feat: adapt res samplers for flow models for eta > 0

Author: wbruna

src/denoiser.hpp

@@ -808,6 +808,18 @@ static std::tuple<float, float, float> get_ancestral_step_flow(float sigma_from,
     return {sigma_down, sigma_up, alpha_scale};
 }
 
+static std::tuple<float, float, float> get_ancestral_step(float sigma_from,
+                                                          float sigma_to,
+                                                          float eta,
+                                                          bool is_flow_denoiser) {
+    if (is_flow_denoiser) {
+        return get_ancestral_step_flow(sigma_from, sigma_to, eta);
+    } else {
+        auto [sigma_down, sigma_up] = get_ancestral_step(sigma_from, sigma_to, eta);
+        return {sigma_down, sigma_up, 1.0f};
+    }
+}
+
 static sd::Tensor<float> sample_euler_ancestral(denoise_cb_t model,
                                                 sd::Tensor<float> x,
                                                 const std::vector<float>& sigmas,
@@ -1148,6 +1160,7 @@ static sd::Tensor<float> sample_res_multistep(denoise_cb_t model,
                                               sd::Tensor<float> x,
                                               const std::vector<float>& sigmas,
                                               std::shared_ptr<RNG> rng,
+                                              bool is_flow_denoiser,
                                               float eta) {
     sd::Tensor<float> old_denoised = x;
     bool have_old_sigma            = false;
@@ -1179,7 +1192,8 @@ static sd::Tensor<float> sample_res_multistep(denoise_cb_t model,
 
         float sigma_from            = sigmas[i];
         float sigma_to              = sigmas[i + 1];
-        auto [sigma_down, sigma_up] = get_ancestral_step(sigma_from, sigma_to, eta);
+
+        auto [sigma_down, sigma_up, alpha_scale] = get_ancestral_step(sigma_from, sigma_to, eta, is_flow_denoiser);
 
         if (sigma_down == 0.0f || !have_old_sigma) {
             x += ((x - denoised) / sigma_from) * (sigma_down - sigma_from);
@@ -1206,7 +1220,10 @@ static sd::Tensor<float> sample_res_multistep(denoise_cb_t model,
             x = sigma_fn(h) * x + h * (b1 * denoised + b2 * old_denoised);
         }
 
-        if (sigmas[i + 1] > 0 && sigma_up > 0.0f) {
+        if (sigma_to > 0.0f && sigma_up > 0.0f) {
+            if (is_flow_denoiser) {
+                x *= alpha_scale;
+            }
             x += sd::Tensor<float>::randn_like(x, rng) * sigma_up;
         }
 
@@ -1221,6 +1238,7 @@ static sd::Tensor<float> sample_res_2s(denoise_cb_t model,
                                        sd::Tensor<float> x,
                                        const std::vector<float>& sigmas,
                                        std::shared_ptr<RNG> rng,
+                                       bool is_flow_denoiser,
                                        float eta) {
     const float c2 = 0.5f;
     auto t_fn      = [](float sigma) -> float { return -logf(sigma); };
@@ -1249,7 +1267,7 @@ static sd::Tensor<float> sample_res_2s(denoise_cb_t model,
         }
         sd::Tensor<float> denoised = std::move(denoised_opt);
 
-        auto [sigma_down, sigma_up] = get_ancestral_step(sigma_from, sigma_to, eta);
+        auto [sigma_down, sigma_up, alpha_scale] = get_ancestral_step(sigma_from, sigma_to, eta, is_flow_denoiser);
 
         sd::Tensor<float> x0 = x;
         if (sigma_down == 0.0f || sigma_from == 0.0f) {
@@ -1278,7 +1296,10 @@ static sd::Tensor<float> sample_res_2s(denoise_cb_t model,
             x                           = x0 + h * (b1 * eps1 + b2 * eps2);
         }
 
-        if (sigmas[i + 1] > 0 && sigma_up > 0.0f) {
+        if (sigma_to > 0.0f && sigma_up > 0.0f) {
+            if (is_flow_denoiser) {
+                x *= alpha_scale;
+            }
             x += sd::Tensor<float>::randn_like(x, rng) * sigma_up;
         }
     }
@@ -1577,9 +1598,9 @@ static sd::Tensor<float> sample_k_diffusion(sample_method_t method,
         case IPNDM_V_SAMPLE_METHOD:
             return sample_ipndm_v(model, std::move(x), sigmas);
         case RES_MULTISTEP_SAMPLE_METHOD:
-            return sample_res_multistep(model, std::move(x), sigmas, rng, eta);
+            return sample_res_multistep(model, std::move(x), sigmas, rng, is_flow_denoiser, eta);
         case RES_2S_SAMPLE_METHOD:
-            return sample_res_2s(model, std::move(x), sigmas, rng, eta);
+            return sample_res_2s(model, std::move(x), sigmas, rng, is_flow_denoiser, eta);
         case ER_SDE_SAMPLE_METHOD:
             return sample_er_sde(model, std::move(x), sigmas, rng, is_flow_denoiser, eta);
         case DDIM_TRAILING_SAMPLE_METHOD: