Add prompt batching to Gemma.cpp.

This CL adds a new function to Gemma that allows for batching of multiple prompts. The function takes a vector of prompts and returns a vector of responses. The prompts are processed in parallel, and the responses are returned in the same order as the prompts.

PiperOrigin-RevId: 648367559

BUILD.bazel

@@ -103,6 +103,10 @@ cc_library(
         "gemma/activations.h",
         "gemma/gemma.h",
     ],
+    exec_properties = {
+        # Avoid linker OOMs when building with sanitizer instrumentation.
+        "mem": "28g",
+    },
     textual_hdrs = [
         # Placeholder for internal file1, do not remove,
         # Placeholder for internal file2, do not remove,
@@ -194,6 +198,7 @@ cc_test(
         ":ops",
         "@googletest//:gtest_main",
         "//compression:io",
+        "@hwy//:hwy",
         "@hwy//:hwy_test_util",
         "@hwy//:thread_pool",
     ],
@@ -370,6 +375,10 @@ cc_test(
         "backprop/backward_test.cc",
         "backprop/test_util.h",
     ],
+    exec_properties = {
+        # Avoid linker OOMs when building with sanitizer instrumentation.
+        "mem": "28g",
+    },
     deps = [
         ":backprop",
         ":backprop_scalar",
@@ -406,6 +415,10 @@ cc_test(
     srcs = [
         "backprop/optimize_test.cc",
     ],
+    exec_properties = {
+        # Avoid linker OOMs when building with sanitizer instrumentation.
+        "mem": "28g",
+    },
     deps = [
         ":backprop",
         ":common",

gemma/benchmark_helper.cc

@@ -18,6 +18,8 @@
 #include <stdio.h>
 #include <time.h>
 
+#include <algorithm>
+#include <cstdio>
 #include <iostream>
 #include <memory>
 #include <ostream>
@@ -34,6 +36,7 @@
 #include "gemma/gemma.h"
 #include "util/app.h"
 #include "util/args.h"
+#include "hwy/aligned_allocator.h"
 #include "hwy/base.h"
 #include "hwy/contrib/thread_pool/thread_pool.h"
 #include "hwy/highway.h"
@@ -72,7 +75,11 @@ GemmaEnv::GemmaEnv(const LoaderArgs& loader, const InferenceArgs& inference,
   } else {
     fprintf(stderr, "Loading model...\n");
     model_ = AllocateGemma(loader_, pool_);
-    kv_cache_ = KVCache::Create(loader_.ModelType());
+
+    kv_caches_.reserve(16);
+    for (int i = 0; i < 16; ++i) {
+      kv_caches_.push_back(new KVCache(KVCache::Create(loader_.ModelType())));
+    }
   }
 
   InitGenerator(inference_args_, gen_);
@@ -107,8 +114,9 @@ std::pair<std::string, size_t> GemmaEnv::QueryModel(
   size_t total_tokens = 0;
 
   const double time_start = hwy::platform::Now();
-  const StreamFunc stream_token = [&res, &total_tokens, &time_start, this](
-                                      int token, float) {
+  const BatchStreamFunc batch_stream_token =
+      [&res, &total_tokens, &time_start, this](
+          size_t query_index, size_t pos, int token, float) {
     ++total_tokens;
     res += StringFromTokens(std::vector<int>{token});
     if (app_.verbosity >= 1 && total_tokens % 128 == 0) {
@@ -123,8 +131,8 @@ std::pair<std::string, size_t> GemmaEnv::QueryModel(
               << "\ttemperature: " << inference_args_.temperature << "\n";
   }
   gcpp::TimingInfo timing_info;
-  runtime_config_.stream_token = stream_token;
-  model_->Generate(runtime_config_, tokens, /*start_pos=*/0, kv_cache_,
+  runtime_config_.batch_stream_token = batch_stream_token;
+  model_->Generate(runtime_config_, tokens, /*start_pos=*/0, *kv_caches_[0],
                    timing_info);
   if (app_.verbosity >= 1) {
     LogSpeedStats(time_start, total_tokens);
@@ -132,12 +140,73 @@ std::pair<std::string, size_t> GemmaEnv::QueryModel(
   return {res, total_tokens};
 }
 
+std::vector<std::pair<std::string, size_t>> GemmaEnv::BatchQueryModel2(
+    const hwy::Span<const hwy::Span<int>>& prompts) {
+  std::vector<std::pair<std::string, size_t>> res(prompts.size());
+  std::fill(res.begin(), res.end(), std::make_pair("", 0));
+  size_t total_tokens = 0;
+
+  const double time_start = hwy::platform::Now();
+  const BatchStreamFunc batch_stream_token =
+      [&res, &total_tokens, &time_start, this](
+          size_t query_index, size_t pos, int token, float) {
+    std::string token_text;
+    HWY_ASSERT(
+        model_->Tokenizer().Decode(std::vector<int>{token}, &token_text));
+    // fprintf(stderr, "Query %zu returned token \"%s\"\n\n", query_index,
+    //         token_text.c_str());
+    std::string single_res = res[query_index].first + token_text;
+    size_t current_token_count = res[query_index].second + 1;
+    res[query_index] = std::make_pair(single_res, current_token_count);
+
+    ++total_tokens;
+    if (app_.verbosity >= 1 && total_tokens % 128 == 0) {
+      LogSpeedStats(time_start, total_tokens);
+    }
+    return true;
+  };
+  if (app_.verbosity >= 2) {
+    std::cout << inference_args_.max_tokens << " "
+              << inference_args_.max_generated_tokens << " "
+              << inference_args_.temperature;
+  }
+  gcpp::TimingInfo timing_info;
+  runtime_config_.batch_stream_token = batch_stream_token;
+  model_->GenerateBatch(runtime_config_, prompts, /*start_pos=*/0, kv_caches_,
+                        timing_info);
+  if (app_.verbosity >= 1) {
+    LogSpeedStats(time_start, total_tokens);
+  }
+  return res;
+}
+
 std::pair<std::string, size_t> GemmaEnv::QueryModel(std::string& input) {
   const std::vector<int> prompt =
       WrapAndTokenize(model_->Tokenizer(), loader_.ModelTrainingType(),
                       /*pos=*/0, input);
   return QueryModel(prompt);
 }
+std::vector<std::pair<std::string, size_t>> GemmaEnv::BatchQueryModel(
+    const std::vector<std::string>& inputs) {
+  std::vector<std::unique_ptr<std::vector<int>>> prompts;
+  prompts.reserve(inputs.size());
+  for (auto& input : inputs) {
+    std::string mutable_prompt = input;
+    prompts.push_back(std::make_unique<std::vector<int>>(
+        WrapAndTokenize(model_->Tokenizer(),
+                                       loader_.ModelTrainingType(),
+                                       /*pos=*/0, mutable_prompt)));
+  }
+  std::vector<hwy::Span<int>> prompt_vector;
+  prompt_vector.reserve(prompts.size());
+  for (auto& prompt : prompts) {
+    prompt_vector.push_back(hwy::Span<int>(
+        prompt->data(), prompt->size()));
+  }
+  hwy::Span<const hwy::Span<int>> prompt_span = hwy::Span<const hwy::Span<int>>(
+      prompt_vector.data(), prompt_vector.size());
+  return BatchQueryModel2(prompt_span);
+}
 
 float GemmaEnv::CrossEntropy(const std::string& input) {
   std::vector<int> prompt = Tokenize(input);

gemma/benchmark_helper.h

@@ -69,8 +69,12 @@ class GemmaEnv {
   // Runs inference on the given input and returns the top-1 result string and
   // the number of tokens that were generated.
   std::pair<std::string, size_t> QueryModel(const std::vector<int>& tokens);
+  std::vector<std::pair<std::string, size_t>> BatchQueryModel2(
+      const hwy::Span<const hwy::Span<int>>& prompts);
   // Adds turn structure to input, tokenizes and calls the above overload.
   std::pair<std::string, size_t> QueryModel(std::string& input);
+  std::vector<std::pair<std::string, size_t>> BatchQueryModel(
+      const std::vector<std::string>& inputs);
 
