non-square VAE tiling (#3)

* refactor tile number calculation

* support non-square tiles

* add env var to change tile overlap

* add safeguards and better error messages for SD_TILE_OVERLAP

* add safeguards and include overlapping factor for SD_TILE_SIZE

* avoid rounding issues when specifying SD_TILE_SIZE as a factor

* lower SD_TILE_OVERLAP limit

Author: wbruna

ggml_extend.hpp

@@ -607,8 +607,38 @@ __STATIC_INLINE__ void ggml_tensor_scale_output(struct ggml_tensor* src) {
 
 typedef std::function<void(ggml_tensor*, ggml_tensor*, bool)> on_tile_process;
 
+__STATIC_INLINE__ void
+sd_tiling_calc_tiles(int &num_tiles_dim, float& tile_overlap_factor_dim, int small_dim, int tile_size, const float tile_overlap_factor) {
+
+    int tile_overlap     = (tile_size * tile_overlap_factor);
+    int non_tile_overlap = tile_size - tile_overlap;
+
+    num_tiles_dim = (small_dim - tile_overlap) / non_tile_overlap;
+    int overshoot_dim = ((num_tiles_dim + 1) * non_tile_overlap + tile_overlap) % small_dim;
+
+    if ((overshoot_dim != non_tile_overlap) && (overshoot_dim <= num_tiles_dim * (tile_size / 2 - tile_overlap))) {
+        // if tiles don't fit perfectly using the desired overlap
+        // and there is enough room to squeeze an extra tile without overlap becoming >0.5
+        num_tiles_dim++;
+    }
+
+    tile_overlap_factor_dim = (float)(tile_size * num_tiles_dim - small_dim) / (float)(tile_size * (num_tiles_dim - 1));
+    if (num_tiles_dim <= 2) {
+        if (small_dim <= tile_size) {
+            num_tiles_dim           = 1;
+            tile_overlap_factor_dim = 0;
+        } else {
+            num_tiles_dim           = 2;
+            tile_overlap_factor_dim = (2 * tile_size - small_dim) / (float)tile_size;
+        }
+    }
+}
+
 // Tiling
-__STATIC_INLINE__ void sd_tiling(ggml_tensor* input, ggml_tensor* output, const int scale, const int tile_size, const float tile_overlap_factor, on_tile_process on_processing) {
+__STATIC_INLINE__ void sd_tiling_non_square(ggml_tensor* input, ggml_tensor* output, const int scale,
+                                            const int p_tile_size_x, const int p_tile_size_y,
+                                            const float tile_overlap_factor, on_tile_process on_processing) {
+
     output = ggml_set_f32(output, 0);
 
     int input_width   = (int)input->ne[0];
@@ -629,62 +659,27 @@ __STATIC_INLINE__ void sd_tiling(ggml_tensor* input, ggml_tensor* output, const
         small_height = input_height;
     }
 
-    int tile_overlap     = (tile_size * tile_overlap_factor);
-    int non_tile_overlap = tile_size - tile_overlap;
-
-    int num_tiles_x = (small_width - tile_overlap) / non_tile_overlap;
-    int overshoot_x = ((num_tiles_x + 1) * non_tile_overlap + tile_overlap) % small_width;
-
-    if ((overshoot_x != non_tile_overlap) && (overshoot_x <= num_tiles_x * (tile_size / 2 - tile_overlap))) {
-        // if tiles don't fit perfectly using the desired overlap
-        // and there is enough room to squeeze an extra tile without overlap becoming >0.5
-        num_tiles_x++;
-    }
-
-    float tile_overlap_factor_x = (float)(tile_size * num_tiles_x - small_width) / (float)(tile_size * (num_tiles_x - 1));
-    if (num_tiles_x <= 2) {
-        if (small_width <= tile_size) {
-            num_tiles_x           = 1;
-            tile_overlap_factor_x = 0;
-        } else {
-            num_tiles_x           = 2;
-            tile_overlap_factor_x = (2 * tile_size - small_width) / (float)tile_size;
-        }
-    }
-
-    int num_tiles_y = (small_height - tile_overlap) / non_tile_overlap;
-    int overshoot_y = ((num_tiles_y + 1) * non_tile_overlap + tile_overlap) % small_height;
+    int num_tiles_x;
+    float tile_overlap_factor_x;
+    sd_tiling_calc_tiles(num_tiles_x, tile_overlap_factor_x, small_width, p_tile_size_x, tile_overlap_factor);
 
