diff --git a/src/model/diffusion/sensenova_u1.h b/src/model/diffusion/sensenova_u1.h
index 929f5d0da..22b789fac 100644
--- a/src/model/diffusion/sensenova_u1.h
+++ b/src/model/diffusion/sensenova_u1.h
@@ -146,7 +146,7 @@ namespace SenseNovaU1 {
             auto x     = ggml_ext_timestep_embedding(ctx->ggml_ctx,
                                                      timesteps,
                                                      static_cast<int>(frequency_embedding_size),
-                                                     10000.f,
+                                                     10000,
                                                      1.f);
             x          = mlp_0->forward(ctx, x);
             x          = ggml_silu_inplace(ctx->ggml_ctx, x);
diff --git a/src/model/diffusion/wan.hpp b/src/model/diffusion/wan.hpp
index 90fd1285b..4b75203d3 100644
--- a/src/model/diffusion/wan.hpp
+++ b/src/model/diffusion/wan.hpp
@@ -20,22 +20,22 @@ namespace WAN {
     constexpr int WAN_GRAPH_SIZE = 10240;
 
     struct WanConfig {
-        std::string model_type                 = "t2v";
-        std::tuple<int, int, int> patch_size   = {1, 2, 2};
-        int64_t text_len                       = 512;
-        int64_t in_dim                         = 16;
-        int64_t dim                            = 2048;
-        int64_t ffn_dim                        = 8192;
-        int freq_dim                           = 256;
-        int64_t text_dim                       = 4096;
-        int64_t out_dim                        = 16;
-        int64_t num_heads                      = 16;
-        int num_layers                         = 32;
-        int vace_layers                        = 0;
-        int64_t vace_in_dim                    = 96;
-        std::map<int, int> vace_layers_mapping = {};
+        std::string model_type                  = "t2v";
+        std::tuple<int, int, int> patch_size    = {1, 2, 2};
+        int64_t text_len                        = 512;
+        int64_t in_dim                          = 16;
+        int64_t dim                             = 2048;
+        int64_t ffn_dim                         = 8192;
+        int freq_dim                            = 256;
+        int64_t text_dim                        = 4096;
+        int64_t out_dim                         = 16;
+        int64_t num_heads                       = 16;
+        int num_layers                          = 32;
+        int vace_layers                         = 0;
+        int64_t vace_in_dim                     = 96;
+        std::map<int, int> vace_layers_mapping  = {};
         int64_t audio_dim                       = 1024;
-        int num_audio_token                     = 4;     // excludes the learned padding token
+        int num_audio_token                     = 4;  // excludes the learned padding token
         std::vector<int> audio_inject_layers    = {};
         std::map<int, int> audio_inject_mapping = {};  // block index -> injector index
         std::string adain_mode                  = "attn_norm";
diff --git a/src/model/vae/tae.hpp b/src/model/vae/tae.hpp
index 95bc0b271..e9257b7c8 100644
--- a/src/model/vae/tae.hpp
+++ b/src/model/vae/tae.hpp
@@ -564,7 +564,7 @@ class TAEHV : public GGMLBlock {
             int64_t chunk_frames = 5 * decoder->t_upscale;
             int64_t pad          = (chunk_frames - (num_frames % chunk_frames)) % chunk_frames;
 
-            result = ggml_ext_pad_ext(ctx->ggml_ctx, ctx->backend, result, 0, 0, 0, 0, 0, 0, 0, pad, false, false);
+            result = ggml_ext_pad_ext(ctx->ggml_ctx, ctx->backend, result, 0, 0, 0, 0, 0, 0, 0, static_cast<int>(pad), false, false);
 
             int64_t num_chunks                  = (num_frames + pad) / chunk_frames;
             auto to_trim                        = decoder->t_upscale - 1;
diff --git a/src/runtime/denoiser.hpp b/src/runtime/denoiser.hpp
index ec89406d3..99c2bf4b0 100644
--- a/src/runtime/denoiser.hpp
+++ b/src/runtime/denoiser.hpp
@@ -2789,9 +2789,9 @@ static sd::Tensor<float> sample_lms(denoise_cb_t model,
         sd::Tensor<float> d_cur = (x - denoised) / sigma;
         x += d_cur * lms_coeff[0];
         if (max_order > 1) {  // if max_order == 1, the history is not used (order always < 2)
-            int hist_size_p1 = hist.size() + 1;
+            int hist_size_p1 = static_cast<int>(hist.size()) + 1;
             if (i) {  // history does not exist at 1st step
-                int hist_max = hist.size() - 1;
+                int hist_max = static_cast<int>(hist.size()) - 1;
                 for (int c = 2; c <= order; c++)
                     x += hist[std::min(hist_max, hist_size_p1 - c + shift)] * lms_coeff[c - 1];
                 // max_order == 4  =>  hist[] index = 2, 1, 0