// Copyright 2024 Google LLC
// SPDX-License-Identifier: Apache-2.0
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
//     https://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.

// Defines Gemma member functions which dynamic-dispatch into the SIMD
// implementations in gemma-inl.h.

#include "gemma/gemma.h"

// Compiles this file for multiple architectures via "foreach_target.h", to
// which we pass the filename via macro 'argument'.
// clang-format off
#undef HWY_TARGET_INCLUDE
#define HWY_TARGET_INCLUDE "gemma/gemma.cc"  // NOLINT
// clang-format on
#include "hwy/foreach_target.h"  // IWYU pragma: keep
#include "hwy/highway.h"
// After highway.h
#include "gemma/attention.h"  // includes highway.h
#include "gemma/gemma-inl.h"
#include "gemma/griffin.h"  // includes highway.h
#include "gemma/vit.h"      // includes highway.h

#ifndef GEMMA_CC_ONCE
#define GEMMA_CC_ONCE

#include <math.h>  // sqrtf
#include <stddef.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>

#include <vector>

#include "gemma/configs.h"
#include "gemma/model_store.h"
#include "gemma/tokenizer.h"
#include "gemma/weights.h"
#include "io/blob_store.h"
#include "io/io.h"  // Path
#include "ops/matmul.h"
#include "paligemma/image.h"
#include "util/threading_context.h"
#include "hwy/aligned_allocator.h"  // Span
#include "hwy/base.h"
#include "hwy/timer.h"

#endif  // GEMMA_CC_ONCE

HWY_BEFORE_NAMESPACE();
namespace gcpp {
namespace HWY_NAMESPACE {

void Attention(LayerAttentionType type, size_t num_tokens,
               const QueriesPos& queries_pos,
               const QueriesPos& queries_prefix_end,
               const hwy::Divisor& div_seq_len, const size_t layer_idx,
               const LayerWeightsPtrs& layer, Activations& activations,
               const KVCaches& kv_caches, MatMulEnv& env) {
  if (type == LayerAttentionType::kGemma) {
    GemmaAttention(num_tokens, queries_pos, &queries_prefix_end, div_seq_len,
                   layer_idx, layer, activations, kv_caches, env,
                   /*flags=*/0);
  } else {
    HWY_DASSERT(type == LayerAttentionType::kGriffinRecurrentBlock);
    // KVCache conv1d_cache and rglru_cache have one row per *Griffin* layer,
    // so map `layer` to the Griffin layer index.
    const size_t griffin_layer =
        activations.weights_config.NumLayersOfTypeBefore(type, layer_idx);
    GriffinRecurrent(queries_pos, num_tokens, griffin_layer, activations,
                     &layer, kv_caches, env);
  }
}

static HWY_NOINLINE void TransformerLayer(
    const size_t num_tokens, const QueriesPos& queries_pos,
    const QueriesPos& queries_prefix_end, const hwy::Divisor& div_seq_len,
    const size_t layer_idx, const LayerWeightsPtrs& layer,
    Activations& activations, const KVCaches& kv_caches, MatMulEnv& env) {
  const LayerConfig& layer_config = layer.layer_config;

  RMSNormBatched(activations.x, layer.pre_attention_norm_scale,
                 activations.pre_att_rms_out);

  Attention(layer_config.type, num_tokens, queries_pos, queries_prefix_end,
            div_seq_len, layer_idx, layer, activations, kv_caches, env);

  PostNorm(layer_config.post_norm, layer.post_attention_norm_scale,
           activations.att_sums);

  ResidualConnection(activations.att_sums, activations.x, layer,
                     /*is_attention=*/true);

  RMSNormBatched(activations.x, layer.pre_ffw_norm_scale,
                 activations.pre_ffw_rms_out);

  if (layer_config.type == LayerAttentionType::kVit) {
    FFWVit(layer, activations, env);
  } else {
    FFWNoVit(layer, activations, env);
  }

  PostNorm(layer_config.post_norm, layer.post_ffw_norm_scale,
           activations.ffw_out);

  ResidualConnection(activations.ffw_out, activations.x, layer,
                     /*is_attention=*/false);
}

// Returns the scale value to use for the embedding (basically sqrt model_dim).
static float EmbeddingScaling(size_t model_dim) {
  // Round to bf16 to match Gemma's Embedder, which casts before mul.
  return hwy::ConvertScalarTo<float>(
      hwy::ConvertScalarTo<BF16>(sqrtf(static_cast<float>(model_dim))));
}

