sampling: add Top-nσ sampler

common/arg.cpp

@@ -912,6 +912,13 @@ common_params_context common_params_parser_init(common_params & params, llama_ex
             params.sampling.min_p = std::stof(value);
         }
     ).set_sparam());
+    add_opt(common_arg(
+        {"--top-nsigma"}, "N",
+        string_format("top-n-sigma sampling (default: %d, -1 = disabled)", params.sampling.top_n_sigma),
+        [](common_params & params, const std::string & value) {
+            params.sampling.top_n_sigma = std::stof(value);
+        }
+    ).set_examples({LLAMA_EXAMPLE_MAIN}).set_sparam());
     add_opt(common_arg(
         {"--xtc-probability"}, "N",
         string_format("xtc probability (default: %.1f, 0.0 = disabled)", (double)params.sampling.xtc_probability),

common/common.h

@@ -134,6 +134,7 @@ struct common_params_sampling {
     int32_t dry_allowed_length = 2;     // tokens extending repetitions beyond this receive penalty
     int32_t dry_penalty_last_n = -1;    // how many tokens to scan for repetitions (0 = disable penalty, -1 = context size)
     int32_t mirostat           = 0;     // 0 = disabled, 1 = mirostat, 2 = mirostat 2.0
+    int32_t top_n_sigma        = -1;    // -1 = disabled
     float   mirostat_tau       = 5.00f; // target entropy
     float   mirostat_eta       = 0.10f; // learning rate
     bool    ignore_eos         = false;

common/sampling.cpp

@@ -134,11 +134,11 @@ std::string common_params_sampling::print() const {
     snprintf(result, sizeof(result),
             "\trepeat_last_n = %d, repeat_penalty = %.3f, frequency_penalty = %.3f, presence_penalty = %.3f\n"
             "\tdry_multiplier = %.3f, dry_base = %.3f, dry_allowed_length = %d, dry_penalty_last_n = %d\n"
-            "\ttop_k = %d, top_p = %.3f, min_p = %.3f, xtc_probability = %.3f, xtc_threshold = %.3f, typical_p = %.3f, temp = %.3f\n"
+            "\ttop_k = %d, top_p = %.3f, min_p = %.3f, xtc_probability = %.3f, xtc_threshold = %.3f, typical_p = %.3f, top_n_sigma = %d, temp = %.3f\n"
             "\tmirostat = %d, mirostat_lr = %.3f, mirostat_ent = %.3f",
             penalty_last_n, penalty_repeat, penalty_freq, penalty_present,
             dry_multiplier, dry_base, dry_allowed_length, dry_penalty_last_n,
-            top_k, top_p, min_p, xtc_probability, xtc_threshold, typ_p, temp,
+            top_k, top_p, min_p, xtc_probability, xtc_threshold, typ_p, top_n_sigma, temp,
             mirostat, mirostat_eta, mirostat_tau);
 
     return std::string(result);
@@ -167,45 +167,50 @@ struct common_sampler * common_sampler_init(const struct llama_model * model, co
                 params.logit_bias.data()));
 