   // Runs inference on the given input and returns the cross entropy, a measure
   // of how well the model predicts the correct output. It is the average
@@ -87,7 +91,7 @@ class GemmaEnv {
   RuntimeConfig& MutableConfig() { return runtime_config_; }
   InferenceArgs& MutableInferenceArgs() { return inference_args_; }
   std::mt19937& MutableGen() { return gen_; }
-  KVCache& MutableKVCache() { return kv_cache_; }
+  KVCache& MutableKVCache() { return *kv_caches_[0]; }
 
  private:
   // Arguments to the model loader: file locations, etc.
@@ -103,7 +107,7 @@ class GemmaEnv {
   // The model to run inference on.
   std::unique_ptr<Gemma> model_;
   // The KV cache to use for inference.
-  KVCache kv_cache_;
+  std::vector<KVCache*> kv_caches_;
   RuntimeConfig runtime_config_;
 };
 

gemma/gemma.cc

@@ -75,23 +75,30 @@ struct Activations {
   static constexpr size_t kQStride = kQKVDim * (kIsMHA ? 3 : 1);
 
   std::array<float, kBatchSize * kModelDim> x;  // input
-
   std::array<float, kBatchSize * kModelDim> pre_att_rms_out;
-  std::array<float, kBatchSize * kHeads * kQStride> q;  // query vector
+  std::array<float, kBatchSize * kHeads * kQStride>
+      q;  // query vector
   std::array<float, kBatchSize * kHeads * TConfig::kSeqLen>
-      att;                                                   // attention vector
-  std::array<float, kBatchSize * kHeads * kQKVDim> att_out;  // attention output
+      att;  // attention vector
+  std::array<float, kBatchSize * kHeads * kQKVDim>
+      att_out;  // attention output
   std::array<float, kHeads * kBatchSize * kModelDim>
       att_post1;  // attention output after linear transformation, per head
   std::array<float, kBatchSize * kModelDim>
       att_post2;  // accumulation of attention outputs over heads
+  std::array<hwy::bfloat16_t, kBatchSize * kModelDim>
+      bf_pre_ffw_rms_out;
+  std::array<float, kBatchSize * TConfig::kFFHiddenDim * 2>
+      ffw_hidden;
 
-  std::array<hwy::bfloat16_t, kBatchSize * kModelDim> bf_pre_ffw_rms_out;
-  std::array<float, kBatchSize * TConfig::kFFHiddenDim * 2> ffw_hidden;
-  std::array<float, kBatchSize * TConfig::kFFHiddenDim> C1;  // MatMul output
+  // For FFW MatMul.
+  std::array<float, kBatchSize * TConfig::kFFHiddenDim> C1;
   std::array<float, kBatchSize * TConfig::kFFHiddenDim> C2;
-  std::array<float, kBatchSize * kModelDim> ffw_out;
 
+  // bf_ version can't be used until GeluMulToBF16 issue in FFW() is resolved.
+  // std::array<hwy::bfloat16_t, kBatchSize * 2 * TConfig::kFFHiddenDim>
+  //     bf_ffw_hidden;
+  std::array<float, kBatchSize * kModelDim> ffw_out;
   std::array<float, kBatchSize * TConfig::kVocabSize> logits;
 
   // For bf16/f32 vectors * bf16 matrix: faster to unpack once beforehand, into
@@ -104,7 +111,8 @@ struct Activations {
   std::array<float, kBatchSize * kGriffinDim> griffin_x;
   std::array<float, kBatchSize * kGriffinDim> griffin_y;
   std::array<float, kBatchSize * kGriffinDim> griffin_gate_x;
-  std::array<float, kBatchSize * kGriffinDim> griffin_multiplier;
+  std::array<float, kBatchSize * kGriffinDim>
+      griffin_multiplier;
 };
 
 namespace {
@@ -116,10 +124,12 @@ struct CreateKVCache {
 
     const size_t size_cache_pos = CachePosSize<TConfig>()();
     if (size_cache_pos != 0) {
-      const size_t seq_len = TConfig::kSeqLen + kPrefillBatchSize;
+      const size_t seq_len =
+          (TConfig::kSeqLen + kPrefillBatchSize);
       kv_cache.kv_cache = hwy::AllocateAligned<float>(seq_len * size_cache_pos);
     }
 
+    // TODO(patrickms): Add query batching support for Griffin.
     if (TConfig::kGriffinLayers) {
       constexpr size_t kConv1dWidth = TConfig::kConv1dWidth;
       const size_t conv1d_cache_size =
@@ -226,19 +236,24 @@ namespace gcpp {
 namespace HWY_NAMESPACE {
 namespace {
 
-template <class TConfig, size_t kBatchSize>
+template <class TConfig, size_t kBatchSize, size_t kQueryBatchSize>
 HWY_NOINLINE void GriffinRecurrent(
-    size_t batch_start, size_t num_tokens, size_t layer,
-    Activations<TConfig, kBatchSize>& activations,
-    const CompressedLayer<TConfig>* layer_weights, KVCache& kv_cache,
-    hwy::ThreadPool& pool) {
+    size_t batch_start, size_t num_tokens, size_t num_queries, size_t layer,
+    Activations<TConfig, kBatchSize * kQueryBatchSize>& activations,
+    const CompressedLayer<TConfig>* layer_weights,
+    const std::vector<KVCache*>& kv_caches, hwy::ThreadPool& pool) {
   PROFILER_ZONE("Gen.Griffin");
+  static_assert(kQueryBatchSize == 1,
+                "Griffin does not support batched queries.");
+  HWY_DASSERT(num_queries == 1);  // TODO: add batch query support for Griffin.
+  KVCache& kv_cache = *kv_caches[0];
   namespace hn = hwy::HWY_NAMESPACE;
   using D = hn::ScalableTag<float>;
   HWY_DASSERT(num_tokens <= kBatchSize);
-  constexpr size_t kModelDim = Activations<TConfig, kBatchSize>::kModelDim;
-  constexpr size_t kConv1dWidth = TConfig::kConv1dWidth;
-  constexpr size_t kHeads = TConfig::kHeads;
+  static constexpr size_t kModelDim =
+      gcpp::Activations<TConfig, kBatchSize * kQueryBatchSize>::kModelDim;
+  static constexpr size_t kConv1dWidth = TConfig::kConv1dWidth;
+  static constexpr size_t kHeads = TConfig::kHeads;
 
   // X / Y linear layers.
   for (size_t batch_idx = 0; batch_idx < num_tokens; ++batch_idx) {
@@ -357,14 +372,19 @@ HWY_NOINLINE void GriffinRecurrent(
   }
 }
 