// `batch_idx` indicates which row of `x` to write to.
// `pos` is the *token*'s position, not the start of the batch, because this is
// called for batches of tokens in prefill, but batches of queries in decode.
//
// For GEMMA_VLM, image tokens are copied into -2 locations (per the Gemma 3
// spec) until we run out of image tokens. This allows for a multi-image prompt
// if -2 locations with appropriate begin/end image tokens are created by the
// calling application.
// Returns new image_token_position.
static HWY_NOINLINE size_t
EmbedMMToken(int token, size_t batch_idx, size_t pos, size_t pos_in_prompt,
             const ModelConfig& model_config, const ModelWeightsPtrs& weights,
             MatStorageT<float>& x, const ImageTokens* image_tokens = nullptr,
             size_t image_token_position = 0) {
  // Image tokens just need to be copied.
  if (model_config.wrapping == PromptWrapping::GEMMA_VLM &&
      image_tokens != nullptr && token == -2 &&
      image_token_position < image_tokens->Rows()) {
    hwy::CopyBytes(image_tokens->Row(image_token_position), x.Row(batch_idx),
                   x.Cols() * x.ElementBytes());
    return image_token_position + 1;
  }

  if (model_config.wrapping == PromptWrapping::PALIGEMMA &&
      image_tokens != nullptr && pos_in_prompt < image_tokens->Rows()) {
    hwy::CopyBytes(image_tokens->Row(pos_in_prompt), x.Row(batch_idx),
                   x.Cols() * x.ElementBytes());
    return image_token_position;
  }

  const size_t model_dim = model_config.model_dim;
  const float emb_scaling = EmbeddingScaling(model_dim);

  HWY_DASSERT(token >= 0);
  HWY_DASSERT(token < static_cast<int>(model_config.vocab_size));

  CallUpcasted(&weights.embedder_input_embedding, [&](const auto* weights_t) {
    // Using `Stride` to compute the offset works for both NUQ (because we use
    // an offset and NUQ is never padded) and padded, because non-NUQ types are
    // seekable, hence the offset can also skip any padding.
    const size_t embedding_ofs = token * weights_t->Stride();
    HWY_ASSERT(weights_t->Cols() == model_dim);
    const auto embedding_span =
        MakeSpan(weights_t->Row(0), embedding_ofs + model_dim);
    const hn::ScalableTag<float> df;
    DecompressAndZeroPad(df, embedding_span, embedding_ofs, x.Row(batch_idx),
                         model_dim);
    MulByConst(emb_scaling * weights_t->Scale(), x.Row(batch_idx), model_dim);
  });

  if (model_config.absolute_pe) {
    AddAbsolutePositionalEmbeddings(x.Row(batch_idx), model_dim, pos);
  }
  return image_token_position;
}

// Incremented in-place by Prefill* and DecodeStepT.
using QueriesMutablePos = hwy::Span<size_t>;

// Populates KV cache for batches of tokens from one query at a time. This is
// called if prompts are longer than the query batch size, and also in
// prefix-LM mode (end > 0), which must see all tokens in one batch.
static HWY_NOINLINE void PrefillTBatch(
    const size_t query_idx_start, const QueriesPromptTokens& queries_prompt,
    const QueriesMutablePos& queries_pos, const QueriesPos& queries_prefix_end,
    const hwy::Divisor& div_seq_len, const ModelConfig& config,
    const RuntimeConfig& runtime_config, const ModelWeightsPtrs& weights,
    Activations& activations, const KVCaches& kv_caches, MatMulEnv& env,
    hwy::BitSet4096<>& non_eos) {
  PROFILER_ZONE("Gen.PrefillT");
  const size_t num_queries = queries_prompt.size();
  HWY_DASSERT(num_queries == queries_pos.size());
  HWY_DASSERT(num_queries == queries_prefix_end.size());
  HWY_DASSERT(num_queries == kv_caches.size());

  // Batches are important for amortizing loading weights over multiple tokens.
  // This is possible in prefill because we know all tokens beforehand, whereas
  // decode depends on the previous output token. However, each prefill batch of
  // a query requires that preceding batches already wrote to the KV cache,
  // hence we sequentially loop over token batches. We can reduce the number of
  // iterations by increasing the batch size, but this also increases arithmetic
  // intensity, and so we are eventually compute-limited. TransformerLayer uses
  // all available threads, so we do not also parallelize over queries, but note
  // that PrefillQBatch uses queries as the batch dimension.
  const size_t max_tbatch_size = runtime_config.prefill_tbatch_size;

