server : check if context can do shifts

ggml-ci

Author: ggerganov

src/llama-kv-cache.cpp

@@ -1579,7 +1579,7 @@ llama_kv_cache_unified_iswa::llama_kv_cache_unified_iswa(
     llama_kv_cache_unified::layer_filter_cb filter_swa  = [&](int32_t il) { return  model.hparams.is_swa(il); };
 
     const uint32_t kv_size_base = kv_size;
-    const uint32_t kv_size_swa  = std::min(kv_size, hparams.n_swa*n_seq_max + n_batch);
+    const uint32_t kv_size_swa  = std::min(kv_size, GGML_PAD(hparams.n_swa*n_seq_max + n_batch + 1, padding));
 
     kv_base = std::make_unique<llama_kv_cache_unified>(model, std::move(filter_base), type_k, type_v, v_trans, offload, kv_size_base, padding);
     kv_swa  = std::make_unique<llama_kv_cache_unified>(model, std::move(filter_swa),  type_k, type_v, v_trans, offload, kv_size_swa,  padding);

tools/server/server.cpp

@@ -2004,6 +2004,23 @@ struct server_context {
             }
         }
 
+        if (!llama_kv_self_can_shift(ctx)) {
+            if (params_base.ctx_shift) {
+                params_base.ctx_shift = false;
+                SRV_WRN("%s\n", "ctx_shift is not supported by this context, it will be disabled");
+            }
+
+            if (params_base.n_cache_reuse) {
+                params_base.n_cache_reuse = 0;
+                SRV_WRN("%s\n", "cache_reuse is not supported by this context, it will be disabled");
+            }
+
+            if (!params_base.speculative.model.path.empty()) {
+                SRV_ERR("%s\n", "err: speculative decode is not supported by this context");
+                return false;
+            }
+        }
+
         return true;
     }