-template <class TConfig, size_t kBatchSize>
-HWY_NOINLINE void Attention(size_t batch_start, size_t num_tokens, size_t layer,
-                            Activations<TConfig, kBatchSize>& activations,
-                            const CompressedLayer<TConfig>* layer_weights,
-                            KVCache& kv_cache, hwy::ThreadPool& pool) {
+template <class TConfig, size_t kBatchSize, size_t kQueryBatchSize>
+HWY_NOINLINE void Attention(
+    size_t batch_and_query_start, size_t num_tokens, size_t num_queries,
+    size_t layer,
+    Activations<TConfig, kBatchSize * kQueryBatchSize>& activations,
+    const CompressedLayer<TConfig>* layer_weights,
+    const std::vector<KVCache*>& kv_caches,
+    hwy::ThreadPool& pool) {
   PROFILER_ZONE("Gen.Attention");
   HWY_DASSERT(num_tokens <= kBatchSize);
-  using TActivations = Activations<TConfig, kBatchSize>;
+  HWY_DASSERT(num_queries <= kQueryBatchSize);
+  HWY_DASSERT(batch_and_query_start % num_queries == 0);
+  using TActivations = Activations<TConfig, kBatchSize * kQueryBatchSize>;
   constexpr size_t kQKVDim = TActivations::kQKVDim;
   constexpr size_t kQStride = TActivations::kQStride;
   constexpr size_t kCachePosSize = CachePosSize<TConfig>()();
@@ -376,15 +396,22 @@ HWY_NOINLINE void Attention(size_t batch_start, size_t num_tokens, size_t layer,
   GEMMA_CONSTEXPR_SQRT const float kQueryScale =
       1.0f / Sqrt(static_cast<float>(kQKVDim));
   constexpr bool kIsMHA = TActivations::kIsMHA;  // Multi-Head Attention
+  const size_t batch_start = batch_and_query_start / num_queries;
+  const size_t num_tokens_and_queries = num_tokens * num_queries;
 
   // If MHA, this also computes KV, which we copy to the KV cache below.
   static_assert(!kIsMHA || TConfig::kInterleaveQKV);  // MHA => interleaved
   MatMul_4x4_Batch<kModelDim, kHeads * kQStride>(
-      num_tokens, activations.pre_att_rms_out.data(),
+      num_tokens_and_queries, activations.pre_att_rms_out.data(),
       layer_weights->qkv_einsum_w.data(), activations.q.data(), pool);
 
-  for (size_t batch_idx = 0; batch_idx < num_tokens; ++batch_idx) {
-    const float* x = activations.pre_att_rms_out.data() + batch_idx * kModelDim;
+  for (size_t batch_and_query_idx = 0;
+       batch_and_query_idx < num_tokens_and_queries; ++batch_and_query_idx) {
+    const float* x = activations.pre_att_rms_out.data() + batch_and_query_idx
+                     * kModelDim;
+    const size_t query_idx = batch_and_query_idx % num_queries;
+    const size_t batch_idx = batch_and_query_idx / num_queries;
+    KVCache& kv_cache = *kv_caches[query_idx];
     // QKV projections:
     if constexpr (!kIsMHA) {
       const size_t pos = batch_start + batch_idx;
@@ -401,18 +428,23 @@ HWY_NOINLINE void Attention(size_t batch_start, size_t num_tokens, size_t layer,
 
   // Positional encodings for kv:
   pool.Run(
-      0, kKVHeads * num_tokens, [&](uint64_t task, size_t thread) HWY_ATTR {
+      0, kKVHeads * num_tokens_and_queries,
+      [&](uint64_t task, size_t thread) HWY_ATTR {
         const size_t head = task % kKVHeads;
-        const size_t batch_idx = task / kKVHeads;
+        const size_t batch_and_query_idx = task / kKVHeads;
+        const size_t query_idx = batch_and_query_idx % num_queries;
+        const size_t batch_idx = batch_and_query_idx / num_queries;
         const size_t pos = batch_start + batch_idx;
         const size_t cache_pos = pos % (kSeqLen + kPrefillBatchSize);
         const size_t kv_offset = cache_pos * kCachePosSize +
                                  layer * kCacheLayerSize + head * kQKVDim * 2;
+        KVCache& kv_cache = *kv_caches[query_idx];
         float* HWY_RESTRICT kv = kv_cache.kv_cache.get() + kv_offset;
         if constexpr (kIsMHA) {
           // For MHA, copy kv into the KV cache from scratch space (see above).
           const float* HWY_RESTRICT q =
-              activations.q.data() + (batch_idx * kHeads + head) * kQStride;
+              activations.q.data() + (batch_and_query_idx * kHeads
+                                      + head) * kQStride;
           // Skip past the Q part of `q`, and copy KV to `kv`.
           memcpy(kv, q + kQKVDim, 2 * kQKVDim * sizeof(float));
         }
@@ -422,17 +454,22 @@ HWY_NOINLINE void Attention(size_t batch_start, size_t num_tokens, size_t layer,
   static_assert((kHeads % kKVHeads) == 0,
                 "query heads must be a multiple of key-value heads");
   constexpr size_t kGroupHeads = kHeads / kKVHeads;
-  pool.Run(0, kHeads * num_tokens, [&](uint64_t task, size_t thread) HWY_ATTR {
+  pool.Run(0, kHeads * num_tokens_and_queries,
+           [&](uint64_t task, size_t thread) HWY_ATTR {
     const size_t head = task % kHeads;
-    const size_t batch_idx = task / kHeads;
+    const size_t batch_and_query_idx = task / kHeads;
+    const size_t query_idx = batch_and_query_idx % num_queries;
+    const size_t batch_idx = batch_and_query_idx / num_queries;
     const size_t head_offset = (head / kGroupHeads) * kQKVDim * 2;
+    KVCache& kv_cache = *kv_caches[query_idx];
     float* HWY_RESTRICT q =
-        activations.q.data() + (batch_idx * kHeads + head) * kQStride;
+        activations.q.data() + (batch_and_query_idx * kHeads + head) * kQStride;
 
     const size_t pos = batch_start + batch_idx;
     // Calculate scores
     float* HWY_RESTRICT head_att =
-        activations.att.data() + head * kSeqLen + batch_idx * kHeads * kSeqLen;
+        activations.att.data() + head * kSeqLen
+        + batch_and_query_idx * kHeads * kSeqLen;
 
     Rope(q, TConfig::kUseHalfRope ? kQKVDim / 2 : kQKVDim, pos);
     MulByConst(kQueryScale, q, kQKVDim);
@@ -451,7 +488,7 @@ HWY_NOINLINE void Attention(size_t batch_start, size_t num_tokens, size_t layer,
 
     // Weighted summation
     float* HWY_RESTRICT att_out = activations.att_out.data() + head * kQKVDim +
-                                  batch_idx * kHeads * kQKVDim;
+                                  batch_and_query_idx * kHeads * kQKVDim;
     hwy::ZeroBytes(att_out, kQKVDim * sizeof(*att_out));
     for (size_t pos2 = start_pos; pos2 <= pos; ++pos2) {
       const size_t cache_pos = pos2 % (kSeqLen + kPrefillBatchSize);
@@ -462,20 +499,23 @@ HWY_NOINLINE void Attention(size_t batch_start, size_t num_tokens, size_t layer,
     }
   });
 
