llama : add note about context size queries

Author: ggerganov

include/llama.h

@@ -461,6 +461,8 @@ extern "C" {
     LLAMA_API bool llama_supports_gpu_offload(void);
     LLAMA_API bool llama_supports_rpc        (void);
 
+    // NOTE: After creating a llama_context, it is recommended to query the actual values using these functions
+    //       In some cases the requested values via llama_context_params may differ from the actual values used by the context
     LLAMA_API uint32_t llama_n_ctx      (const struct llama_context * ctx);
     LLAMA_API uint32_t llama_n_ctx_seq  (const struct llama_context * ctx);
     LLAMA_API uint32_t llama_n_batch    (const struct llama_context * ctx);
@@ -586,7 +588,7 @@ extern "C" {
     LLAMA_API int32_t llama_adapter_meta_val_str_by_index(const struct llama_adapter_lora * adapter, int32_t i, char * buf, size_t buf_size);
 
     // Manually free a LoRA adapter
-    // Note: loaded adapters will be free when the associated model is deleted
+    // NOTE: loaded adapters will be free when the associated model is deleted
     LLAMA_API void llama_adapter_lora_free(struct llama_adapter_lora * adapter);
 
     // Get the invocation tokens if the current lora is an alora

src/llama-context.cpp

@@ -112,14 +112,24 @@ llama_context::llama_context(
         }
     }
 
-    cparams.n_ctx_seq = cparams.kv_unified ? cparams.n_ctx : cparams.n_ctx / cparams.n_seq_max;
+    if (cparams.kv_unified) {
+        cparams.n_ctx_seq = cparams.n_ctx;
+    } else {
+        cparams.n_ctx_seq = cparams.n_ctx / cparams.n_seq_max;
+    }
 
     if (cparams.n_ctx_seq > hparams.n_ctx_train) {
         LLAMA_LOG_WARN("%s: capping n_ctx_seq (%u) to n_ctx_train (%u)\n", __func__, cparams.n_ctx_seq, hparams.n_ctx_train);
 
         cparams.n_ctx_seq = hparams.n_ctx_train;
     }
 
+    if (cparams.kv_unified) {
+        cparams.n_ctx = cparams.n_ctx_seq;
+    } else {
+        cparams.n_ctx = cparams.n_ctx_seq * cparams.n_seq_max;
+    }
+
     LLAMA_LOG_INFO("%s: n_seq_max     = %u\n",   __func__, cparams.n_seq_max);
     LLAMA_LOG_INFO("%s: n_ctx         = %u\n",   __func__, cparams.n_ctx);
     LLAMA_LOG_INFO("%s: n_ctx_seq     = %u\n",   __func__, cparams.n_ctx_seq);