     if (params.mirostat == 0) {
-        for (const auto & cnstr : params.samplers) {
-            switch (cnstr) {
-                case COMMON_SAMPLER_TYPE_DRY:
-                    {
-                        std::vector<const char *> c_breakers;
-                        c_breakers.reserve(params.dry_sequence_breakers.size());
-                        for (const auto & str : params.dry_sequence_breakers) {
-                            c_breakers.push_back(str.c_str());
+        if(params.top_n_sigma >= 0) {
+            llama_sampler_chain_add(result->chain, llama_sampler_init_temp(params.temp));
+            llama_sampler_chain_add(result->chain, llama_sampler_init_top_n_sigma(params.top_n_sigma));
+        } else {
+            for (const auto & cnstr : params.samplers) {
+                switch (cnstr) {
+                    case COMMON_SAMPLER_TYPE_DRY:
+                        {
+                            std::vector<const char *> c_breakers;
+                            c_breakers.reserve(params.dry_sequence_breakers.size());
+                            for (const auto & str : params.dry_sequence_breakers) {
+                                c_breakers.push_back(str.c_str());
+                            }
+
+                            llama_sampler_chain_add(result->chain, llama_sampler_init_dry      (vocab, llama_model_n_ctx_train(model), params.dry_multiplier, params.dry_base, params.dry_allowed_length, params.dry_penalty_last_n, c_breakers.data(), c_breakers.size()));
                         }
-
-                        llama_sampler_chain_add(result->chain, llama_sampler_init_dry      (vocab, llama_model_n_ctx_train(model), params.dry_multiplier, params.dry_base, params.dry_allowed_length, params.dry_penalty_last_n, c_breakers.data(), c_breakers.size()));
-                    }
-                    break;
-                case COMMON_SAMPLER_TYPE_TOP_K:
-                    llama_sampler_chain_add(result->chain, llama_sampler_init_top_k    (params.top_k));
-                    break;
-                case COMMON_SAMPLER_TYPE_TOP_P:
-                    llama_sampler_chain_add(result->chain, llama_sampler_init_top_p    (params.top_p, params.min_keep));
-                    break;
-                case COMMON_SAMPLER_TYPE_MIN_P:
-                    llama_sampler_chain_add(result->chain, llama_sampler_init_min_p    (params.min_p, params.min_keep));
-                    break;
-                case COMMON_SAMPLER_TYPE_XTC:
-                    llama_sampler_chain_add(result->chain, llama_sampler_init_xtc      (params.xtc_probability, params.xtc_threshold, params.min_keep, params.seed));
-                    break;
-                case COMMON_SAMPLER_TYPE_TYPICAL_P:
-                    llama_sampler_chain_add(result->chain, llama_sampler_init_typical  (params.typ_p, params.min_keep));
-                    break;
-                case COMMON_SAMPLER_TYPE_TEMPERATURE:
-                    llama_sampler_chain_add(result->chain, llama_sampler_init_temp_ext (params.temp, params.dynatemp_range, params.dynatemp_exponent));
-                    break;
-                case COMMON_SAMPLER_TYPE_INFILL:
-                    llama_sampler_chain_add(result->chain, llama_sampler_init_infill   (vocab));
-                    break;
-                case COMMON_SAMPLER_TYPE_PENALTIES:
-                    llama_sampler_chain_add(result->chain, llama_sampler_init_penalties(params.penalty_last_n, params.penalty_repeat, params.penalty_freq, params.penalty_present));
-                    break;
-                default:
-                    GGML_ASSERT(false && "unknown sampler type");
+                        break;
+                    case COMMON_SAMPLER_TYPE_TOP_K:
+                        llama_sampler_chain_add(result->chain, llama_sampler_init_top_k    (params.top_k));
+                        break;
+                    case COMMON_SAMPLER_TYPE_TOP_P:
+                        llama_sampler_chain_add(result->chain, llama_sampler_init_top_p    (params.top_p, params.min_keep));
+                        break;
+                    case COMMON_SAMPLER_TYPE_MIN_P:
+                        llama_sampler_chain_add(result->chain, llama_sampler_init_min_p    (params.min_p, params.min_keep));
+                        break;
+                    case COMMON_SAMPLER_TYPE_XTC:
+                        llama_sampler_chain_add(result->chain, llama_sampler_init_xtc      (params.xtc_probability, params.xtc_threshold, params.min_keep, params.seed));
+                        break;
+                    case COMMON_SAMPLER_TYPE_TYPICAL_P:
+                        llama_sampler_chain_add(result->chain, llama_sampler_init_typical  (params.typ_p, params.min_keep));
+                        break;
+                    case COMMON_SAMPLER_TYPE_TEMPERATURE:
+                        llama_sampler_chain_add(result->chain, llama_sampler_init_temp_ext (params.temp, params.dynatemp_range, params.dynatemp_exponent));
+                        break;
+                    case COMMON_SAMPLER_TYPE_INFILL:
+                        llama_sampler_chain_add(result->chain, llama_sampler_init_infill   (vocab));
+                        break;
+                    case COMMON_SAMPLER_TYPE_PENALTIES:
+                        llama_sampler_chain_add(result->chain, llama_sampler_init_penalties(params.penalty_last_n, params.penalty_repeat, params.penalty_freq, params.penalty_present));
+                        break;
+                    default:
+                        GGML_ASSERT(false && "unknown sampler type");
+                }
             }
         }
         llama_sampler_chain_add(result->chain, llama_sampler_init_dist(params.seed));