  // For each query. `qi` is within the batch, not the global query index.
  for (size_t qi = 0; qi < num_queries; ++qi) {
    non_eos.Set(qi);

    // Single query at a time, so pass slices of the spans because
    // GemmaAttention will only access the first KV cache and position.
    QueriesPos single_query_pos(&queries_pos[qi], 1);
    QueriesPos single_query_prefix_end(&queries_prefix_end[qi], 1);
    KVCaches single_kv_cache(&kv_caches[qi], 1);

    const size_t prompt_size = queries_prompt[qi].size();
    // In autoregressive mode, we don't need to prefill the last token, so - 1.
    size_t prefill_this_query = prompt_size - 1;
    const size_t prefix_end_this_query = queries_prefix_end[qi];
    // We can't attend beyond the prompt_size.
    HWY_ASSERT(prefix_end_this_query <= prompt_size);
    // Special case: if the prefix includes the last token, we need to prefill
    // the last token, too. However, we need to rewind this for the generation
    // of the first token. So we need to keep track of this.
    // TODO: consider implementing masking instead of this logic?
    const bool attend_to_last_token =
        (prefill_this_query < prefix_end_this_query);
    if (attend_to_last_token) {
      // The difference can be at most 1.
      prefill_this_query += 1;
      HWY_ASSERT(prefill_this_query == prefix_end_this_query);
    }
    // In prefix-LM mode, we need to look at all the tokens for the prefix in
    // one iteration through the layers, so we need a large enough batch size.
    HWY_ASSERT(prefix_end_this_query == 0 ||
               max_tbatch_size >= prefill_this_query);

    // For each batch of tokens in the query:
    for (size_t tbatch_start = 0; tbatch_start < prefill_this_query;
         tbatch_start += max_tbatch_size) {
      const size_t tbatch_size =
          HWY_MIN(max_tbatch_size, prefill_this_query - tbatch_start);
      activations.SetBatchSize(tbatch_size);

      // Fill activations.x (much faster than TransformerLayer).
      size_t image_token_position = 0;
      for (size_t ti = 0; ti < tbatch_size; ++ti) {
        const size_t pos = queries_pos[qi] + ti;
        const size_t pos_in_prompt = tbatch_start + ti;
        const int token = queries_prompt[qi][pos_in_prompt];
        image_token_position = EmbedMMToken(
            token, ti, pos, pos_in_prompt, config, weights, activations.x,
            runtime_config.image_tokens, image_token_position);
      }

      // Transformer with one batch of tokens from a single query.
      for (size_t layer_idx = 0; layer_idx < config.layer_configs.size();
           ++layer_idx) {
        TransformerLayer(tbatch_size, single_query_pos, single_query_prefix_end,
                         div_seq_len, layer_idx, *weights.GetLayer(layer_idx),
                         activations, single_kv_cache, env);
      }

      // NOTE: we unconditionally call StreamToken, even if EOS.
      for (size_t ti = 0; ti < tbatch_size; ++ti) {
        const size_t pos = queries_pos[qi] + ti;
        const size_t pos_in_prompt = tbatch_start + ti;
        const int token = queries_prompt[qi][pos_in_prompt];
        if (pos_in_prompt < prompt_size - 1) {
          runtime_config.StreamToken(query_idx_start + qi, pos, token, 0.0f);
        } else {
          // The last token will be streamed later and we should only get here
          // if we need to attend to the last token because it is in the prefix.
          HWY_ASSERT(attend_to_last_token);
        }
      }

      queries_pos[qi] += tbatch_size;
    }  // for tbatch_start
    if (attend_to_last_token) {
      // We need to rewind the position for the last token that we only
      // attended to to make sure the prefix LM sees everything.
      // This means we duplicate work on the last prompt token in autoregressive
      // decoding. Alternatives: (1) real masking; (2) always prefill the last
      // token and only generate the next one from the already prefilled
      // activations.
      queries_pos[qi] -= 1;
    }
  }
}

