Merge commit '12a4a47e6aaf691492644c39da453745aaee1672' into concedo_experimental

# Conflicts:
#	ci/run.sh
#	examples/model-conversion/scripts/causal/run-converted-model-embeddings-logits.sh
#	examples/model-conversion/scripts/causal/run-converted-model.sh
#	examples/model-conversion/scripts/embedding/run-converted-model.sh
#	ggml/src/ggml-cann/ggml-cann.cpp
#	ggml/src/ggml-hexagon/ggml-hexagon.cpp
#	ggml/src/ggml-opencl/ggml-opencl.cpp
#	ggml/src/ggml-sycl/ggml-sycl.cpp
#	ggml/src/ggml-webgpu/ggml-webgpu.cpp
#	ggml/src/ggml-zdnn/ggml-zdnn.cpp
#	ggml/src/ggml-zendnn/ggml-zendnn.cpp
#	tests/CMakeLists.txt
#	tests/test-chat-parser.cpp
#	tests/test-chat-peg-parser.cpp
#	tests/test-chat.cpp
#	tools/cli/cli.cpp

src/llama-mmap.cpp

@@ -265,7 +265,8 @@ struct llama_file::impl {
                         continue;  // Interrupted by signal, retry
                     }
                     // Fallback to std::fread in case the DMA controller cannot access the buffer
-                    if (errno == EFAULT) {
+                    if (errno == EFAULT || errno == EINVAL) {
+                        LLAMA_LOG_WARN("%s: Falling back to buffered IO due to %s\n", __func__, strerror(errno));
                         auto curr_off = tell();
                         close(fd);
                         fd = -1;
@@ -384,6 +385,9 @@ int llama_file::file_id() const {
 #ifdef _WIN32
     return _fileno(pimpl->fp);
 #else
+    if (pimpl->fd != -1) {
+        return pimpl->fd;
+    }
 #if defined(fileno)
     return fileno(pimpl->fp);
 #else

src/llama-model-loader.cpp

@@ -543,12 +543,18 @@ llama_model_loader::llama_model_loader(
     files.emplace_back(new llama_file(fname.c_str(), "rb", use_direct_io));
     contexts.emplace_back(ctx);
 
-    use_direct_io = use_direct_io && files.back()->has_direct_io();
-
-    // Disable mmap in case Direct I/O is enabled and available
-    if (use_direct_io && use_mmap) {
-        use_mmap = false;
-        LLAMA_LOG_WARN("%s: direct I/O is enabled, disabling mmap\n", __func__);
+    if (use_mmap && use_direct_io) {
+        if (files.back()->has_direct_io()) {
+            // Disable mmap, as DirectIO is available
+            use_mmap = false;
+            LLAMA_LOG_WARN("%s: direct I/O is enabled, disabling mmap\n", __func__);
+        } else {
+            // Disable DirectIO and reopen file using std::fopen for mmap
+            use_direct_io = false;
+            files.pop_back();
+            files.emplace_back(new llama_file(fname.c_str(), "rb", false));
+            LLAMA_LOG_WARN("%s: direct I/O is not available, using mmap\n", __func__);
+        }
     }
 
     // Save tensors data offset of the main file.

src/llama-model.cpp

@@ -1823,7 +1823,12 @@ void llama_model::load_hparams(llama_model_loader & ml) {
                 if (hparams.expert_gating_func == LLAMA_EXPERT_GATING_FUNC_TYPE_NONE) {
                     // for compatibility with existing DeepSeek V2 and V2.5 GGUFs
                     // that have no expert_gating_func model parameter set
-                    hparams.expert_gating_func = LLAMA_EXPERT_GATING_FUNC_TYPE_SOFTMAX;
+                    if ((hparams.n_layer == 47 || hparams.n_layer == 48) && n_vocab == 154880) {
+                        // GLM 4.7 Lite
+                        hparams.expert_gating_func = LLAMA_EXPERT_GATING_FUNC_TYPE_SIGMOID;
+                    } else {
+                        hparams.expert_gating_func = LLAMA_EXPERT_GATING_FUNC_TYPE_SOFTMAX;
+                    }
                 }
 
                 if (ml.get_key(LLM_KV_ROPE_SCALING_YARN_LOG_MUL, hparams.rope_yarn_log_mul, 0.0f)) {