Merge 5b62c11e49 into 1fe009caed

src/whisper.cpp

@@ -602,6 +602,39 @@ struct whisper_hparams {
     float   eps           = 1e-5f;
 };
 
+static bool whisper_is_valid_ftype(int32_t ftype) {
+    switch ((ggml_ftype) ftype) {
+        case GGML_FTYPE_ALL_F32:
+        case GGML_FTYPE_MOSTLY_F16:
+        case GGML_FTYPE_MOSTLY_BF16:
+        case GGML_FTYPE_MOSTLY_Q4_0:
+        case GGML_FTYPE_MOSTLY_Q4_1:
+        case GGML_FTYPE_MOSTLY_Q1_0:
+        case GGML_FTYPE_MOSTLY_Q5_0:
+        case GGML_FTYPE_MOSTLY_Q5_1:
+        case GGML_FTYPE_MOSTLY_Q8_0:
+        case GGML_FTYPE_MOSTLY_MXFP4:
+        case GGML_FTYPE_MOSTLY_NVFP4:
+        case GGML_FTYPE_MOSTLY_Q2_K:
+        case GGML_FTYPE_MOSTLY_Q3_K:
+        case GGML_FTYPE_MOSTLY_Q4_K:
+        case GGML_FTYPE_MOSTLY_Q5_K:
+        case GGML_FTYPE_MOSTLY_Q6_K:
+        case GGML_FTYPE_MOSTLY_IQ2_XXS:
+        case GGML_FTYPE_MOSTLY_IQ2_XS:
+        case GGML_FTYPE_MOSTLY_IQ3_XXS:
+        case GGML_FTYPE_MOSTLY_IQ1_S:
+        case GGML_FTYPE_MOSTLY_IQ1_M:
+        case GGML_FTYPE_MOSTLY_IQ4_NL:
+        case GGML_FTYPE_MOSTLY_IQ4_XS:
+        case GGML_FTYPE_MOSTLY_IQ3_S:
+        case GGML_FTYPE_MOSTLY_IQ2_S:
+            return true;
+        default:
+            return false;
+    }
+}
+
 // audio encoding layer
 struct whisper_layer_encoder {
     // encoder.blocks.*.attn_ln
@@ -1552,11 +1585,11 @@ static bool whisper_model_load(struct whisper_model_loader * loader, whisper_con
 
         // for the big tensors, we have the option to store the data in 16-bit floats or quantized
         // in order to save memory and also to speed up the computation
-        wctx.wtype = ggml_ftype_to_ggml_type((ggml_ftype) (model.hparams.ftype));
-        if (wctx.wtype == GGML_TYPE_COUNT) {
+        if (!whisper_is_valid_ftype(model.hparams.ftype)) {
             WHISPER_LOG_ERROR("%s: invalid model (bad ftype value %d)\n", __func__, model.hparams.ftype);
             return false;
         }
+        wctx.wtype = ggml_ftype_to_ggml_type((ggml_ftype) (model.hparams.ftype));
 
         WHISPER_LOG_INFO("%s: n_vocab       = %d\n", __func__, hparams.n_vocab);
         WHISPER_LOG_INFO("%s: n_audio_ctx   = %d\n", __func__, hparams.n_audio_ctx);

tests/CMakeLists.txt

@@ -96,6 +96,14 @@ target_link_libraries(${UTF8_TEST} PRIVATE common)
 add_test(NAME ${UTF8_TEST} COMMAND ${UTF8_TEST})
 set_tests_properties(${UTF8_TEST} PROPERTIES LABELS "unit")
 
+# Model loading unit test
+set(MODEL_LOAD_TEST test-model-load)
+add_executable(${MODEL_LOAD_TEST} ${MODEL_LOAD_TEST}.cpp)
+target_include_directories(${MODEL_LOAD_TEST} PRIVATE ../include ../ggml/include)
+target_link_libraries(${MODEL_LOAD_TEST} PRIVATE whisper)
+add_test(NAME ${MODEL_LOAD_TEST} COMMAND ${MODEL_LOAD_TEST})
+set_tests_properties(${MODEL_LOAD_TEST} PROPERTIES LABELS "unit")
+
 # VAD test tests VAD in isolation
 set(VAD_TEST test-vad)
 add_executable(${VAD_TEST} ${VAD_TEST}.cpp)

tests/test-model-load.cpp

@@ -0,0 +1,53 @@
+#include "ggml.h"
+#include "whisper.h"
+
+#include <cstdint>
+#include <cstdio>
+#include <fstream>
+
+static void write_u32(std::ofstream & fout, uint32_t value) {
+    fout.write(reinterpret_cast<const char *>(&value), sizeof(value));
+}
+
+static void write_i32(std::ofstream & fout, int32_t value) {
+    fout.write(reinterpret_cast<const char *>(&value), sizeof(value));
+}
+
+int main() {
+    const char * path = "invalid-ftype.ggml";
+
+    {
+        std::ofstream fout(path, std::ios::binary);
+        if (!fout) {
+            return 1;
+        }
+
+        write_u32(fout, GGML_FILE_MAGIC);
+        write_i32(fout, 1);    // n_vocab
+        write_i32(fout, 1500); // n_audio_ctx
+        write_i32(fout, 384);  // n_audio_state
+        write_i32(fout, 6);    // n_audio_head
+        write_i32(fout, 4);    // n_audio_layer
+        write_i32(fout, 448);  // n_text_ctx
+        write_i32(fout, 384);  // n_text_state
+        write_i32(fout, 6);    // n_text_head
+        write_i32(fout, 4);    // n_text_layer
+        write_i32(fout, 80);   // n_mels
+        write_i32(fout, 5);    // invalid ftype
+
+        if (!fout) {
+            return 2;
+        }
+    }
+
+    whisper_context_params params = whisper_context_default_params();
+    whisper_context * ctx = whisper_init_from_file_with_params(path, params);
+    if (ctx != nullptr) {
+        whisper_free(ctx);
+        std::remove(path);
+        return 3;
+    }
+
+    std::remove(path);
+    return 0;
+}