// Embeds token and calls each TransformerLayer. `queries_token` is the previous
// token from each query, and `queries_pos` are their position in the sequence.
// Called by query-batched `PrefillQBatch` and `DecodeStepT`, but not the
// token-batched `PrefillTBatch`.
static HWY_NOINLINE void Transformer(
    const QueriesToken& queries_token, const QueriesMutablePos& queries_pos,
    const QueriesPos& queries_prefix_end, const hwy::Divisor& div_seq_len,
    const ModelConfig& config, const RuntimeConfig& runtime_config,
    const ModelWeightsPtrs& weights, Activations& activations,
    const KVCaches& kv_caches, MatMulEnv& env) {
  const size_t num_queries = queries_token.size();
  HWY_DASSERT(num_queries == queries_pos.size());
  HWY_DASSERT(num_queries == queries_prefix_end.size());

  if (HWY_UNLIKELY(runtime_config.layers_output)) {
    for (size_t qi = 0; qi < num_queries; ++qi) {
      const float token_f = queries_token[qi];
      runtime_config.layers_output(qi, queries_pos[qi], "tokens", -1, &token_f,
                                   1);
    }
  }

  size_t image_token_position = 0;
  for (size_t qi = 0; qi < num_queries; ++qi) {
    image_token_position =
        EmbedMMToken(queries_token[qi], qi, queries_pos[qi],
                     /*pos_in_prompt=*/0, config, weights, activations.x,
                     runtime_config.image_tokens, image_token_position);
  }

  for (size_t layer_idx = 0; layer_idx < weights.c_layers.size(); ++layer_idx) {
    TransformerLayer(/*num_tokens=*/1, queries_pos, queries_prefix_end,
                     div_seq_len, layer_idx, *weights.GetLayer(layer_idx),
                     activations, kv_caches, env);

    if (HWY_UNLIKELY(runtime_config.activations_observer)) {
      runtime_config.activations_observer(queries_pos, layer_idx, activations);
    }
  }
}

// Populates KV cache for the batch queries, one token at a time. Only called
// for autoregressive (non-prefix-LM) prefill, so `queries_prefix_end` == 0.
static HWY_NOINLINE void PrefillQBatch(
    const size_t query_idx_start, const QueriesPromptTokens& queries_prompt,
    const QueriesMutablePos& queries_pos, const QueriesPos& queries_prefix_end,
    const size_t max_prompt_size, const hwy::Divisor& div_seq_len,
    const ModelConfig& config, const RuntimeConfig& runtime_config,
    const ModelWeightsPtrs& weights, Activations& activations,
    const KVCaches& kv_caches, MatMulEnv& env, hwy::BitSet4096<>& non_eos) {
  PROFILER_ZONE("Gen.Prefill");
  const size_t num_queries = queries_prompt.size();
  HWY_DASSERT(num_queries == queries_pos.size());
  HWY_DASSERT(num_queries == queries_prefix_end.size());
  HWY_DASSERT(num_queries == activations.x.Rows());
  HWY_DASSERT(num_queries == kv_caches.size());

  hwy::BitSet4096<> prefill_active;
  for (size_t qi = 0; qi < num_queries; ++qi) {
    prefill_active.Set(qi);

    HWY_DASSERT(queries_prefix_end[qi] == 0);
    (void)queries_prefix_end;
  }
  non_eos = prefill_active;

  // In autoregressive mode, we don't prefill the last token, hence - 1.
  for (size_t pos_in_prompt = 0; pos_in_prompt < max_prompt_size - 1;
       ++pos_in_prompt) {
    // Streams that have already finished prefill no longer interleave/stream.
    for (size_t qi = 0; qi < num_queries; ++qi) {
      if (pos_in_prompt >= queries_prompt[qi].size() - 1) {
        prefill_active.Clear(qi);
        activations.gen_tokens[qi] = config.eos_id;
      }
    }

    // Batch := interleaved tokens, one from each non-EOS query.
    prefill_active.Foreach([&](size_t qi) {
      activations.gen_tokens[qi] = queries_prompt[qi][pos_in_prompt];
    });

    // One token from each query in the batch. Increments queries_pos.
    // Do not call DecodeStepT because it computes logits for token
    // probabilities, which are not required for the prompt tokens.
    Transformer(QueriesToken(activations.gen_tokens.data(), num_queries),
                queries_pos, queries_prefix_end, div_seq_len, config,
                runtime_config, weights, activations, kv_caches, env);

    prefill_active.Foreach([&](size_t qi) {
      const int token = queries_prompt[qi][pos_in_prompt];
      // Ignore any user request to stop during prefill.
      (void)runtime_config.StreamToken(query_idx_start + qi, queries_pos[qi],
                                       token, 0.0f);
      queries_pos[qi] += 1;
    });
  }  // pos_in_prompt
}