examples/main/README.md

@@ -265,6 +265,14 @@ Being experimental and unique, XTC is disabled by default. The recommended combi
 
 Example usage: `--xtc-probability 0.5 --xtc-threshold 0.1`
 
+### Top-nσ Sampling
+
+-   `--top-nsigma N`: Limit the next token selection to a subset of tokens with pre-softmax logits that are within n * σ less than the max logit (default: -1, -1 = disabled).
+
+Top-nσ sampling is a text generation method that selects tokens based on a statistical threshold in pre-softmax logits. It works by only sampling from tokens with logits that are within n * σ of the maximum logit. This method helps maintain a stable sampling space regardless of temperature scaling, allowing it to perform well on reasoning tasks even in high temperatures. Without complex probability manipulation, it efficiently filters tokens directly on the pre-softmax logits. A higher value for top-nsigma (e.g., 5) will take more noisy tokens into consideration, while a lower value (e.g., 1) will focous on the more informative region of the sampling space.
+
+Example usage: `--top-nsigma 1`
+
 ### Logit Bias
 
 -   `-l TOKEN_ID(+/-)BIAS, --logit-bias TOKEN_ID(+/-)BIAS`: Modify the likelihood of a token appearing in the generated text completion.

include/llama.h

@@ -1163,6 +1163,9 @@ extern "C" {
     /// @details XTC sampler as described in https://github.com/oobabooga/text-generation-webui/pull/6335
     LLAMA_API struct llama_sampler * llama_sampler_init_xtc        (float   p, float   t,     size_t min_keep, uint32_t seed);
 
+    /// @details Top n sigma sampling as described in academic paper "Top-nσ: Not All Logits Are You Need" https://arxiv.org/pdf/2411.07641
+    LLAMA_API struct llama_sampler * llama_sampler_init_top_n_sigma(int32_t n);
+
     /// @details Mirostat 1.0 algorithm described in the paper https://arxiv.org/abs/2007.14966. Uses tokens instead of words.
     /// @param candidates A vector of `llama_token_data` containing the candidate tokens, their probabilities (p), and log-odds (logit) for the current position in the generated text.
     /// @param tau  The target cross-entropy (or surprise) value you want to achieve for the generated text. A higher value corresponds to more surprising or less predictable text, while a lower value corresponds to less surprising or more predictable text.

src/llama-sampling.cpp

@@ -1648,6 +1648,73 @@ struct llama_sampler * llama_sampler_init_penalties(
     };
 }
 