-  for (size_t batch_idx = 0; batch_idx < num_tokens; ++batch_idx) {
+  for (size_t batch_and_query_idx = 0;
+       batch_and_query_idx < num_tokens_and_queries; ++batch_and_query_idx) {
     // TODO(szabadka) Use a single MatVecAdd like in GriffinRecurrent() after
     // rearranging the weights.
     float* HWY_RESTRICT att_out =
-        activations.att_out.data() + batch_idx * kHeads * kQKVDim;
+        activations.att_out.data() + batch_and_query_idx * kHeads * kQKVDim;
     float* HWY_RESTRICT layer_out =
-        activations.att_post2.data() + batch_idx * kModelDim;
+        activations.att_post2.data() + batch_and_query_idx * kModelDim;
     MatVecT</*kAdd=*/TConfig::kSoftmaxAttnOutputBiases, kModelDim, kQKVDim>(
         layer_weights->attn_vec_einsum_w, 0, att_out,
         layer_weights->attention_output_biases.data(),
         activations.even_odd.data(), layer_out, pool);
     for (size_t head = 1; head < kHeads; ++head) {
+      // TODO(patrickms): Check this calculation
       float* HWY_RESTRICT head_out =
-          activations.att_post1.data() + head * kBatchSize * kModelDim;
+          activations.att_post1.data() +
+          head * kBatchSize * kQueryBatchSize * kModelDim;
       // TODO: requires MatMul support for offsets.
       MatVec<kModelDim, kQKVDim>(
           layer_weights->attn_vec_einsum_w, head * kModelDim * kQKVDim,
@@ -583,103 +623,138 @@ HWY_NOINLINE void EmbedToken(int token, size_t token_idx, size_t pos,
   };
 }
 
-template <class TConfig, size_t kBatchSize>
+template <class TConfig, size_t kBatchSize, size_t kQueryBatchSize>
 HWY_NOINLINE void TransformerLayer(
-    size_t num_tokens, size_t pos, size_t layer,
+    size_t num_tokens, size_t num_queries, size_t pos, size_t layer,
     const CompressedLayer<TConfig>* layer_weights,
-    Activations<TConfig, kBatchSize>& activations, KVCache& kv_cache,
-    hwy::ThreadPool& pool) {
+    Activations<TConfig, kBatchSize * kQueryBatchSize>& activations,
+    const std::vector<KVCache*>& kv_caches, hwy::ThreadPool& pool) {
   constexpr size_t kModelDim = TConfig::kModelDim;
+  const size_t num_tokens_and_queries = num_tokens * num_queries;
   auto type = TConfig::kLayerConfig[layer];
   size_t layer_of_type =
       NumLayersOfTypeBefore(TConfig::kLayerConfig, type, layer);
-  RMSNormBatched<kBatchSize>(num_tokens, activations.x.data(),
-                             layer_weights->pre_attention_norm_scale.data(),
-                             activations.pre_att_rms_out.data(), kModelDim);
+  RMSNormBatched<kBatchSize * kQueryBatchSize>(
+      num_tokens_and_queries, activations.x.data(),
+      layer_weights->pre_attention_norm_scale.data(),
+      activations.pre_att_rms_out.data(), kModelDim);
   if (type == LayerAttentionType::kGemma) {
-    Attention(pos, num_tokens, layer_of_type, activations, layer_weights,
-              kv_cache, pool);
+    Attention<TConfig, kBatchSize, kQueryBatchSize>(
+        pos, num_tokens, num_queries, layer_of_type, activations,
+              layer_weights, kv_caches, pool);
   } else {
-    GriffinRecurrent(pos, num_tokens, layer_of_type, activations, layer_weights,
-                     kv_cache, pool);
+    // This Griffin layers should never exist unless the model is a Griffin
+    // model. This conditional prevents the compiler from generating code for
+    // this branch when building a non-Griffin model, since we have static
+    // asserts about the query batch size for Griffin layers.
+    if constexpr (TConfig::kGriffinLayers > 0) {
+      GriffinRecurrent<TConfig, kBatchSize, kQueryBatchSize>(
+          pos, num_tokens, num_queries, layer_of_type, activations,
+                      layer_weights, kv_caches, pool);
+    }
   }
   if (TConfig::kPostNormScale) {
-    RMSNormInplaceBatched<kBatchSize>(
-        num_tokens, layer_weights->post_attention_norm_scale.data(),
+    RMSNormInplaceBatched<kBatchSize * kQueryBatchSize>(
+        num_tokens_and_queries,
+        layer_weights->post_attention_norm_scale.data(),
         activations.att_post2.data(), kModelDim);
   }
-  AddFromBatched<kBatchSize>(num_tokens, activations.att_post2.data(),
-                             activations.x.data(), kModelDim);
-  RMSNormBatched<kBatchSize>(num_tokens, activations.x.data(),
-                             layer_weights->pre_ffw_norm_scale.data(),
-                             activations.bf_pre_ffw_rms_out.data(), kModelDim);
-  FFW(activations, num_tokens, layer_weights, pool);
+  AddFromBatched<kBatchSize * kQueryBatchSize>(num_tokens_and_queries,
+                                               activations.att_post2.data(),
+                                               activations.x.data(), kModelDim);
+  RMSNormBatched<kBatchSize * kQueryBatchSize>(
+      num_tokens_and_queries, activations.x.data(),
+      layer_weights->pre_ffw_norm_scale.data(),
+      activations.bf_pre_ffw_rms_out.data(), kModelDim);
+  FFW<TConfig, kBatchSize * kQueryBatchSize>(
+      activations, num_tokens_and_queries, layer_weights, pool);
   if (TConfig::kPostNormScale) {
-    RMSNormInplaceBatched<kBatchSize>(num_tokens,
-                                      layer_weights->post_ffw_norm_scale.data(),
-                                      activations.ffw_out.data(), kModelDim);
+    RMSNormInplaceBatched<kBatchSize * kQueryBatchSize>(
+        num_tokens_and_queries, layer_weights->post_ffw_norm_scale.data(),
+        activations.ffw_out.data(), kModelDim);
   }
-  AddFromBatched<kBatchSize>(num_tokens, activations.ffw_out.data(),
-                             activations.x.data(), kModelDim);
+  AddFromBatched<kBatchSize * kQueryBatchSize>(
+      num_tokens_and_queries, activations.ffw_out.data(),
+      activations.x.data(), kModelDim);
 }
 
-template <class TConfig, size_t kBatchSize>
-HWY_NOINLINE void Prefill(const int* tokens, size_t num_tokens, size_t pos,
-                          const CompressedWeights<TConfig>& weights,
-                          Activations<TConfig, kBatchSize>& activations,
-                          KVCache& kv_cache, hwy::ThreadPool& pool) {
+template <class TConfig, size_t kBatchSize, size_t kQueryBatchSize>
+HWY_NOINLINE void Prefill(
+    const int* tokens, size_t num_tokens, size_t num_queries, size_t pos,
+    const CompressedWeights<TConfig>& weights,
+    Activations<TConfig, kBatchSize * kQueryBatchSize>& activations,
+    const std::vector<KVCache*>& kv_caches, hwy::ThreadPool& pool) {
+  HWY_DASSERT(num_queries <= kQueryBatchSize);
+  const size_t minibatch_size = std::min(num_tokens, kBatchSize);
   PROFILER_ZONE("Gen.Prefill\\Att\\FFW");
+  // TODO(patrickms): Try to hoist pool.Run out of the loop.
+  for (size_t i = 0; i < num_tokens; i += minibatch_size) {
+    const size_t offset = i * num_queries;
+    const size_t current_token_count = std::min(
+        minibatch_size, num_tokens - i);
+    pool.Run(0, current_token_count * num_queries,
+            [&](const uint64_t token_idx, size_t /*thread*/) HWY_ATTR {
+              EmbedToken<TConfig, kBatchSize * kQueryBatchSize>(
+                  tokens[token_idx + offset], token_idx, pos + offset,
+                  weights, activations);
+            });
 
