pali112/ollama
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

update patches

Browse Source
This commit is contained in:
inforithmics
2025-12-26 16:42:31 +01:00
parent 18fdcc94e5
commit c7d1f258aa
20 changed files with 100 additions and 110 deletions
Download Patch File Download Diff File Expand all files Collapse all files

52
llama/patches/0004-solar-pro.patch
View File

@@ -19,10 +19,10 @@ adds support for the Solar Pro architecture
create mode 100644 src/models/solar.cpp
diff --git a/src/CMakeLists.txt b/src/CMakeLists.txt
index 4192af7c0..bd44d73e7 100644
index 1e155534b..159f429e8 100644
--- a/src/CMakeLists.txt
+++ b/src/CMakeLists.txt
@@ -125,6 +125,7 @@ add_library(llama
@@ -127,6 +127,7 @@ add_library(llama
models/seed-oss.cpp
models/smallthinker.cpp
models/smollm3.cpp
@@ -31,10 +31,10 @@ index 4192af7c0..bd44d73e7 100644
models/starcoder.cpp
models/starcoder2.cpp
diff --git a/src/llama-arch.cpp b/src/llama-arch.cpp
index 8caf80afc..2ce8ffec0 100644
index 75013d8d3..22b30bfcc 100644
--- a/src/llama-arch.cpp
+++ b/src/llama-arch.cpp
@@ -87,6 +87,7 @@ static const std::map<llm_arch, const char *> LLM_ARCH_NAMES = {
@@ -88,6 +88,7 @@ static const std::map<llm_arch, const char *> LLM_ARCH_NAMES = {
{ LLM_ARCH_GRANITE_MOE, "granitemoe" },
{ LLM_ARCH_GRANITE_HYBRID, "granitehybrid" },
{ LLM_ARCH_CHAMELEON, "chameleon" },
@@ -42,7 +42,7 @@ index 8caf80afc..2ce8ffec0 100644
{ LLM_ARCH_WAVTOKENIZER_DEC, "wavtokenizer-dec" },
{ LLM_ARCH_PLM, "plm" },
{ LLM_ARCH_BAILINGMOE, "bailingmoe" },
@@ -208,6 +209,7 @@ static const std::map<llm_kv, const char *> LLM_KV_NAMES = {
@@ -212,6 +213,7 @@ static const std::map<llm_kv, const char *> LLM_KV_NAMES = {
{ LLM_KV_ATTENTION_OUTPUT_SCALE, "%s.attention.output_scale" },
{ LLM_KV_ATTENTION_TEMPERATURE_LENGTH, "%s.attention.temperature_length" },
{ LLM_KV_ATTENTION_TEMPERATURE_SCALE, "%s.attention.temperature_scale" },
@@ -50,7 +50,7 @@ index 8caf80afc..2ce8ffec0 100644
{ LLM_KV_ATTENTION_KEY_LENGTH_MLA, "%s.attention.key_length_mla" },
{ LLM_KV_ATTENTION_VALUE_LENGTH_MLA, "%s.attention.value_length_mla" },
@@ -339,6 +341,7 @@ static const std::map<llm_tensor, const char *> LLM_TENSOR_NAMES = {
@@ -344,6 +346,7 @@ static const std::map<llm_tensor, const char *> LLM_TENSOR_NAMES = {
{ LLM_TENSOR_ATTN_QKV, "blk.%d.attn_qkv" },
{ LLM_TENSOR_LAYER_OUT_NORM, "blk.%d.layer_output_norm" },
{ LLM_TENSOR_ATTN_OUT_NORM, "blk.%d.attn_output_norm" },
@@ -58,7 +58,7 @@ index 8caf80afc..2ce8ffec0 100644
{ LLM_TENSOR_POS_EMBD, "position_embd" },
{ LLM_TENSOR_FFN_ACT, "blk.%d.ffn.act" },
{ LLM_TENSOR_TOKEN_EMBD_NORM, "token_embd_norm" },
@@ -2176,6 +2179,22 @@ static std::set<llm_tensor> llm_get_tensor_names(llm_arch arch) {