+// top-n-sigma
+
+struct llama_sampler_top_n_sigma {
+    const int32_t n;
+};
+
+static const char * llama_sampler_top_n_sigma_name(const struct llama_sampler * /*smpl*/) {
+    return "top-n-sigma";
+}
+
+static void llama_sampler_top_n_sigma_apply(struct llama_sampler * smpl, llama_token_data_array * cur_p) {
+    const auto * ctx = (llama_sampler_top_n_sigma *) smpl->ctx;
+
+    // find max logit and calculate mean
+    float max = cur_p->data[0].logit;
+    float logits_sum = 0;
+    for (size_t i = 0; i < cur_p->size; ++i) {
+        if(cur_p->data[i].logit > max){
+            max = cur_p->data[i].logit;
+        }
+        logits_sum += cur_p->data[i].logit;
+    }
+    float mean = logits_sum/cur_p->size;
+
+    // calculate standard deviation
+    float acc = 0;
+    for(size_t i = 0; i < cur_p->size; ++i){
+        acc += pow(cur_p->data[i].logit - mean, 2);
+    }
+    float std = sqrt(acc/cur_p->size);
+
+    //apply mask
+    for(size_t i = 0; i < cur_p->size; ++i){
+        if(cur_p->data[i].logit < max - ctx->n * std) {
+            cur_p->data[i].logit = -INFINITY;
+        }
+    }
+    llama_sampler_softmax_impl(cur_p);
+}
+
+static struct llama_sampler * llama_sampler_top_n_sigma_clone(const struct llama_sampler * smpl){
+    const auto * ctx = (const llama_sampler_top_n_sigma *) smpl->ctx;
+    return llama_sampler_init_top_n_sigma(ctx->n);
+}
+
+static void llama_sampler_top_n_sigma_free(struct llama_sampler * smpl) {
+    delete (llama_sampler_top_n_sigma *) smpl->ctx;
+}
+
+static struct llama_sampler_i llama_sampler_top_n_sigma_i = {
+    /* .name   = */ llama_sampler_top_n_sigma_name,
+    /* .accept = */ nullptr,
+    /* .apply  = */ llama_sampler_top_n_sigma_apply,
+    /* .reset  = */ nullptr,
+    /* .clone  = */ llama_sampler_top_n_sigma_clone,
+    /* .free   = */ llama_sampler_top_n_sigma_free,
+};
+
+struct llama_sampler * llama_sampler_init_top_n_sigma(int32_t n) {
+    return new llama_sampler {
+        /* .iface = */ &llama_sampler_top_n_sigma_i,
+        /* .ctx   = */ new llama_sampler_top_n_sigma {
+            /* .n = */ n,
+                        },
+    };
+}
+
 // DRY
 
 struct llama_sampler_dry {

tests/test-sampling.cpp

@@ -182,6 +182,17 @@ static void test_dry(
     tester.check();
 }
 
+static void test_top_n_sigma(const std::vector<float> & probs, const std::vector<float> & probs_expected, int n) {
+    sampler_tester tester(probs, probs_expected);
+
+    DUMP(&tester.cur_p);
+    tester.apply(llama_sampler_init_top_n_sigma(n));
+    tester.apply(llama_sampler_init_dist (0));
+    DUMP(&tester.cur_p);
+
+    tester.check();
+}
+
 static void test_sampler_queue(const size_t n_vocab, const std::string & samplers_sequence, const int top_k, const float top_p, const float min_p
 ) {
     sampler_tester tester(n_vocab);
@@ -349,6 +360,10 @@ int main(void) {
     test_dry({0.2f, 0.2f, 0.2f, 0.2f, 0.2f}, {0, 1, 2, 0, 1}, {0.241818f, 0.241818f, 0.241818f, 0.241818f, 0.032727f}, 2.0f, 1.1f, 2, 5, {});
     test_dry({0.2f, 0.2f, 0.2f, 0.2f, 0.2f}, {0, 1, 2, 3, 4, 0, 1}, {0.2f, 0.2f, 0.2f, 0.2f, 0.2f}, 1.0f, 1.1f, 4, 7, {});
 
+    test_top_n_sigma({0.1f, 0.2f, 0.3f, 0.4f}, {0.571429f, 0.428571f, 0.0f, 0.0f}, 1);
+    test_top_n_sigma({0.1f, 0.2f, 0.3f, 0.4f}, {1.0f, 0.0f, 0.0f, 0.0f}, 0);
+    test_top_n_sigma({0.1f, 0.2f, 0.3f, 0.4f}, {0.4f, 0.3f, 0.2f, 0.1f}, 3);
+
     test_sampler_queue(10000, "k", 10000, 1.0f, 1.0f);
     test_sampler_queue(10000, "k",     1, 1.0f, 1.0f);
     test_sampler_queue(10000, "p", 10000, 1.0f, 1.0f);