// TODO: inline.
void RangeChecks(const ModelConfig& weights_config,
                 size_t& max_generated_tokens, const size_t prompt_size) {
  if (!weights_config.use_local_attention) {
    if (max_generated_tokens > weights_config.seq_len) {
      HWY_WARN("max_generated_tokens %zu > kSeqLen %u, truncating.",
               max_generated_tokens, weights_config.seq_len);
      max_generated_tokens = weights_config.seq_len;
    }
  }
  HWY_ASSERT(prompt_size > 0);
}

// Also writes the token to activations.gen_tokens for subsequent DecodeStepT,
// and updates `non_eos` if the query is at the end of its sequence.
static void StreamAndUpdateEOS(const size_t qi, const size_t pos, int token,
                               const float prob, const ModelConfig& config,
                               const RuntimeConfig& runtime_config,
                               Activations& activations,
                               hwy::BitSet4096<>& non_eos) {
  HWY_DASSERT(non_eos.Get(qi));

  // User decided to stop: set next token to primary EOS.
  if (HWY_UNLIKELY(!runtime_config.StreamToken(qi, pos, token, prob))) {
    token = config.eos_id;
  }

  // Primary or secondary EOS: mark query as EOS.
  if (HWY_UNLIKELY(config.IsEOS(token))) non_eos.Clear(qi);

  activations.gen_tokens[qi] = token;
}

// For a batch of queries, runs Transformer, computes logits, samples and
// streams the token.
static void DecodeStepT(
    const size_t query_idx_start, const QueriesPromptTokens& queries_prompt,
    const QueriesMutablePos& queries_mutable_pos,
    const QueriesPos& queries_prefix_end, const hwy::Divisor div_seq_len,
    const ModelConfig& config, const RuntimeConfig& runtime_config,
    const ModelWeightsPtrs& weights, const SampleFunc& sample_token,
    Activations& activations, const KVCaches& kv_caches, MatMulEnv& env,
    hwy::BitSet4096<>& non_eos, TimingInfo& timing_info) {
  const size_t num_queries = queries_prompt.size();
  HWY_DASSERT(num_queries == activations.x.Rows());

  Transformer(QueriesToken(activations.gen_tokens.data(), num_queries),
              queries_mutable_pos, queries_prefix_end, div_seq_len, config,
              runtime_config, weights, activations, kv_caches, env);

  RMSNormInplaceBatched(weights.final_norm_scale, activations.x);

  if (HWY_UNLIKELY(runtime_config.activations_observer)) {
    runtime_config.activations_observer(queries_mutable_pos, -1, activations);
  }

  {
    PROFILER_ZONE("Gen.EmbeddingMatmul");
    // Compute logits from last layer activations.
    CallMatMul(activations.x, weights.embedder_input_embedding,
               /*add=*/nullptr, env, activations.logits);
  }
  PROFILER_ZONE("Gen.Softcap+Sample+Stream");
  non_eos.Foreach([&](size_t qi) {
    float* HWY_RESTRICT logits = activations.logits.Row(qi);
    MaybeLogitsSoftCap(config.final_cap, logits, config.vocab_size);
    const TokenAndProb tp = sample_token(logits, config.vocab_size);
    timing_info.NotifyGenerated();

    StreamAndUpdateEOS(query_idx_start + qi, queries_mutable_pos[qi], tp.token,
                       tp.prob, config, runtime_config, activations, non_eos);

    if (non_eos.Get(qi)) queries_mutable_pos[qi] += 1;
  });
}

static HWY_INLINE SampleFunc
ChooseSampleFunc(const RuntimeConfig& runtime_config) {
  // If user provided a sample_func, use it.
  if (runtime_config.sample_func) return runtime_config.sample_func;

  // Fast path for top-1 with no accept_token.
  if (runtime_config.top_k == 1 && !runtime_config.accept_token) {
    return [](float* logits, size_t vocab_size) HWY_ATTR -> TokenAndProb {
      PROFILER_ZONE("Gen.Sample Top1");
      return Top1OfSoftmax(logits, vocab_size);
    };
  }