@@ -2217,6 +2220,22 @@ static std::set<llm_tensor> llm_get_tensor_names(llm_arch arch) {
return {
LLM_TENSOR_TOKEN_EMBD,
};
@@ -81,7 +81,7 @@ index 8caf80afc..2ce8ffec0 100644
default:
GGML_ABORT("unknown architecture for tensor mapping");
}
@@ -2344,6 +2363,7 @@ static const std::map<llm_tensor, llm_tensor_info> LLM_TENSOR_INFOS = {
@@ -2385,6 +2404,7 @@ static const std::map<llm_tensor, llm_tensor_info> LLM_TENSOR_INFOS = {
{LLM_TENSOR_LAUREL_POST_NORM, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL}},
// this tensor is loaded for T5, but never used
{LLM_TENSOR_DEC_CROSS_ATTN_REL_B, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_NONE}},
@@ -90,10 +90,10 @@ index 8caf80afc..2ce8ffec0 100644
{LLM_TENSOR_POS_NET_NORM, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL}},
{LLM_TENSOR_POS_NET_NORM1, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL}},
diff --git a/src/llama-arch.h b/src/llama-arch.h
index 6cbf9b1f8..14d461c76 100644
index 27bdedc83..06c903bb6 100644
--- a/src/llama-arch.h
+++ b/src/llama-arch.h
@@ -91,6 +91,7 @@ enum llm_arch {
@@ -92,6 +92,7 @@ enum llm_arch {
LLM_ARCH_GRANITE_MOE,
LLM_ARCH_GRANITE_HYBRID,
LLM_ARCH_CHAMELEON,
@@ -101,7 +101,7 @@ index 6cbf9b1f8..14d461c76 100644
LLM_ARCH_WAVTOKENIZER_DEC,
LLM_ARCH_PLM,
LLM_ARCH_BAILINGMOE,
@@ -212,6 +213,7 @@ enum llm_kv {
@@ -216,6 +217,7 @@ enum llm_kv {
LLM_KV_ATTENTION_OUTPUT_SCALE,
LLM_KV_ATTENTION_TEMPERATURE_LENGTH,
LLM_KV_ATTENTION_TEMPERATURE_SCALE,
@@ -109,7 +109,7 @@ index 6cbf9b1f8..14d461c76 100644
LLM_KV_ATTENTION_KEY_LENGTH_MLA,
LLM_KV_ATTENTION_VALUE_LENGTH_MLA,
@@ -465,6 +467,7 @@ enum llm_tensor {
@@ -470,6 +472,7 @@ enum llm_tensor {
LLM_TENSOR_ENC_OUTPUT_NORM,
LLM_TENSOR_CLS,
LLM_TENSOR_CLS_OUT,
@@ -137,7 +137,7 @@ index fe1fa4341..aabff2f06 100644
if (il < n_layer) {
return swa_layers[il];
diff --git a/src/llama-hparams.h b/src/llama-hparams.h
index f6e95b5d2..c6e673276 100644
index 42def73f0..d3c53b5f2 100644
--- a/src/llama-hparams.h
+++ b/src/llama-hparams.h
@@ -65,6 +65,8 @@ struct llama_hparams {
@@ -149,7 +149,7 @@ index f6e95b5d2..c6e673276 100644
uint32_t n_layer_dense_lead = 0;
uint32_t n_lora_q = 0;
uint32_t n_lora_kv = 0;
@@ -259,6 +261,9 @@ struct llama_hparams {
@@ -260,6 +262,9 @@ struct llama_hparams {
uint32_t n_pos_per_embd() const;
@@ -160,10 +160,10 @@ index f6e95b5d2..c6e673276 100644
bool has_kv(uint32_t il) const;
diff --git a/src/llama-model-loader.cpp b/src/llama-model-loader.cpp
index ca2ea2461..8916a6242 100644
index 5003b4fbf..243b296b5 100644
--- a/src/llama-model-loader.cpp
+++ b/src/llama-model-loader.cpp
@@ -466,7 +466,7 @@ namespace GGUFMeta {