-  pool.Run(
-      0, num_tokens, [&](const uint64_t token_idx, size_t /*thread*/) HWY_ATTR {
-        EmbedToken(tokens[token_idx], token_idx, pos, weights, activations);
-      });
-
-  for (size_t layer = 0; layer < TConfig::kLayers; ++layer) {
-    const auto* layer_weights = weights.GetLayer(layer);
-    TransformerLayer(num_tokens, pos, layer, layer_weights, activations,
-                     kv_cache, pool);
+    for (size_t layer = 0; layer < TConfig::kLayers; ++layer) {
+      const auto* layer_weights = weights.GetLayer(layer);
+      TransformerLayer<TConfig, kBatchSize, kQueryBatchSize>(
+          current_token_count, num_queries, pos + offset , layer, layer_weights,
+          activations, kv_caches, pool);
+    }
   }
 }
 
 // Compute the transformer for a batch of input tokens. During generation,
 // we usually have num_tokens == 1 (and also kBatchSize == 1).
-template <class TConfig, size_t kBatchSize>
-HWY_NOINLINE void Transformer(const int* tokens, size_t num_tokens, size_t pos,
-                              const CompressedWeights<TConfig>& weights,
-                              Activations<TConfig, kBatchSize>& activations,
-                              KVCache& kv_cache, hwy::ThreadPool& pool,
-                              const LayersOutputFunc& layers_output) {
+template <class TConfig, size_t kBatchSize, size_t kQueryBatchSize>
+HWY_NOINLINE void Transformer(
+    const int* tokens, size_t num_tokens, size_t num_queries, size_t pos,
+    const CompressedWeights<TConfig>& weights,
+    Activations<TConfig, kBatchSize * kQueryBatchSize>& activations,
+    const std::vector<KVCache*>& kv_caches,
+    hwy::ThreadPool& pool,
+    const LayersOutputFunc& layers_output) {
   HWY_ASSERT(num_tokens <= kBatchSize);
+  const size_t num_tokens_and_queries = num_tokens * num_queries;
   if (layers_output) {
-    for (size_t token_idx = 0; token_idx < num_tokens; ++token_idx) {
+    for (size_t token_idx = 0; token_idx < num_tokens_and_queries;
+         ++token_idx) {
       float token_f = tokens[token_idx];
       layers_output(pos + token_idx, "Tokens", &token_f, 1);
     }
   }
   constexpr size_t kModelDim = TConfig::kModelDim;
-  for (size_t token_idx = 0; token_idx < num_tokens; ++token_idx) {
-    EmbedToken(tokens[token_idx], token_idx, pos, weights, activations);
+  for (size_t token_idx = 0; token_idx < num_tokens_and_queries; ++token_idx) {
+    EmbedToken<TConfig, kBatchSize * kQueryBatchSize>(
+        tokens[token_idx], token_idx, pos, weights, activations);
   }
 
   for (size_t layer = 0; layer < TConfig::kLayers; ++layer) {
     const CompressedLayer<TConfig>* layer_weights = weights.GetLayer(layer);
-    TransformerLayer(num_tokens, pos, layer, layer_weights, activations,
-                     kv_cache, pool);
+    TransformerLayer<TConfig, kBatchSize, kQueryBatchSize>(
+        num_tokens, num_queries, pos, layer, layer_weights,
+        activations, kv_caches, pool);
 
     if (layers_output) {
       const std::string block_name = "blocks." + std::to_string(layer);
-      for (size_t token_idx = 0; token_idx < num_tokens; ++token_idx) {
+      for (size_t token_idx = 0; token_idx < num_tokens_and_queries;
+           ++token_idx) {
         layers_output(pos + token_idx, block_name,
                       activations.x.data() + token_idx * kModelDim, kModelDim);
       }
     }
   }
 
-  RMSNormInplaceBatched<kBatchSize>(num_tokens, weights.final_norm_scale.data(),
-                                    activations.x.data(), kModelDim);
+  RMSNormInplaceBatched<kBatchSize * kQueryBatchSize>(
+      num_tokens * num_queries, weights.final_norm_scale.data(),
+      activations.x.data(), kModelDim);
   if (layers_output) {
-    for (size_t token_idx = 0; token_idx < num_tokens; ++token_idx) {
+    for (size_t token_idx = 0; token_idx < num_tokens_and_queries;
+         ++token_idx) {
       layers_output(pos + token_idx, "final_norm",
                     activations.x.data() + token_idx * kModelDim, kModelDim);
     }
@@ -719,31 +794,69 @@ void RangeChecks(size_t& max_tokens, size_t& max_generated_tokens,
 }
 
 template <class TConfig, size_t kBatchSize>
-Activations<TConfig, kBatchSize>& GetActivations(const ByteStorageT& state_u8) {
-  return *reinterpret_cast<Activations<TConfig, kBatchSize>*>(state_u8.get());
+Activations<TConfig, kBatchSize>& GetActivations(
+    const ByteStorageT& state_u8) {
+  return *reinterpret_cast<Activations<TConfig, kBatchSize>*>(
+      state_u8.get());
 }
 
 }  // namespace
 
 // Placeholder for internal test3, do not remove
 
-template <class TConfig>
+bool StreamToken(size_t query_idx, size_t pos, int token, float weight,
+                 const RuntimeConfig& runtime_config) {
+  if (runtime_config.batch_stream_token) {
+    return runtime_config.batch_stream_token(query_idx, pos, token, weight);
+  }
+  return runtime_config.stream_token(token, weight);
+}
+
+template <class TConfig, size_t kQueryBatchSize>
 void GenerateT(const ByteStorageT& weights_u8, const ByteStorageT& prefill_u8,
                const ByteStorageT& decode_u8,
                const RuntimeConfig& runtime_config,
-               const std::vector<int>& prompt, size_t pos, KVCache& kv_cache,
-               hwy::ThreadPool& pool, TimingInfo& timing_info) {
+               const hwy::Span<const hwy::Span<int>>& prompts, size_t pos,
+               const size_t query_index_offset,
+               const std::vector<KVCache*>& kv_caches, hwy::ThreadPool& pool,
+               TimingInfo& timing_info) {
+  constexpr size_t kAdjustedPrefillBatchSize =
+      std::max((size_t)1, kPrefillBatchSize / kQueryBatchSize);
+  static_assert(kAdjustedPrefillBatchSize >= kMinAdjustedPrefillBatchSize);
+  const size_t num_queries = prompts.size();
+  HWY_DASSERT(num_queries <= kQueryBatchSize);
+  pos *= num_queries;  // position in (num_queries) interleaved token sequence.
   const CompressedWeights<TConfig>& weights =
       *reinterpret_cast<const CompressedWeights<TConfig>*>(weights_u8.get());
   auto& prefill_activations =
-      GetActivations<TConfig, kPrefillBatchSize>(prefill_u8);
-  auto& activations = GetActivations<TConfig, 1>(decode_u8);
+      GetActivations<TConfig,
+                     kAdjustedPrefillBatchSize * kQueryBatchSize>(prefill_u8);
+  auto& activations = GetActivations<TConfig, kQueryBatchSize>(decode_u8);
+
+  size_t min_prompt_size =  (size_t)-1;
+  size_t max_prompt_size = 0;
+  for (int i=0; i < prompts.size(); ++i) {
+    min_prompt_size = std::min(min_prompt_size, prompts[i].size());
+    max_prompt_size = std::max(max_prompt_size, prompts[i].size());
+  }
+
+  std::vector<int> prompt;
+  prompt.reserve(max_prompt_size * prompts.size());
+  for (int i = 0; i < max_prompt_size; ++i) {
+    for (int j=0; j < prompts.size(); ++j) {
+      if (i < prompts[j].size()) {
+        prompt.push_back(prompts[j][i]);
+      } else {
+        prompt.push_back(0);
+      }
+    }
+  }
 