  // General case: Softmax with top-k sampling.
  return [&runtime_config](float* logits,
                           size_t vocab_size) HWY_ATTR -> TokenAndProb {
    PROFILER_ZONE("Gen.Sample general");
    return FusedSoftmaxAndSampleTopK(
        logits, runtime_config.top_k, vocab_size, *runtime_config.gen,
        runtime_config.temperature, runtime_config.accept_token);
  };
}

// Generates one continuation for each query in `queries_prompt`, which is one
// qbatch whose size is at most the `batch_size` passed to `activations` ctor.
//
// `queries_pos` stores the KV cache position for each query. In the first turn
// of a chat, pos = 0; we increment each query's position after each token.
//
// `query_idx_start` is the query_idx of the first query in the batch, so that
// `StreamFunc` gets the global query index, not relative to the batch.
static void GenerateT(
    const size_t query_idx_start, const QueriesPromptTokens& queries_prompt,
    const QueriesPos& queries_pos_in, const QueriesPos& queries_prefix_end,
    const ModelConfig& config, const RuntimeConfig& runtime_config,
    const ModelWeightsPtrs& weights, Activations& activations,
    const KVCaches& kv_caches, MatMulEnv& env, TimingInfo& timing_info) {
  const size_t num_queries = queries_prompt.size();
  HWY_ASSERT(num_queries <= 4096);  // non_eos uses `BitSet4096`.
  HWY_ASSERT(num_queries == queries_pos_in.size());
  HWY_ASSERT(num_queries == queries_prefix_end.size());
  HWY_ASSERT(num_queries <= activations.x.Rows());
  HWY_ASSERT(num_queries == kv_caches.size());

  // Griffin assumes that the recurrent block cache is zero-initialized.
  for (size_t i = 0; i < kv_caches.size(); ++i) {
    if (queries_pos_in[i] == 0) {
      kv_caches[i].ZeroGriffinCache();  // No-op for non-Griffin models.
    }
  }

  // Copy so we can increment without requiring users to pass in a mutable span.
  std::vector<size_t> queries_pos_copy(queries_pos_in.cbegin(),
                                       queries_pos_in.cend());
  const QueriesMutablePos queries_mutable_pos(queries_pos_copy.data(),
                                              queries_pos_copy.size());

  size_t max_prompt_size = 0;
  bool all_prefix_end_are_zero = true;
  size_t prefill_tokens = 0;
  for (size_t qi = 0; qi < num_queries; ++qi) {
    const PromptTokens& prompt = queries_prompt[qi];
    max_prompt_size = HWY_MAX(max_prompt_size, prompt.size());

    // Prefill stops before size - 1 because the last prompt token is the
    // first input token for generation.
    prefill_tokens += prompt.size() - 1;

    // Sanity check: prompts should not be empty, nor start with EOS.
    HWY_ASSERT(prompt.size() != 0 && prompt[0] != config.eos_id);

    all_prefix_end_are_zero &= queries_prefix_end[qi] == 0;
  }

  const hwy::Divisor div_seq_len(static_cast<uint32_t>(kv_caches[0].seq_len));

  // Lacks a constructor to bulk-set, hence initialized by Prefill* which have
  // qi loops anyway.
  hwy::BitSet4096<> non_eos;

  timing_info.prefill_start = hwy::platform::Now();
  // Batch over the larger of prompt length, or queries.
  if ((num_queries > max_prompt_size) && all_prefix_end_are_zero) {
    activations.SetBatchSize(num_queries);  // required before PrefillQBatch
    PrefillQBatch(query_idx_start, queries_prompt, queries_mutable_pos,
                  queries_prefix_end, max_prompt_size, div_seq_len, config,
                  runtime_config, weights, activations, kv_caches, env,
                  non_eos);
  } else {
    PrefillTBatch(query_idx_start, queries_prompt, queries_mutable_pos,
                  queries_prefix_end, div_seq_len, config, runtime_config,
                  weights, activations, kv_caches, env, non_eos);
    activations.SetBatchSize(num_queries);  // Restore after PrefillTBatch.
  }
  HWY_DASSERT(num_queries == non_eos.Count());
  timing_info.NotifyPrefill(prefill_tokens);
  // queries_pos have been incremented by Prefill.

