Smoothing factor backport

kalomaze · Apr 2, 2024 · b5dbcf6 · b5dbcf6
1 parent f87f7b8
commit b5dbcf6
Show file tree

Hide file tree

Showing 5 changed files with 23 additions and 6 deletions.
diff --git a/common/sampling.cpp b/common/sampling.cpp
@@ -132,6 +132,7 @@ static void sampler_queue(
     const float         temp              = params.temp;
     const float         dynatemp_range    = params.dynatemp_range;
     const float         dynatemp_exponent = params.dynatemp_exponent;
+    const float         smoothing_factor  = params.smoothing_factor;
     const int32_t       top_k             = params.top_k;
     const float         top_p             = params.top_p;
     const float         min_p             = params.min_p;
@@ -147,10 +148,10 @@ static void sampler_queue(
             case llama_sampler_type::TOP_P    : llama_sample_top_p    (ctx_main, &cur_p, top_p,     min_keep); break;
             case llama_sampler_type::MIN_P    : llama_sample_min_p    (ctx_main, &cur_p, min_p,     min_keep); break;
             case llama_sampler_type::TEMPERATURE:
-                if (dynatemp_range > 0) {
+                if (dynatemp_range > 0 || smoothing_factor > 0) {
                     float dynatemp_min = std::max(0.0f, temp - dynatemp_range);
                     float dynatemp_max = std::max(0.0f, temp + dynatemp_range);
-                    llama_sample_entropy(ctx_main, &cur_p, dynatemp_min, dynatemp_max, dynatemp_exponent);
+                    llama_sample_entropy(ctx_main, &cur_p, dynatemp_min, dynatemp_max, dynatemp_exponent, smoothing_factor);
                 } else {
                     llama_sample_temp(ctx_main, &cur_p, temp);
                 }

diff --git a/common/sampling.h b/common/sampling.h
@@ -31,6 +31,7 @@ typedef struct llama_sampling_params {
     float       temp                  = 0.80f;    // <= 0.0 to sample greedily, 0.0 to not output probabilities
     float       dynatemp_range        = 0.00f;    // 0.0 = disabled
     float       dynatemp_exponent     = 1.00f;    // controls how entropy maps to temperature in dynamic temperature sampler
+    float       smoothing_factor      = 0.0f;     // controls the quadratic adjustment in smooth sampling
     int32_t     penalty_last_n        = 64;       // last n tokens to penalize (0 = disable penalty, -1 = context size)
     float       penalty_repeat        = 1.00f;    // 1.0 = disabled
     float       penalty_freq          = 0.00f;    // 0.0 = disabled

diff --git a/examples/server/server.cpp b/examples/server/server.cpp
@@ -839,6 +839,7 @@ struct server_context {
         slot.sparams.temp              = json_value(data, "temperature",       default_sparams.temp);
         slot.sparams.dynatemp_range    = json_value(data, "dynatemp_range",    default_sparams.dynatemp_range);
         slot.sparams.dynatemp_exponent = json_value(data, "dynatemp_exponent", default_sparams.dynatemp_exponent);
+        slot.sparams.smoothing_factor  = json_value(data, "smoothing_factor",  default_sparams.smoothing_factor);
         slot.sparams.penalty_last_n    = json_value(data, "repeat_last_n",     default_sparams.penalty_last_n);
         slot.sparams.penalty_repeat    = json_value(data, "repeat_penalty",    default_sparams.penalty_repeat);
         slot.sparams.penalty_freq      = json_value(data, "frequency_penalty", default_sparams.penalty_freq);

diff --git a/llama.cpp b/llama.cpp
@@ -12183,14 +12183,27 @@ void llama_sample_typical(struct llama_context * ctx, llama_token_data_array * c
     }
 }
 
-void llama_sample_entropy(struct llama_context * ctx, llama_token_data_array * candidates_p, float min_temp, float max_temp, float exponent_val) {
+void llama_sample_entropy(struct llama_context* ctx, llama_token_data_array* candidates_p, float min_temp, float max_temp, float exponent_val, float smoothing_factor) {
     const int64_t t_start_sample_us = ggml_time_us();
 
     // no need to do anything if there is only one (or zero) candidates
-    if(candidates_p->size <= 1) {
+    if (candidates_p->size <= 1) {
         return;
     }
 
+    // Apply smoothing if smoothing_factor is > 0. Do not change base implementation otherwise.
+    if (smoothing_factor > 0 && candidates_p->size > 1) {
+        llama_sample_softmax(ctx, candidates_p);
+        float h = candidates_p->data[0].logit; // Find the maximum logit for h to be added after the transformation
+
+        // Apply quadratic transformation using the smoothing_factor
+        for (size_t i = 0; i < candidates_p->size; ++i) {
+            float logit_shifted = candidates_p->data[i].logit - h;
+            candidates_p->data[i].logit = -smoothing_factor * logit_shifted * logit_shifted + h;
+        }
+        llama_sample_softmax(ctx, candidates_p);
+    }
+
     // Calculate maximum possible entropy
     float max_entropy = -logf(1.0f / candidates_p->size);
 

diff --git a/llama.h b/llama.h
@@ -864,13 +864,14 @@ extern "C" {
                            float   p,
                           size_t   min_keep);
 
-    /// @details Dynamic temperature implementation described in the paper https://arxiv.org/abs/2309.02772.
+    /// @details Dynamic temperature implementation + Smooth Sampling implementations wrapped into one function, no research papers available
     LLAMA_API void llama_sample_entropy(
             struct llama_context * ctx,
           llama_token_data_array * candidates_p,
                            float   min_temp,
                            float   max_temp,
-                           float   exponent_val);
+                           float   exponent_val,
+                           float   smoothing_factor);
 
     LLAMA_API void llama_sample_temp(
             struct llama_context * ctx,