model: add GLM 5.2 Indexer support (#25407)

* Start building graph - reuse deepseek32

* Enable kv cache and rotation for glm_dsa architecture

Just follow Deepseek 3.2 for now.

* Reuse prev_top_k for "shared" indexer layers

* GLM 5.2 uses LLAMA_ROPE_TYPE_NORM for the indexer.

This is transformers' `apply_rotary_pos_emb_interleave`

* Default indexer types to GLM pattern

Previous converted GGUFs like https://huggingface.co/unsloth/GLM-5.2-GGUF write indexer weights to _all_ layers, even if they are only required for "full" types. This PR relies on a new key "%s.attention.indexer.types"; if absent, it will use the default GLM 5.2 schedule as defined in https://huggingface.co/zai-org/GLM-5.2/blob/main/config.json#L26.

Note that conversion is not saving this key yet.

* Save indexer types to gguf, restore on load

* Use ggml_lightning_indexer when cparams.fused_lid

Co-authored-by: fairydreaming <166155368+fairydreaming@users.noreply.github.com>

* GLM 5 and 5.1 use full indexers

Co-authored-by: fairydreaming <166155368+fairydreaming@users.noreply.github.com>

* Fix indentation

* Ensure array is zero-filled

* Prefer explicit std::fill

* Assert prev_top_k exists for shared indexer

---------

Co-authored-by: fairydreaming <166155368+fairydreaming@users.noreply.github.com>

Author: pcuenca

conversion/glm.py

@@ -237,6 +237,9 @@ def set_gguf_parameters(self):
         self.gguf_writer.add_indexer_head_count(self.hparams["index_n_heads"])
         self.gguf_writer.add_indexer_key_length(self.hparams["index_head_dim"])
         self.gguf_writer.add_indexer_top_k(self.hparams["index_topk"])
+        if (indexer_types := self.hparams.get("indexer_types")) is not None:
+            indexer_types = [t == "full" for t in indexer_types]
+            self.gguf_writer.add_indexer_types(indexer_types)
 
 
 @ModelBase.register("SolarOpenForCausalLM")

gguf-py/gguf/constants.py

@@ -200,6 +200,7 @@ class Indexer:
             HEAD_COUNT = "{arch}.attention.indexer.head_count"
             KEY_LENGTH = "{arch}.attention.indexer.key_length"
             TOP_K      = "{arch}.attention.indexer.top_k"
+            TYPES      = "{arch}.attention.indexer.types"
 
     class HyperConnection:
         COUNT                = "{arch}.hyper_connection.count"

gguf-py/gguf/gguf_writer.py

@@ -793,6 +793,10 @@ def add_indexer_key_length(self, length: int) -> None:
     def add_indexer_top_k(self, top_k: int) -> None:
         self.add_uint32(Keys.Attention.Indexer.TOP_K.format(arch=self.arch), top_k)
 
+    def add_indexer_types(self, value: Sequence[bool]) -> None:
+        key = Keys.Attention.Indexer.TYPES.format(arch=self.arch)
+        self.add_array(key, value)
+
     def add_max_alibi_bias(self, bias: float) -> None:
         self.add_float32(Keys.Attention.MAX_ALIBI_BIAS.format(arch=self.arch), bias)
 

src/llama-arch.cpp

@@ -253,6 +253,7 @@ static const std::map<llm_kv, const char *> LLM_KV_NAMES = {
     { LLM_KV_ATTENTION_INDEXER_HEAD_COUNT,           "%s.attention.indexer.head_count"           },
     { LLM_KV_ATTENTION_INDEXER_KEY_LENGTH,           "%s.attention.indexer.key_length"           },
     { LLM_KV_ATTENTION_INDEXER_TOP_K,                "%s.attention.indexer.top_k"                },
+    { LLM_KV_ATTENTION_INDEXER_TYPES,                "%s.attention.indexer.types"                },
     { LLM_KV_ATTENTION_OUTPUT_GROUP_COUNT,           "%s.attention.output_group_count"           },
     { LLM_KV_ATTENTION_OUTPUT_LORA_RANK,             "%s.attention.output_lora_rank"             },
     { LLM_KV_ATTENTION_COMPRESS_ROPE_FREQ_BASE,      "%s.attention.compress_rope_freq_base"      },

src/llama-arch.h

@@ -258,6 +258,7 @@ enum llm_kv {
     LLM_KV_ATTENTION_INDEXER_HEAD_COUNT,
     LLM_KV_ATTENTION_INDEXER_KEY_LENGTH,
     LLM_KV_ATTENTION_INDEXER_TOP_K,
+    LLM_KV_ATTENTION_INDEXER_TYPES,
     LLM_KV_ATTENTION_OUTPUT_GROUP_COUNT,
     LLM_KV_ATTENTION_OUTPUT_LORA_RANK,
     LLM_KV_ATTENTION_COMPRESS_ROPE_FREQ_BASE,

src/llama-hparams.cpp

@@ -248,6 +248,14 @@ bool llama_hparams::is_mla() const {
     return n_embd_head_k_mla_impl != 0 && n_embd_head_v_mla_impl != 0;
 }
 