   constexpr size_t kVocabSize = TConfig::kVocabSize;
-  size_t prompt_size = prompt.size();
+
   size_t max_tokens = runtime_config.max_tokens;
   size_t max_generated_tokens = runtime_config.max_generated_tokens;
-  RangeChecks<TConfig>(max_tokens, max_generated_tokens, prompt_size);
+  RangeChecks<TConfig>(max_tokens, max_generated_tokens, max_prompt_size);
   if (pos >= max_tokens) {
     fprintf(stderr, "Warning: pos %zu >= max_tokens %zu, aborting.\n", pos,
             max_tokens);
@@ -760,6 +873,9 @@ void GenerateT(const ByteStorageT& weights_u8, const ByteStorageT& prefill_u8,
                                       runtime_config.accept_token);
   };
 
+  std::vector<bool> reached_eos(num_queries);
+  std::fill(reached_eos.begin(), reached_eos.end(), false);
+
   // pos indexes the KV cache. In the first turn of a chat, pos = 0.
   //
   // After the first turn, pos gets passed in with > 0 corresponding to the
@@ -772,23 +888,44 @@ void GenerateT(const ByteStorageT& weights_u8, const ByteStorageT& prefill_u8,
   // In single-turn (non-chat) usage, pos and pos_offset start at 0 and are
   // always equal.
   size_t pos_offset = 0;  // offset relative to pos
+  // Used to keep track of how many tokens are processed per prompt,
+  // so that we know when to start generating tokens.
+  size_t single_prompt_pos_offset = 0;
   const double prefill_start = hwy::platform::Now();
 
   // Prefill stops before prompt_size - 1 since the last prompt token is the
   // first input token for generation.
-  while (pos_offset < prompt_size - 1) {
-    const size_t batch_size =
-        std::min(kPrefillBatchSize, prompt_size - 1 - pos_offset);
+  while (single_prompt_pos_offset < min_prompt_size - 1) {
+    const size_t batch_size = std::min(
+        kPrefillBatchSize, min_prompt_size - 1 - single_prompt_pos_offset);
+    const size_t batch_and_query_size = batch_size * num_queries;
     HWY_DASSERT(batch_size <= kPrefillBatchSize);
-    HWY_DASSERT(pos_offset + batch_size <= prompt_size - 1);
+    HWY_DASSERT(single_prompt_pos_offset + batch_size <= min_prompt_size - 1);
+    HWY_DASSERT(pos_offset + batch_size <= (min_prompt_size - 1) * num_queries);
     const int* batch_tokens = prompt.data() + pos_offset;
-    Prefill(batch_tokens, batch_size, pos, weights, prefill_activations,
-            kv_cache, pool);
+    Prefill<TConfig, kAdjustedPrefillBatchSize, kQueryBatchSize>(
+        batch_tokens, batch_size, num_queries, pos, weights,
+            prefill_activations, kv_caches, pool);
     for (size_t idx = 0; idx < batch_size; ++idx) {
-      if (!runtime_config.stream_token(batch_tokens[idx], 0.0f)) return;
+      bool all_tokens_eos = true;
+      for (size_t query_idx = 0; query_idx < num_queries; ++query_idx) {
+        if (reached_eos[query_idx]) continue;
+        if (StreamToken(
+                query_idx + query_index_offset, single_prompt_pos_offset,
+                batch_tokens[idx * num_queries + query_idx], 0.0f,
+                runtime_config)) {
+          all_tokens_eos = false;
+        } else {
+          reached_eos[query_idx] = true;
+        }
+      }
+      if (all_tokens_eos) {
+        return;
+      }
     }
-    pos += batch_size;
-    pos_offset += batch_size;
+    pos += batch_and_query_size;
+    pos_offset += batch_and_query_size;
+    single_prompt_pos_offset += batch_size;
   }
 
   timing_info.prefill_tok_sec =
@@ -796,45 +933,81 @@ void GenerateT(const ByteStorageT& weights_u8, const ByteStorageT& prefill_u8,
 
