctx_v4, llama_batch_get_one(tmp.data(), tmp.size()));`
`3342`	`3342`	`if(er!=0)`
`3343`	`3343`	`{`