-    if ((overshoot_y != non_tile_overlap) && (overshoot_y <= num_tiles_y * (tile_size / 2 - tile_overlap))) {
-        // if tiles don't fit perfectly using the desired overlap
-        // and there is enough room to squeeze an extra tile without overlap becoming >0.5
-        num_tiles_y++;
-    }
-
-    float tile_overlap_factor_y = (float)(tile_size * num_tiles_y - small_height) / (float)(tile_size * (num_tiles_y - 1));
-    if (num_tiles_y <= 2) {
-        if (small_height <= tile_size) {
-            num_tiles_y           = 1;
-            tile_overlap_factor_y = 0;
-        } else {
-            num_tiles_y           = 2;
-            tile_overlap_factor_y = (2 * tile_size - small_height) / (float)tile_size;
-        }
-    }
+    int num_tiles_y;
+    float tile_overlap_factor_y;
+    sd_tiling_calc_tiles(num_tiles_y, tile_overlap_factor_y, small_height, p_tile_size_y, tile_overlap_factor);
 
     LOG_DEBUG("num tiles : %d, %d ", num_tiles_x, num_tiles_y);
     LOG_DEBUG("optimal overlap : %f, %f (targeting %f)", tile_overlap_factor_x, tile_overlap_factor_y, tile_overlap_factor);
 
     GGML_ASSERT(input_width % 2 == 0 && input_height % 2 == 0 && output_width % 2 == 0 && output_height % 2 == 0);  // should be multiple of 2
 
-    int tile_overlap_x     = (int32_t)(tile_size * tile_overlap_factor_x);
-    int non_tile_overlap_x = tile_size - tile_overlap_x;
+    int tile_overlap_x     = (int32_t)(p_tile_size_x * tile_overlap_factor_x);
+    int non_tile_overlap_x = p_tile_size_x - tile_overlap_x;
 
-    int tile_overlap_y     = (int32_t)(tile_size * tile_overlap_factor_y);
-    int non_tile_overlap_y = tile_size - tile_overlap_y;
+    int tile_overlap_y     = (int32_t)(p_tile_size_y * tile_overlap_factor_y);
+    int non_tile_overlap_y = p_tile_size_y - tile_overlap_y;
 
-    int tile_size_x = tile_size < small_width ? tile_size : small_width;
-    int tile_size_y = tile_size < small_height ? tile_size : small_height;
+    int tile_size_x = p_tile_size_x < small_width ? p_tile_size_x : small_width;
+    int tile_size_y = p_tile_size_y < small_height ? p_tile_size_y : small_height;
 
     int input_tile_size_x  = tile_size_x;
     int input_tile_size_y  = tile_size_y;
@@ -773,6 +768,11 @@ __STATIC_INLINE__ void sd_tiling(ggml_tensor* input, ggml_tensor* output, const
     ggml_free(tiles_ctx);
 }
 
+__STATIC_INLINE__ void sd_tiling(ggml_tensor* input, ggml_tensor* output, const int scale,
+    const int tile_size, const float tile_overlap_factor, on_tile_process on_processing) {
+    sd_tiling_non_square(input, output, scale, tile_size, tile_size, tile_overlap_factor, on_processing);
+}
+
 __STATIC_INLINE__ struct ggml_tensor* ggml_group_norm_32(struct ggml_context* ctx,
                                                          struct ggml_tensor* a) {
     const float eps = 1e-6f;  // default eps parameter

stable-diffusion.cpp

@@ -1427,23 +1427,91 @@ class StableDiffusionGGML {
                                                  x->ne[3]);  // channels
         int64_t t0          = ggml_time_ms();
 
