llama : remove obsolete n_pad from memory instances (#31200)

Every caller passes n_pad = 1 and the only two users clamp it to 256.
Drop the parameter, the members and the trivially true kv_size assert.

Assisted-by: dsh:llama.cpp/ggml-org/Qwen3.8-Flash-Next-GGUF (MacBook, M5 Max, 128GB)
This commit is contained in:
Georgi Gerganov authored and GitHub committed 2026-10-27 07:53:44 +02:00
1 parent 0d5ca60098
commit 7a2e5aba90
19 files changed
+21 -67

No files matched your search

-17
View File
@@ -2842,7 +2842,6 @@ llama_memory_i * llama_model::create_memory(const llama_memory_params & params,
cparams.kv_unified,
cparams.n_ctx_seq,
cparams.n_seq_max,
1,
hparams.n_swa,
hparams.swa_type,
nullptr,
@@ -2870,7 +2869,6 @@ llama_memory_i * llama_model::create_memory(const llama_memory_params & params,
cparams.kv_unified,
cparams.n_ctx_seq,
cparams.n_seq_max,
1,
hparams.n_swa,
hparams.swa_type,
nullptr,
@@ -2896,7 +2894,6 @@ llama_memory_i * llama_model::create_memory(const llama_memory_params & params,
cparams.kv_unified,
cparams.n_ctx_seq,
cparams.n_seq_max,
1,
hparams.n_swa,
hparams.swa_type,
filter_mla,
@@ -2935,7 +2932,6 @@ llama_memory_i * llama_model::create_memory(const llama_memory_params & params,
/* attn_type_v */ params.type_v,
/* attn_v_trans */ !cparams.flash_attn,
/* attn_kv_size */ cparams.n_ctx_seq,
/* attn_n_pad */ 1,
/* attn_n_swa */ hparams.n_swa,
/* attn_swa_type */ hparams.swa_type,
/* recurrent_type_r */ GGML_TYPE_F32,
@@ -2964,7 +2960,6 @@ llama_memory_i * llama_model::create_memory(const llama_memory_params & params,
cparams.kv_unified,
cparams.n_ctx_seq,
cparams.n_seq_max,
1,
hparams.n_swa,
hparams.swa_type,
nullptr,
@@ -2985,7 +2980,6 @@ llama_memory_i * llama_model::create_memory(const llama_memory_params & params,
cparams.kv_unified,
cparams.n_ctx_seq,
cparams.n_seq_max,
1,
hparams.n_swa,
hparams.swa_type,
nullptr,
@@ -3013,7 +3007,6 @@ llama_memory_i * llama_model::create_memory(const llama_memory_params & params,
cparams.kv_unified,
cparams.n_ctx_seq,
cparams.n_seq_max,
1,
hparams.n_swa,
hparams.swa_type,
nullptr,
@@ -3040,7 +3033,6 @@ llama_memory_i * llama_model::create_memory(const llama_memory_params & params,
cparams.n_ctx_seq,
cparams.n_seq_max,
cparams.n_ubatch,
1,
filter_mla,
filter_lid,
nullptr);
@@ -3068,7 +3060,6 @@ llama_memory_i * llama_model::create_memory(const llama_memory_params & params,
cparams.n_ctx_seq,
cparams.n_seq_max,
cparams.n_ubatch,
1,
nullptr,
filter_mtp,
nullptr,
@@ -3086,7 +3077,6 @@ llama_memory_i * llama_model::create_memory(const llama_memory_params & params,
cparams.n_ctx_seq,
cparams.n_seq_max,
cparams.n_ubatch,
1,
cparams.n_rs_seq,
nullptr,
nullptr);
@@ -3110,7 +3100,6 @@ llama_memory_i * llama_model::create_memory(const llama_memory_params & params,
cparams.n_ctx_seq,
cparams.n_seq_max,
cparams.n_ubatch,
1,
nullptr,
nullptr,
nullptr,
@@ -3198,7 +3187,6 @@ llama_memory_i * llama_model::create_memory(const llama_memory_params & params,
/* attn_swa_full */ params.swa_full,
/* attn_kv_size */ cparams.n_ctx_seq,
/* attn_n_ubatch */ cparams.n_ubatch,
/* attn_n_pad */ 1,
/* recurrent_type_r */ GGML_TYPE_F32,
/* recurrent_type_s */ GGML_TYPE_F32,
/* recurrent_rs_size */ std::max((uint32_t) 1, cparams.n_seq_max),
@@ -3217,7 +3205,6 @@ llama_memory_i * llama_model::create_memory(const llama_memory_params & params,
/* attn_type_v */ params.type_v,
/* attn_v_trans */ !cparams.flash_attn,
/* attn_kv_size */ cparams.n_ctx_seq,
/* attn_n_pad */ 1,
/* attn_n_swa */ hparams.n_swa,
/* attn_swa_type */ hparams.swa_type,
/* recurrent_type_k */ GGML_TYPE_F32,
@@ -3238,7 +3225,6 @@ llama_memory_i * llama_model::create_memory(const llama_memory_params & params,
/* attn_type_v */ params.type_v,
/* attn_v_trans */ !cparams.flash_attn,
/* attn_kv_size */ cparams.n_ctx_seq,
/* attn_n_pad */ 1,
/* attn_n_swa */ hparams.n_swa,
/* attn_swa_type */ hparams.swa_type,
/* recurrent_type_k */ GGML_TYPE_F32,
@@ -3310,7 +3296,6 @@ llama_memory_i * llama_model::create_memory(const llama_memory_params & params,
cparams.n_ctx_seq,
cparams.n_seq_max,
cparams.n_ubatch,
1,
mem_other,
filter,
reuse,
@@ -3328,7 +3313,6 @@ llama_memory_i * llama_model::create_memory(const llama_memory_params & params,
cparams.n_ctx_seq,
cparams.n_seq_max,
cparams.n_ubatch,
1,
nullptr,
filter,
reuse,
@@ -3348,7 +3332,6 @@ llama_memory_i * llama_model::create_memory(const llama_memory_params & params,
cparams.kv_unified,
cparams.n_ctx_seq,
cparams.n_seq_max,
1,
hparams.n_swa,
hparams.swa_type,
nullptr,