  // Stream the last prompt token from each query, fill activations.gen_tokens.
  for (size_t qi = 0; qi < num_queries; ++qi) {
    const size_t last_token_pos_in_prompt =
        queries_mutable_pos[qi] - queries_pos_in[qi];
    StreamAndUpdateEOS(query_idx_start + qi, queries_mutable_pos[qi],
                       queries_prompt[qi][last_token_pos_in_prompt], 0.0f,
                       config, runtime_config, activations, non_eos);
    // No incrementing queries_mutable_pos[qi].
  }

  size_t max_gen_steps = runtime_config.max_generated_tokens;
  RangeChecks(config, max_gen_steps, max_prompt_size);

  const SampleFunc sample_token = ChooseSampleFunc(runtime_config);

  {
    timing_info.generate_start = hwy::platform::Now();
    for (size_t gen = 0; gen < max_gen_steps && non_eos.Any(); ++gen) {
      DecodeStepT(query_idx_start, queries_prompt, queries_mutable_pos,
                  queries_prefix_end, div_seq_len, config, runtime_config,
                  weights, sample_token, activations, kv_caches, env, non_eos,
                  timing_info);
    }
    timing_info.NotifyGenerateDone();
  }
}

void GenerateSingleT(const PromptTokens& prompt, size_t pos, size_t prefix_end,
                     const ModelConfig& config,
                     const RuntimeConfig& runtime_config,
                     const ModelWeightsPtrs& weights, KVCache& kv_cache,
                     MatMulEnv& env, TimingInfo& timing_info) {
  constexpr size_t kNumQueries = 1;
  const size_t qbatch_start = 0;

  const size_t max_batch_size =
      HWY_MAX(kNumQueries, runtime_config.prefill_tbatch_size);
  // TODO: move into Gemma?
  Activations activations(config, max_batch_size, env.row_ptrs);

  const QueriesPromptTokens queries_prompt(&prompt, kNumQueries);
  QueriesPos queries_pos(&pos, kNumQueries);
  const QueriesPos queries_prefix_end(&prefix_end, kNumQueries);
  const KVCaches kv_caches{&kv_cache, kNumQueries};

  GenerateT(qbatch_start, queries_prompt, queries_pos, queries_prefix_end,
            config, runtime_config, weights, activations, kv_caches, env,
            timing_info);
}

// Splits the input into batches of at most `runtime_config.decode_qbatch_size`
// queries, and calls `GenerateT` on each batch.
void GenerateBatchT(const QueriesPromptTokens& queries_prompt,
                    const QueriesPos& queries_pos,
                    const QueriesPos& queries_prefix_end,
                    const ModelConfig& config,
                    const RuntimeConfig& runtime_config,
                    const ModelWeightsPtrs& weights, const KVCaches& kv_caches,
                    MatMulEnv& env, TimingInfo& timing_info) {
  const size_t num_queries = queries_prompt.size();
  HWY_ASSERT(queries_pos.size() == num_queries);
  HWY_ASSERT(kv_caches.size() >= num_queries);

  const size_t max_qbatch_size = runtime_config.decode_qbatch_size;
  const size_t max_batch_size =
      HWY_MAX(max_qbatch_size, runtime_config.prefill_tbatch_size);
  Activations activations(config, max_batch_size, env.row_ptrs);

  for (size_t qbatch_start = 0; qbatch_start < num_queries;
       qbatch_start += max_qbatch_size) {
    // Generate one batch of tokens from `qbatch_size` queries.
    const size_t qbatch_size =
        HWY_MIN(num_queries - qbatch_start, max_qbatch_size);
    const QueriesPromptTokens qbatch_prompts(&queries_prompt[qbatch_start],
                                             qbatch_size);
    QueriesPos qbatch_pos(&queries_pos[qbatch_start], qbatch_size);
    const QueriesPos qbatch_prefix_end(&queries_prefix_end[qbatch_start],
                                       qbatch_size);
    const KVCaches qbatch_kv(&kv_caches[qbatch_start], qbatch_size);
    GenerateT(qbatch_start, qbatch_prompts, qbatch_pos, qbatch_prefix_end,
              config, runtime_config, weights, activations, qbatch_kv, env,
              timing_info);
  }
}

void GenerateImageTokensT(const ModelConfig& config,
                          const RuntimeConfig& runtime_config,
                          const ModelWeightsPtrs& weights, const Image& image,
                          ImageTokens& image_tokens, MatMulEnv& env) {
  if (config.vit_config.layer_configs.empty()) {
    HWY_ABORT("Model does not support generating image tokens.");
  }
  RuntimeConfig prefill_runtime_config = runtime_config;
  ModelConfig vit_config = GetVitConfig(config);
  prefill_runtime_config.prefill_tbatch_size =
      vit_config.seq_len / (vit_config.pool_dim * vit_config.pool_dim);
  Activations prefill_activations(vit_config, vit_config.seq_len, env.row_ptrs);
  // Weights are for the full PaliGemma model, not just the ViT part.
  PrefillVit(config, weights, prefill_runtime_config, image, image_tokens,
             prefill_activations, env);
}