+bool llama_hparams::is_indexer_full(uint32_t il) const {
+    if (il < n_layer()) {
+        return is_indexer_full_impl[il];
+    }
+
+    GGML_ABORT("%s: il (%u) out of bounds (n_layer: %u)\n", __func__, il, n_layer());
+}
+
 uint32_t llama_hparams::n_embd_head_k_mla() const {
     return is_mla() ? n_embd_head_k_mla_impl : n_embd_head_k();
 }

src/llama-hparams.h

@@ -227,6 +227,10 @@ struct llama_hparams {
     uint32_t indexer_head_size = 0;
     uint32_t indexer_top_k     = 0;
 
+    // Indexer is "full" (1) or "shared" (0)
+    // Shared indexers reuse top-k from previous full layer
+    std::array<uint32_t, LLAMA_MAX_LAYERS> is_indexer_full_impl;
+
     // DeepSeek-V4
     uint32_t dsv4_o_group_count        = 0;
     uint32_t dsv4_o_lora_rank          = 0;
@@ -302,6 +306,8 @@ struct llama_hparams {
 
     bool is_swa(uint32_t il) const;
 
+    bool is_indexer_full(uint32_t il) const;
+
     void set_recr_pattern(uint32_t n_pattern, bool dense_first = false);
 
     // whether or not the given layer is recurrent (for hybrid models)

src/llama-kv-cache.cpp

@@ -323,7 +323,7 @@ llama_kv_cache::llama_kv_cache(
             hparams.n_embd_head_k() % 64 == 0;
 
         // always create Hadamard rotation tensors for DeepSeek lightning indexers
-        if ((model.arch == LLM_ARCH_DEEPSEEK32 || model.arch == LLM_ARCH_DEEPSEEK4) &&
+        if ((model.arch == LLM_ARCH_DEEPSEEK32 || model.arch == LLM_ARCH_DEEPSEEK4 || model.arch == LLM_ARCH_GLM_DSA) &&
                 hparams.n_embd_head_k_full == hparams.indexer_head_size) {
             attn_rot_k = true;
         }

src/llama-model-saver.cpp

@@ -281,6 +281,7 @@ void llama_model_saver::add_kv_from_model() {
     add_kv(LLM_KV_ATTENTION_INDEXER_HEAD_COUNT,      hparams.indexer_n_head);
     add_kv(LLM_KV_ATTENTION_INDEXER_KEY_LENGTH,      hparams.indexer_head_size);
     add_kv(LLM_KV_ATTENTION_INDEXER_TOP_K,           hparams.indexer_top_k);
+    add_kv(LLM_KV_ATTENTION_INDEXER_TYPES,           hparams.is_indexer_full_impl, true);
     add_kv(LLM_KV_ATTENTION_RECURRENT_LAYERS,        hparams.is_recr_impl, true);
 
     const float rope_scaling_factor = hparams.rope_freq_scale_train == 1.0f ? 0.0f : 1.0f/hparams.rope_freq_scale_train;

src/llama-model.cpp

@@ -1129,6 +1129,7 @@ void llama_model_base::load_hparams(llama_model_loader & ml) {
     std::fill(hparams.rope_sections.begin(), hparams.rope_sections.end(), 0);
     std::fill(hparams.is_swa_impl.begin(),   hparams.is_swa_impl.end(), 0);
     std::fill(hparams.is_recr_impl.begin(),  hparams.is_recr_impl.end(),  llm_arch_is_recurrent(ml.get_arch()) ? 1 : 0);
+    std::fill(hparams.is_indexer_full_impl.begin(), hparams.is_indexer_full_impl.end(), 0);
 
     std::fill(hparams.xielu_alpha_n.begin(), hparams.xielu_alpha_n.end(), 0.0f);
     std::fill(hparams.xielu_alpha_p.begin(), hparams.xielu_alpha_p.end(), 0.0f);
@@ -2065,6 +2066,7 @@ llama_memory_i * llama_model::create_memory(const llama_memory_params & params,
                 res = nullptr;
             } break;
         case LLM_ARCH_DEEPSEEK32:
+        case LLM_ARCH_GLM_DSA:
             {
                 res = new llama_kv_cache_dsa(
                         *this,