add whisper model check (#189)

intel · Mar 22, 2024 · 66bcc8b · 66bcc8b
1 parent ed6e8ad
commit 66bcc8b
Show file tree

Hide file tree

Showing 2 changed files with 11 additions and 4 deletions.
diff --git a/neural_speed/application/quant_whisper.cpp b/neural_speed/application/quant_whisper.cpp
@@ -23,6 +23,9 @@
 #include "models/model_utils/quant_utils.h"
 #include "common.h"
 
+#define F_OK 0
+
+inline bool exists_model(const std::string& name) { return (access(name.c_str(), F_OK) != -1); }
 int main(int argc, char** argv) {
   quant_params q_params;
   if (quant_params_parse(argc, argv, q_params) == false) {
@@ -52,9 +55,13 @@ int main(int argc, char** argv) {
   // load the model
   {
     const int64_t t_start_us = ne_time_us();
-
-    if (!whisper_model_quantize(fname_inp, fname_out, ne_ftype(ftype))) {
-      fprintf(stderr, "%s: failed to quantize model from '%s'\n", __func__, fname_inp.c_str());
+    if (exists_model(fname_inp)) {
+      if (!whisper_model_quantize(fname_inp, fname_out, ne_ftype(ftype))) {
+        fprintf(stderr, "%s: failed to quantize model from '%s'\n", __func__, fname_inp.c_str());
+        return 1;
+      }
+    } else {
+      fprintf(stderr, "%s: model is not exist '%s'\n", __func__, fname_inp.c_str());
       return 1;
     }
 

diff --git a/neural_speed/models/gemma/gemma_utils.cpp b/neural_speed/models/gemma/gemma_utils.cpp
@@ -83,7 +83,7 @@ void Gemma::init(const char* path_model, model_context* ctx, int n_gpu_layer_, b
   n_head = hparams.n_head;
   n_expert = hparams.n_experts;
   n_expert_used = hparams.n_experts_used;
-  scratch = gemma_mem_req(n_layer, lctx.scratch_size_ratio);
+  scratch = gemma_mem_req(n_layer, lctx.scratch_size_ratio * 1.5);
   model.scratchs = scratch;
 }