// NOLINTNEXTLINE(google-readability-namespace-comments)
}  // namespace HWY_NAMESPACE
}  // namespace gcpp
HWY_AFTER_NAMESPACE();

#if HWY_ONCE
namespace gcpp {
HWY_EXPORT(GenerateSingleT);
HWY_EXPORT(GenerateBatchT);
HWY_EXPORT(GenerateImageTokensT);

MatMulEnv MakeMatMulEnv(const ThreadingArgs& threading_args) {
  ThreadingContext::SetArgs(threading_args);
  return MatMulEnv(ThreadingContext::Get());
}

Gemma::Gemma(const LoaderArgs& loader, const InferenceArgs& inference,
             MatMulEnv& env)
    : env_(env),
      reader_(loader.weights),
      model_(reader_, loader.tokenizer, loader.wrapping),
      weights_(model_.Config()),
      chat_template_(model_.Tokenizer(), model_.Config().model) {
  weights_.ReadFromBlobs(model_, reader_, loader, inference, mat_owners_,
                         env.ctx.pools.Pool());
  reader_.CloseFile();
}

Gemma::~Gemma() = default;

void Gemma::Save(const Path& weights_path, hwy::ThreadPool& pool) const {
  BlobWriter writer;
  const std::vector<uint32_t> serialized_mat_ptrs =
      weights_.AddTensorDataToWriter(writer);
  WriteSingleFile(model_.Config(), model_.Tokenizer(), serialized_mat_ptrs,
                  writer, env_.ctx.pools.Pool(), weights_path);
}

void Gemma::Generate(const RuntimeConfig& runtime_config,
                     const PromptTokens& prompt, size_t pos, size_t prefix_end,
                     KVCache& kv_cache, TimingInfo& timing_info) const {
  env_.ctx.pools.MaybeStartSpinning(runtime_config.use_spinning);

  HWY_DYNAMIC_DISPATCH(GenerateSingleT)(prompt, pos, prefix_end,
                                        model_.Config(), runtime_config,
                                        weights_, kv_cache, env_, timing_info);

  env_.ctx.pools.MaybeStopSpinning(runtime_config.use_spinning);
}

void Gemma::GenerateBatch(const RuntimeConfig& runtime_config,
                          const QueriesPromptTokens& queries_prompt,
                          const QueriesPos& queries_pos,
                          const QueriesPos& queries_prefix_end,
                          const KVCaches& kv_caches,
                          TimingInfo& timing_info) const {
  // If we did not get passed prefix ends (size 0), assume 0 and pass that on.
  QueriesPos queries_prefix_end_or_zeros = queries_prefix_end;
  std::vector<size_t> prefix_end_vec;
  if (queries_prefix_end.size() == 0) {  // hwy::Span lacks empty()
    prefix_end_vec.resize(queries_prompt.size(), 0);
    queries_prefix_end_or_zeros =
        QueriesPos(prefix_end_vec.data(), prefix_end_vec.size());
  }

  env_.ctx.pools.MaybeStartSpinning(runtime_config.use_spinning);

  HWY_DYNAMIC_DISPATCH(GenerateBatchT)(
      queries_prompt, queries_pos, queries_prefix_end_or_zeros, model_.Config(),
      runtime_config, weights_, kv_caches, env_, timing_info);

  env_.ctx.pools.MaybeStopSpinning(runtime_config.use_spinning);
}

void Gemma::GenerateImageTokens(const RuntimeConfig& runtime_config,
                                const Image& image,
                                ImageTokens& image_tokens) const {
  env_.ctx.pools.MaybeStartSpinning(runtime_config.use_spinning);

  HWY_DYNAMIC_DISPATCH(GenerateImageTokensT)(
      model_.Config(), runtime_config, weights_, image, image_tokens, env_);

  env_.ctx.pools.MaybeStopSpinning(runtime_config.use_spinning);
}

}  // namespace gcpp
#endif  // HWY_ONCE