A high-performance inference engine for AI models
Fix sampling tests
Aleksei Savin4dae7d51fce2
diff --git a/crates/uzu/tests/common/mod.rs b/crates/uzu/tests/common/mod.rsindex e912e4743..78cdbed95 100644--- a/crates/uzu/tests/common/mod.rs+++ b/crates/uzu/tests/common/mod.rs@@ -2,8 +2,7 @@use std::{collections::HashMap, path::PathBuf};-use uzu::VERSION;-use uzu::speculators::speculator::Speculator;+use uzu::{VERSION, speculators::speculator::Speculator};pub const MODEL_DIR_NAME: &str = "Llama-3.2-1B-Instruct";pub const MODEL_FILE_NAME: &str = "model.safetensors";diff --git a/crates/uzu/tests/sampling_test.rs b/crates/uzu/tests/sampling_test.rsindex 63a4770d9..0ce9d1424 100644--- a/crates/uzu/tests/sampling_test.rs+++ b/crates/uzu/tests/sampling_test.rs@@ -107,6 +107,8 @@ fn test_argmax_sampling_with_strategy(strategy: ArgmaxStrategy) {.encode(&logits_buffer,None,+ 0,+ None,&output_buffer,SamplingMethod::Greedy,batch_size,@@ -247,6 +249,8 @@ fn test_topp_sampling_from_prob_exact_match(.encode(&logits_buf,Some(&seeds_buf),+ 0,+ None,&output_buf,SamplingMethod::Stochastic {temperature: None,@@ -374,6 +378,8 @@ fn test_topp_sampling_statistical_large() {.encode(&logits_buf,Some(&seeds_buf),+ 0,+ None,&output_buf,SamplingMethod::Stochastic {temperature: None,@@ -482,6 +488,8 @@ fn perf_topp_128k_vocab() {.encode(&logits_buf,Some(&seeds_buf),+ 0,+ None,&output_buf,SamplingMethod::Stochastic {temperature: None,@@ -587,6 +595,8 @@ fn perf_argmax_128k_vocab_with_strategy(strategy: ArgmaxStrategy) {.encode(&logits_buf,Some(&seeds_buf),+ 0,+ None,&output_buf,SamplingMethod::Greedy,BATCH,@@ -708,6 +718,8 @@ fn test_categorical_sampling() {.encode(&logits_buffer,Some(&seeds_buffer),+ 0,+ None,&output_buffer,SamplingMethod::Stochastic {temperature: None,@@ -851,6 +863,8 @@ fn test_categorical_sampling_statistical() {.encode(&logits_buffer,Some(&seeds_buffer),+ 0,+ None,&output_buffer,SamplingMethod::Stochastic {temperature: None,@@ -957,6 +971,8 @@ fn perf_categorical_128k_vocab() {.encode(&logits_buf,Some(&seeds_buf),+ 0,+ None,&output_buf,SamplingMethod::Stochastic {temperature: None,@@ -1291,6 +1307,7 @@ fn test_gumbel_gpu_cpu_match() {.encode_gumbel(&logits_buffer,&seeds_buffer,+ 0,&gumbel_logits_buffer,BATCH as u32,VOCAB as u32,