Merge pull request #108 from intel/a32543254-patch-4

A32543254 patch 4
intel · Feb 1, 2024 · d06db34 · d06db34
2 parents 29a1158 + 0ab0165
commit d06db34
Show file tree

Hide file tree

Showing 3 changed files with 24 additions and 0 deletions.
diff --git a/docs/supported_models.md b/docs/supported_models.md
@@ -43,6 +43,15 @@ Neural Speed supports the following models:
     <td>✅</td>
     <td>✅</td>
     <td>Latest</td>
+  </tr>
+    <td><a href="https://huggingface.co/upstage/SOLAR-10.7B-Instruct-v1.0" target="_blank" rel="noopener noreferrer">Solar-10.7B</a></td>
+    <td>✅</td>
+    <td>✅</td>
+    <td>✅</td>
+    <td>✅</td>
+    <td>✅</td>
+    <td>✅</td>
+    <td>Latest</td>
   </tr>
   <tr>
     <td><a href="https://huggingface.co/EleutherAI/gpt-j-6b" target="_blank" rel="noopener noreferrer">GPT-J-6B</a></td>
@@ -317,6 +326,14 @@ Neural Speed supports the following models:
     <td>✅</td>
     <td>✅</td>
     <td>✅</td>
+  </tr>
+    <tr>
+    <td><a href="https://huggingface.co/upstage/SOLAR-10.7B-Instruct-v1.0" target="_blank" rel="noopener noreferrer">Solar-10.7B</td>
+    <td>✅</td>
+    <td>✅</td>
+    <td>✅</td>
+    <td>✅</td>
+    <td>✅</td>
   </tr>
     </tr>
     <tr>

diff --git a/neural_speed/core/ne_layers.c b/neural_speed/core/ne_layers.c
@@ -10901,6 +10901,7 @@ void ne_graph_compute(struct ne_context* ctx, struct ne_cgraph* cgraph) {
 }
 
 void ne_graph_profiling(const struct ne_cgraph* cgraph) {
+#ifdef NS_PERF
   int64_t perf_total_per_op_us[NE_OP_COUNT] = {0};
 
   NE_PRINT("=== GRAPH Profiling ===\n");
@@ -10923,6 +10924,10 @@ void ne_graph_profiling(const struct ne_cgraph* cgraph) {
   }
   NE_PRINT("perf_total_per_op_us[%24s] = %7.3f ms\n", "INNER PRODUCT", (double)ip_duration / 1000.0);
   NE_PRINT("========================================\n");
+
+#else
+  NE_PRINT("\n[Warning] To collect profiling data, please recompile with NS_PROFILING=ON.\n");
+#endif
 }
 
 void ne_graph_reset(struct ne_cgraph* cgraph) {

diff --git a/setup.py b/setup.py
@@ -33,6 +33,7 @@ def check_env_flag(name: str, default: bool = False) -> bool:
 
 NS_WITH_AVX2 = check_env_flag("NS_WITH_AVX2", 'avx512f' not in cpu_flags)
 """ Whether to limit the max ISA used to AVX2; otherwise AVX512 will be used; set to ON/OFF """
+NS_PROFILING_ENV = os.environ.get("NS_PROFILING", "OFF")
 
 cwd = os.path.dirname(os.path.abspath(__file__))
 
@@ -104,6 +105,7 @@ def build_extension(self, ext: CMakeExtension) -> None:
             f"-DNS_WITH_AVX2={'ON' if NS_WITH_AVX2 else 'OFF'}",
             f"-DNS_WITH_TESTS=OFF",
             f"-DNS_PYTHON_API=ON",
+            f"-DNS_PROFILING={NS_PROFILING_ENV}",
         ]
         if sys.platform == "linux":  # relative_rpath
             cmake_args.append('-DCMAKE_BUILD_RPATH=$ORIGIN/')