   // Start generation.
   const double gen_start = hwy::platform::Now();
-  HWY_DASSERT(pos_offset == prompt_size - 1);
+  HWY_DASSERT(single_prompt_pos_offset == min_prompt_size - 1);
   size_t pos_gen_start = pos_offset;
   int token = prompt.at(pos_offset);
-  // The loop below is not yet prepared for batch size > 1.
+  std::vector<int>::const_iterator first = prompt.begin() + pos_offset;
+  std::vector<int>::const_iterator last = first + num_queries;
+  std::vector<int> gen_tokens(first, last);
+  // The loop below is not yet prepared for decode batch size > 1.
   HWY_ASSERT(kDecodeBatchSize == 1);
-  if (!runtime_config.stream_token(token, 0.0f)) return;
+  bool all_tokens_eos = true;
+  for (size_t i=0; i < num_queries; ++i) {
+    if (reached_eos[i]) continue;
+    if (StreamToken(i + query_index_offset,
+                    single_prompt_pos_offset, gen_tokens[i], 0.0f,
+                    runtime_config)) {
+      all_tokens_eos = false;
+    } else {
+      reached_eos[i] = true;
+    }
+  }
+  if (all_tokens_eos) {
+    return;
+  }
   for (size_t generate_pos = 0;
-       pos < max_tokens && generate_pos < max_generated_tokens;
-       ++pos, ++pos_offset, ++generate_pos) {
-    Transformer(&token, kDecodeBatchSize, pos, weights, activations, kv_cache,
-                pool, runtime_config.layers_output);
+       generate_pos < max_tokens && generate_pos < max_generated_tokens;
+       ++single_prompt_pos_offset, ++generate_pos) {
+    Transformer<TConfig, kDecodeBatchSize, kQueryBatchSize>(
+        gen_tokens.data(), kDecodeBatchSize, num_queries, pos, weights,
+        activations, kv_caches, pool, runtime_config.layers_output);
     float token_logit = 0.0f;
     // The condition below is always true if we are doing Prefill above.
     // We keep it here for clarity so that the code is correct even if Prefill
     // is disabled.
-    const bool is_generating_phase = pos_offset >= prompt_size - 1;
-    if (is_generating_phase) {
-      PROFILER_ZONE("Gen.Embedding");
-      // Compute logits from last layer activations.
-      MatVec<kVocabSize, TConfig::kModelDim>(
-          weights.embedder_input_embedding, 0, activations.x.data(),
-          activations.even_odd.data(), activations.logits.data(), pool);
-      LogitsSoftCap(30.0f, activations.logits.data(), kVocabSize);
-      // Barrier: must have all logits so we can subtract max.
-      Softmax(activations.logits.data(), kVocabSize);
-      token = sample_token(activations.logits.data(), kVocabSize);
-      token_logit = activations.logits[token];
-      if (generate_pos == 0) {
-        timing_info.time_to_first_token = hwy::platform::Now() - gen_start;
+    bool all_tokens_eos = true;
+    float* x = activations.x.data();
+    float* logits = activations.logits.data();
+    for (size_t i = 0; i < num_queries; ++i, ++pos, ++pos_offset,
+                x += TConfig::kModelDim, logits += kVocabSize) {
+      const size_t prompt_size = prompts[i].size();
+      const bool is_generating_phase =
+          (single_prompt_pos_offset >= prompt_size - 1);
+      if (is_generating_phase) {
+        PROFILER_ZONE("Gen.Embedding");
+        // Compute logits from last layer activations.
+        MatVec<kVocabSize, TConfig::kModelDim>(
+            weights.embedder_input_embedding, 0, x, activations.even_odd.data(),
+            logits, pool);
+        LogitsSoftCap(30.0f, logits, kVocabSize);
+        // Barrier: must have all logits so we can subtract max.
+        Softmax(logits, kVocabSize);
+        token = sample_token(logits, kVocabSize);
+        token_logit = logits[token];
+        if (generate_pos == 0) {
+          timing_info.time_to_first_token = hwy::platform::Now() - gen_start;
+        }
+      } else {
+        // We would take this branch if we were not doing Prefill but would
+        // process the tokens of the prompt one at a time.
+        token = prompt.at(pos_offset);
+        token_logit = 0.0f;
       }
-    } else {
-      // We would take this branch if we were not doing Prefill but would
-      // process the tokens of the prompt one at a time.
-      token = prompt.at(pos_offset + 1);
+
+      if (!reached_eos[i]) {
+        if (!StreamToken(i + query_index_offset, single_prompt_pos_offset+1,
+                         token, token_logit, runtime_config)) {
+          token = runtime_config.eos_id;
+        }
+        if (token != runtime_config.eos_id) {
+          all_tokens_eos = false;
+        } else {
+          reached_eos[i] = true;
+        }
+      }
+      gen_tokens[i] = token;
     }
-    if (!runtime_config.stream_token(token, token_logit)) {
-      token = runtime_config.eos_id;
-    }
-    if (token == runtime_config.eos_id) {
+    if (all_tokens_eos) {
       break;
     }
   }
@@ -842,6 +1015,46 @@ void GenerateT(const ByteStorageT& weights_u8, const ByteStorageT& prefill_u8,
                             (hwy::platform::Now() - gen_start);
 }
 
+template <class TConfig>
+void GenerateOneQueryT(const ByteStorageT& weights_u8,
+                       const ByteStorageT& prefill_u8,
+                       const ByteStorageT& decode_u8,
+                       const RuntimeConfig& runtime_config,
+                       const std::vector<int>& prompt, size_t pos,
+                       KVCache& kv_cache, hwy::ThreadPool& pool,
+                       TimingInfo& timing_info) {
+  std::vector<hwy::Span<int>> prompt_vector = {
+    hwy::Span<int>(const_cast<int*>(prompt.data()), prompt.size())};
+  const hwy::Span<const hwy::Span<int>> prompts(
+      prompt_vector.data(), prompt_vector.size());
+  std::vector<KVCache*> kv_caches = {&kv_cache};
+  GenerateT<TConfig, 1>(weights_u8, prefill_u8, decode_u8,
+                        runtime_config, prompts, pos, 0,
+                        kv_caches, pool, timing_info);
+}
+
+template <class TConfig>
+void GenerateBatchT(const ByteStorageT& weights_u8,
+                    const ByteStorageT& prefill_u8,
+                    const ByteStorageT& decode_u8,
+                    const RuntimeConfig& runtime_config,
+                    const hwy::Span<const hwy::Span<int>>& prompts,
+                    size_t pos, const std::vector<KVCache*>& kv_caches,
+                    hwy::ThreadPool& pool,
+                    TimingInfo& timing_info) {
+  // Disable query batching for Griffin models.
+  constexpr size_t kQueryBatchSize =
+      (TConfig::kGriffinLayers > 0) ? 1 : kBatchedQueryBatchSize;
+  for (size_t i = 0; i < prompts.size(); i += kQueryBatchSize) {
+    const size_t num_queries = std::min(prompts.size() - i, kQueryBatchSize);
+    const hwy::Span<const hwy::Span<int>> current_prompts(
+      prompts.data() + i, num_queries);
+    GenerateT<TConfig, kQueryBatchSize>(weights_u8, prefill_u8, decode_u8,
+                                        runtime_config, current_prompts,
+                                        pos, i, kv_caches, pool, timing_info);
+  }
+}
+
 }  // namespace HWY_NAMESPACE
 }  // namespace gcpp
 HWY_AFTER_NAMESPACE();
@@ -853,8 +1066,13 @@ namespace {
 template <typename TConfig>
 struct AllocateState {
   void operator()(ByteStorageT& prefill, ByteStorageT& decode) const {
-    prefill = AllocateSizeof<Activations<TConfig, kPrefillBatchSize>>();
-    decode = AllocateSizeof<Activations<TConfig, kDecodeBatchSize>>();
+    // When batching queries, the prefill batch size is reduced by a factor
+    // of kBatchedQueryBatchSize
+    prefill = AllocateSizeof<
+        Activations<TConfig,
+                    kMinAdjustedPrefillBatchSize * kBatchedQueryBatchSize>>();
+    decode = AllocateSizeof<
+        Activations<TConfig, kDecodeBatchSize * kBatchedQueryBatchSize>>();
   }
 };
 
@@ -895,13 +1113,28 @@ void Gemma::Generate(const RuntimeConfig& runtime_config,
   pool_.SetWaitMode(hwy::PoolWaitMode::kSpin);
 
