From 10bd640aaee876c9d6341ad2ceb43a5024e78161 Mon Sep 17 00:00:00 2001 From: Daniel Bevenius Date: Thu, 4 Dec 2025 08:26:33 +0100 Subject: [PATCH] Revert "sampling : stop short if backend sampler sampled a token" This reverts commit 87b2719eca55b30afff600fc7f61c6cce9452cbf. --- src/llama-graph.cpp | 1 - tests/test-backend-sampler.cpp | 2 -- 2 files changed, 3 deletions(-) diff --git a/src/llama-graph.cpp b/src/llama-graph.cpp index c0ff7d1791..a621c4ebf5 100644 --- a/src/llama-graph.cpp +++ b/src/llama-graph.cpp @@ -2100,7 +2100,6 @@ void llm_graph_context::build_sampling() const { if (data.sampled != nullptr) { res->t_sampled[seq_id] = data.sampled; ggml_build_forward_expand(gf, data.sampled); - continue; } if (data.probs != nullptr) { diff --git a/tests/test-backend-sampler.cpp b/tests/test-backend-sampler.cpp index eb3a0e248d..f56cce6350 100644 --- a/tests/test-backend-sampler.cpp +++ b/tests/test-backend-sampler.cpp @@ -441,8 +441,6 @@ static void test_backend_temp_sampling(const char * model_path) { struct llama_sampler_chain_params backend_chain_params = llama_sampler_chain_default_params(); struct llama_sampler * backend_sampler_chain = llama_sampler_chain_init(backend_chain_params); llama_sampler_chain_add(backend_sampler_chain, llama_sampler_init_temp(temp)); - llama_sampler_chain_add(backend_sampler_chain, llama_sampler_init_top_k(40)); - llama_sampler_chain_add(backend_sampler_chain, llama_sampler_init_dist(18)); std::vector backend_sampler_configs = { { seq_id, backend_sampler_chain },