-        int tile_size = 32;
-        // TODO: arg instead of env?
+        // TODO: args instead of env for tile size / overlap?
+
+        float tile_overlap = 0.5f;
+        const char* SD_TILE_OVERLAP = getenv("SD_TILE_OVERLAP");
+        if (SD_TILE_OVERLAP != nullptr) {
+            std::string sd_tile_overlap_str = SD_TILE_OVERLAP;
+            try {
+                tile_overlap = std::stof(sd_tile_overlap_str);
+                if (tile_overlap < 0.0) {
+                    LOG_WARN("SD_TILE_OVERLAP too low, setting it to 0.0");
+                    tile_overlap = 0.0;
+                }
+                else if (tile_overlap > 0.5) {
+                    LOG_WARN("SD_TILE_OVERLAP too high, setting it to 0.5");
+                    tile_overlap = 0.5;
+                }
+            } catch (const std::invalid_argument&) {
+                LOG_WARN("SD_TILE_OVERLAP is invalid, keeping the default");
+            } catch (const std::out_of_range&) {
+                LOG_WARN("SD_TILE_OVERLAP is out of range, keeping the default");
+            }
+        }
+
+        int tile_size_x = 32;
+        int tile_size_y = 32;
         const char* SD_TILE_SIZE = getenv("SD_TILE_SIZE");
         if (SD_TILE_SIZE != nullptr) {
+            // format is AxB, or just A (equivalent to AxA)
+            // A and B can be integers (tile size) or floating point
+            // floating point <= 1 means simple fraction of the latent dimension
+            // floating point > 1 means number of tiles across that dimension
+            // a single number gets applied to both
+            auto get_tile_factor = [tile_overlap](const std::string& factor_str) {
+                float factor = std::stof(factor_str);
+                if (factor > 1.0)
+                    factor = 1 / (factor - factor * tile_overlap + tile_overlap);
+                return factor;
+            };
+            const int latent_x = W / (decode ? 1 : 8);
+            const int latent_y = H / (decode ? 1 : 8);
+            const int min_tile_dimension = 4;
             std::string sd_tile_size_str = SD_TILE_SIZE;
+            size_t x_pos = sd_tile_size_str.find('x');
             try {
-                tile_size = std::stoi(sd_tile_size_str);
+                int tmp_x = tile_size_x, tmp_y = tile_size_y;
+                if (x_pos != std::string::npos) {
+                    std::string tile_x_str = sd_tile_size_str.substr(0, x_pos);
+                    std::string tile_y_str = sd_tile_size_str.substr(x_pos + 1);
+                    if (tile_x_str.find('.') != std::string::npos) {
+                        tmp_x = std::round(latent_x * get_tile_factor(tile_x_str));
+                    }
+                    else {
+                        tmp_x = std::stoi(tile_x_str);
+                    }
+                    if (tile_y_str.find('.') != std::string::npos) {
+                        tmp_y = std::round(latent_y * get_tile_factor(tile_y_str));
+                    }
+                    else {
+                        tmp_y = std::stoi(tile_y_str);
+                    }
+                }
+                else {
+                    if (sd_tile_size_str.find('.') != std::string::npos) {
+                        float tile_factor = get_tile_factor(sd_tile_size_str);
+                        tmp_x = std::round(latent_x * tile_factor);
+                        tmp_y = std::round(latent_y * tile_factor);
+                    }
+                    else {
+                        tmp_x = tmp_y = std::stoi(sd_tile_size_str);
+                    }
+                }
+                tile_size_x = std::max(std::min(tmp_x, latent_x), min_tile_dimension);
+                tile_size_y = std::max(std::min(tmp_y, latent_y), min_tile_dimension);
             } catch (const std::invalid_argument&) {
-                LOG_WARN("Invalid");
+                LOG_WARN("SD_TILE_SIZE is invalid, keeping the default");
             } catch (const std::out_of_range&) {
-                LOG_WARN("OOR");
+                LOG_WARN("SD_TILE_SIZE is out of range, keeping the default");
             }
         }
+
         if(!decode){
             // TODO: also use and arg for this one?
             // to keep the compute buffer size consistent
-            tile_size*=1.30539;
+            tile_size_x*=1.30539;
+            tile_size_y*=1.30539;
         }
         if (!use_tiny_autoencoder) {
             if (decode) {
@@ -1452,11 +1520,17 @@ class StableDiffusionGGML {
                 ggml_tensor_scale_input(x);
             }
             if (vae_tiling) {
+                if (SD_TILE_SIZE != nullptr) {
+                    LOG_INFO("VAE Tile size: %dx%d", tile_size_x, tile_size_y);
+                }
+                if (SD_TILE_OVERLAP != nullptr) {
+                    LOG_INFO("VAE Tile overlap: %.2f", tile_overlap);
+                }
                 // split latent in 32x32 tiles and compute in several steps
                 auto on_tiling = [&](ggml_tensor* in, ggml_tensor* out, bool init) {
                     first_stage_model->compute(n_threads, in, decode, &out);
                 };
-                sd_tiling(x, result, 8, tile_size, 0.5f, on_tiling);
+                sd_tiling_non_square(x, result, 8, tile_size_x, tile_size_y, tile_overlap, on_tiling);
             } else {
                 first_stage_model->compute(n_threads, x, decode, &result);
             }