   GEMMA_EXPORT_AND_DISPATCH(
-      model_type_, weight_type_, GenerateT,
+      model_type_, weight_type_, GenerateOneQueryT,
       (weights_u8_, prefill_u8_, decode_u8_, runtime_config, prompt, start_pos,
        kv_cache, pool_, timing_info));
 
   pool_.SetWaitMode(hwy::PoolWaitMode::kBlock);
 }
 
+void Gemma::GenerateBatch(const RuntimeConfig& runtime_config,
+                          const hwy::Span<const hwy::Span<int>>& prompts,
+                          size_t start_pos,
+                          const std::vector<KVCache*>& kv_caches,
+                          TimingInfo& timing_info) {
+  pool_.SetWaitMode(hwy::PoolWaitMode::kSpin);
+
+  GEMMA_EXPORT_AND_DISPATCH(
+      model_type_, weight_type_, GenerateBatchT,
+      (weights_u8_, prefill_u8_, decode_u8_, runtime_config, prompts, start_pos,
+       kv_caches, pool_, timing_info));
+
+  pool_.SetWaitMode(hwy::PoolWaitMode::kBlock);
+}
+
 std::vector<int> WrapAndTokenize(const GemmaTokenizer& tokenizer,
                                  const ModelTraining training, size_t pos,
                                  std::string& prompt) {

gemma/gemma.h

@@ -32,6 +32,9 @@ namespace gcpp {
 
 constexpr size_t kPrefillBatchSize = 16;
 constexpr size_t kDecodeBatchSize = 1;
+constexpr size_t kBatchedQueryBatchSize = 16;
+constexpr size_t kMinAdjustedPrefillBatchSize =
+    HWY_MAX((size_t)1, kPrefillBatchSize / kBatchedQueryBatchSize);
 constexpr bool kSystemPrompt = false;
 
 struct KVCache {
@@ -72,6 +75,11 @@ class GemmaTokenizer {
 // probability is 0.0f. StreamFunc should return false to stop generation and
 // true to continue generation.
 using StreamFunc = std::function<bool(int, float)>;
+// BatchStreamFunc is called with (query_idx, pos, token, probability). 
+// For prompt tokens,
+// probability is 0.0f. StreamFunc should return false to stop generation and
+// true to continue generation.
+using BatchStreamFunc = std::function<bool(size_t, size_t, int, float)>;
 // If not empty, AcceptFunc is called with token. It should return false for
 // tokens you don't want to generate and true for tokens you want to generate.
 using AcceptFunc = std::function<bool(int, float)>;
@@ -93,6 +101,7 @@ struct RuntimeConfig {
   int verbosity;
   std::mt19937* gen;
   StreamFunc stream_token;
+  BatchStreamFunc batch_stream_token;
   AcceptFunc accept_token;  // if empty, accepts all tokens.
   SampleFunc sample_func;   // if empty, uses SampleTopK.
   LayersOutputFunc layers_output;  // if not empty, called after each layer.
@@ -125,6 +134,11 @@ class Gemma {
                 const std::vector<int>& prompt, size_t start_pos,
                 KVCache& kv_cache, TimingInfo& timing_info);
 
+  void GenerateBatch(const RuntimeConfig& runtime_config,
+                     const hwy::Span<const hwy::Span<int>>& prompts,
+                     size_t start_pos, const std::vector<KVCache*>& kv_caches,
+                     TimingInfo& timing_info);
+
  private:
   hwy::ThreadPool& pool_;
 

gemma/gemma_test.cc

@@ -17,9 +17,11 @@
 
 #include <stdio.h>
 
+#include <memory>
 #include <string>
 #include <vector>
 
+#include "hwy/aligned_allocator.h"
 #include "gemma/benchmark_helper.h"
 #include "gemma/common.h"
 #include "hwy/tests/hwy_gtest.h"
@@ -44,13 +46,55 @@ class GemmaTest : public ::testing::Test {
     return response;
   }
 
-  void TestQuestions(const char* kQA[][2], size_t num_questions) {
+  std::vector<std::string> BatchGemmaReply(
+      const std::vector<std::string>& inputs) {
+    s_env->SetMaxGeneratedTokens(64);
+    s_env->MutableConfig().temperature = 0.0f;  // deterministic
+    s_env->MutableConfig().verbosity = 0;
+    // Using the turn structure worsens results.
+    std::vector<std::unique_ptr<std::vector<int>>> prompts;
+    prompts.reserve(inputs.size());
+    for (auto input_string : inputs) {
+      std::string mutable_input_string = input_string;
+      prompts.push_back(std::make_unique<std::vector<int>>(
+          s_env->TokenizeAndPrependBOS(input_string)));
+    }
+    std::vector<hwy::Span<int>> prompt_vector;
+    for (auto& prompt : prompts) {
+      prompt_vector.push_back(hwy::Span<int>(
+          prompt->data(), prompt->size()));
+    }
+    hwy::Span<const hwy::Span<int>> prompt_span =
+        hwy::Span<const hwy::Span<int>>(
+            prompt_vector.data(), prompt_vector.size());
+    std::vector<std::string> replies;
+    for (auto [response, n] : s_env->BatchQueryModel2(prompt_span)) {
+      replies.push_back(response);
+    }
+    return replies;
+  }
+
+  void TestQuestions(const char* kQA[][2], size_t num_questions, bool batch) {
     if (!s_env->GetModel()) return;
-    for (size_t i = 0; i < num_questions; ++i) {
-      fprintf(stderr, "Question %zu\n\n", i + 1);
-      std::string response = GemmaReply(kQA[i][0]);
-      fprintf(stderr, "'%s'\n\n", response.c_str());
-      EXPECT_TRUE(response.find(kQA[i][1]) != std::string::npos);  // NOLINT
+    if (batch) {
+      std::vector<std::string> inputs;
+      for (size_t i = 0; i < num_questions; ++i) {
+        fprintf(stderr, "Batch Question %zu\n\n", i + 1);
+        inputs.push_back(kQA[i][0]);
+      }
+      std::vector<std::string> responses = BatchGemmaReply(inputs);
+      for (size_t i = 0; i < num_questions; ++i) {
+        std::string response = responses.at(i);
+        fprintf(stderr, "Batch answer %zu '%s'\n\n", i + 1, response.c_str());
+        EXPECT_TRUE(response.find(kQA[i][1]) != std::string::npos);  // NOLINT
+      }
+    } else {
+      for (size_t i = 0; i < num_questions; ++i) {
+        fprintf(stderr, "Question %zu\n\n", i + 1);
+        std::string response = GemmaReply(kQA[i][0]);
+        fprintf(stderr, "'%s'\n\n", response.c_str());
+        EXPECT_TRUE(response.find(kQA[i][1]) != std::string::npos);  // NOLINT
+      }
     }
   }
 };
@@ -58,10 +102,16 @@ class GemmaTest : public ::testing::Test {
 TEST_F(GemmaTest, Geography) {
   static const char* kQA[][2] = {
       {"What is the capital of Hungary?", "Budapest"},
+      {"What is the capital of Australia?", "Canberra"},
       {"How many states does the US have?", "50"},
   };
   static const size_t kNum = sizeof(kQA) / sizeof(kQA[0]);
-  TestQuestions(kQA, kNum);
+  TestQuestions(kQA, kNum, /* batch= */ false);
+  static const char* kQA_single_question[][2] = {
+      {"What is the capital of Australia?", "Canberra"},
+  };
+  TestQuestions(kQA_single_question, 1, /* batch= */ true);
+  TestQuestions(kQA, kNum, /* batch= */ true);
 }
 
 TEST_F(GemmaTest, History) {
@@ -69,7 +119,7 @@ TEST_F(GemmaTest, History) {
       {"When was the battle of Hastings?", "1066"},
   };
   static const size_t kNum = sizeof(kQA) / sizeof(kQA[0]);
-  TestQuestions(kQA, kNum);
+  TestQuestions(kQA, kNum, /* batch= */ false);
 }
 
 TEST_F(GemmaTest, Arithmetic) {
@@ -78,7 +128,7 @@ TEST_F(GemmaTest, Arithmetic) {
       {"what is 7 * 8?", "56"},
   };
   static const size_t kNum = sizeof(kQA) / sizeof(kQA[0]);
-  TestQuestions(kQA, kNum);
+  TestQuestions(kQA, kNum, /* batch= */ false);
 }
 
 static const char kJingleBells[] = R"(