sd: sync with master-816-487de75 (#2395)

otherarch/sdcpp/src/name_conversion.cpp

@@ -185,6 +185,20 @@ std::string convert_cond_stage_model_name(std::string name, std::string prefix)
 }
 
 std::string convert_qwen3_vl_vision_name(std::string name) {
+    static const std::vector<std::pair<std::string, std::string>> qwen3_vl_deepstack_name_map{
+        {"v.deepstack_merger_list.", "deepstack_merger_list."},
+        {"v.deepstack.5.", "deepstack_merger_list.0."},
+        {"v.deepstack.8.", "deepstack_merger_list.0."},
+        {"v.deepstack.11.", "deepstack_merger_list.1."},
+        {"v.deepstack.16.", "deepstack_merger_list.1."},
+        {"v.deepstack.17.", "deepstack_merger_list.2."},
+        {"v.deepstack.24.", "deepstack_merger_list.2."},
+        {"fc1.", "linear_fc1."},
+        {"fc2.", "linear_fc2."},
+        {"ffn_up.", "linear_fc1."},
+        {"ffn_down.", "linear_fc2."},
+        {"ffn_norm.", "norm."},
+    };
     static const std::vector<std::pair<std::string, std::string>> qwen3_vl_vision_name_map{
         {"mm.0.", "merger.linear_fc1."},
         {"mm.2.", "merger.linear_fc2."},
@@ -201,6 +215,10 @@ std::string convert_qwen3_vl_vision_name(std::string name) {
         {"ln1.", "norm1."},
         {"ln2.", "norm2."},
     };
+    if (contains(name, "v.deepstack_merger_list.") || contains(name, "v.deepstack.")) {
+        replace_with_name_map(name, qwen3_vl_deepstack_name_map);
+        return name;
+    }
     replace_with_name_map(name, qwen3_vl_vision_name_map);
     return name;
 }
@@ -1181,6 +1199,7 @@ std::string convert_sep_to_dot(std::string name) {
         "x_embedder",
         "cross_attn",
         "output_proj",
+        "token_refiner",
     };
 
     // record the positions of underscores that should NOT be replaced

otherarch/sdcpp/src/runtime/latent-preview.h

@@ -4,6 +4,33 @@
 #include "core/tensor.hpp"
 #include "ggml.h"
 
+const float minimax_latent_rgb_proj[24][3] = {
+    {0.19819857f, 0.11584999f, 0.07929777f},
+    {-0.16047224f, -0.10601170f, -0.15996324f},
+    {0.47391951f, 0.37602475f, 0.20267826f},
+    {-0.09857441f, -0.27435449f, -0.51681751f},
+    {-0.18930605f, -0.10512278f, -0.28571478f},
+    {-0.15639569f, -0.18000929f, -0.25432852f},
+    {-0.07176921f, -0.10901598f, -0.06654253f},
+    {-0.05014077f, -0.05839826f, -0.05516087f},
+    {-0.05201424f, -0.04351913f, -0.01507579f},
+    {0.24750438f, 0.13307422f, 0.17684120f},
+    {0.07377446f, 0.10235858f, 0.11707827f},
+    {0.02908304f, 0.06587022f, 0.10643690f},
+    {-0.00670531f, -0.03857879f, 0.01750151f},
+    {-0.07119107f, -0.03083323f, -0.01995450f},
+    {-0.08612627f, -0.07253841f, -0.01442890f},
+    {0.08793202f, 0.08681750f, 0.02994647f},
+    {0.00876893f, 0.02721868f, 0.00091178f},
+    {-0.03484412f, -0.02711262f, -0.00110101f},
+    {-0.00679772f, -0.01844275f, -0.01683359f},
+    {0.04287028f, 0.01601068f, 0.04037397f},
+    {-0.00493432f, -0.00230528f, 0.00353911f},
+    {0.01495088f, 0.00292306f, 0.00416671f},
+    {0.00495307f, 0.05066542f, 0.05210543f},
+    {-0.02154842f, -0.01518524f, 0.00442402f}};
+float minimax_latent_rgb_bias[3] = {0.07776964f, -0.01580954f, -0.06561434f};
+
 const float ltxav_latent_rgb_proj[128][3] = {
     {-0.0293802f, -0.0362516f, -0.0291386f},
     {0.0117735f, 0.0223435f, 0.018856f},

otherarch/sdcpp/src/stable-diffusion.cpp

@@ -2572,6 +2572,14 @@ public:
                     LOG_WARN("No latent to RGB projection known for this model");
                     return;
                 }
+            } else if (channels == 24) {
+                if(sd_version_is_minimax_h3(version)){
+                    latent_rgb_proj = minimax_latent_rgb_proj;
+                    latent_rgb_bias = minimax_latent_rgb_bias;
+                } else {
+                    LOG_WARN("No latent to RGB projection known for this model");
+                    return;
+                }
             } else if (channels == 16) {
                 if (sd_version_is_sd3(version)) {
                     latent_rgb_proj = sd3_latent_rgb_proj;
@@ -5925,6 +5933,18 @@ SD_API bool generate_image(sd_ctx_t* sd_ctx,
         return false;
     }
 
+    // MiniMax-H3 is video-only. Its denoiser always splits the packed latent into a video and an
+    // audio half, and only generate_video ever computes the audio length, so reaching this
+    // function with an H3 checkpoint is guaranteed to die on
+    // GGML_ASSERT(!audio_input_cache.empty()) with a core dump, after the several minutes it
+    // takes to load the weights, and with nothing in the output pointing at the missing --mode.
+    // (The AnimateDiff path below routes vid_gen back through here, but that is SD1.5 plus a
+    // motion module, never H3.)
+    if (sd_version_is_minimax_h3(sd_ctx->sd->version)) {
+        LOG_ERROR("MiniMax-H3 is a video model and cannot be run in img_gen mode; use --mode vid_gen");
+        return false;
+    }
+
     sd_ctx->sd->reset_cancel_flag();
 
     int64_t t0                    = ggml_time_ms();