@@ -489,7 +489,7 @@ namespace GGUFMeta {
template bool llama_model_loader::get_key_or_arr<std::array<int, 4>>(enum llm_kv kid, std::array<int, 4> & result, uint32_t n, bool required);
template bool llama_model_loader::get_key_or_arr<std::array<uint32_t, 512>>(enum llm_kv kid, std::array<uint32_t, 512> & result, uint32_t n, bool required);
template bool llama_model_loader::get_key_or_arr<std::array<float, 512>>(enum llm_kv kid, std::array<float, 512> & result, uint32_t n, bool required);
@@ -173,10 +173,10 @@ index ca2ea2461..8916a6242 100644
llama_model_loader::llama_model_loader(
const std::string & fname,
diff --git a/src/llama-model.cpp b/src/llama-model.cpp
index ae8207ee1..00cd579e0 100644
index 69075742c..bdee9b6e6 100644
--- a/src/llama-model.cpp
+++ b/src/llama-model.cpp
@@ -1995,6 +1995,21 @@ void llama_model::load_hparams(llama_model_loader & ml) {
@@ -2028,6 +2028,21 @@ void llama_model::load_hparams(llama_model_loader & ml) {
default: type = LLM_TYPE_UNKNOWN;
}
} break;
@@ -198,7 +198,7 @@ index ae8207ee1..00cd579e0 100644
case LLM_ARCH_WAVTOKENIZER_DEC:
{
ml.get_key(LLM_KV_ATTENTION_LAYERNORM_EPS, hparams.f_norm_eps);
@@ -5429,6 +5444,34 @@ bool llama_model::load_tensors(llama_model_loader & ml) {
@@ -5510,6 +5525,34 @@ bool llama_model::load_tensors(llama_model_loader & ml) {
layer.ffn_norm = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, 0);
@@ -233,7 +233,7 @@ index ae8207ee1..00cd579e0 100644
layer.ffn_gate = create_tensor(tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd, n_ff}, 0);
layer.ffn_down = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), { n_ff, n_embd}, 0);
layer.ffn_up = create_tensor(tn(LLM_TENSOR_FFN_UP, "weight", i), {n_embd, n_ff}, 0);
@@ -7534,6 +7577,10 @@ ggml_cgraph * llama_model::build_graph(const llm_graph_params & params) const {
@@ -7664,6 +7707,10 @@ ggml_cgraph * llama_model::build_graph(const llm_graph_params & params) const {
{
llm = std::make_unique<llm_build_chameleon>(*this, params);
} break;
@@ -244,7 +244,7 @@ index ae8207ee1..00cd579e0 100644
case LLM_ARCH_WAVTOKENIZER_DEC:
{
llm = std::make_unique<llm_build_wavtokenizer_dec>(*this, params);
@@ -7798,6 +7845,7 @@ llama_rope_type llama_model_rope_type(const llama_model * model) {
@@ -7932,6 +7979,7 @@ llama_rope_type llama_model_rope_type(const llama_model * model) {
case LLM_ARCH_GRANITE_MOE:
case LLM_ARCH_GRANITE_HYBRID:
case LLM_ARCH_CHAMELEON:
@@ -253,10 +253,10 @@ index ae8207ee1..00cd579e0 100644
case LLM_ARCH_NEO_BERT:
case LLM_ARCH_SMOLLM3:
diff --git a/src/llama-model.h b/src/llama-model.h
index c6eb95318..b378b23ec 100644
index 9c00eec75..858af51bb 100644
--- a/src/llama-model.h
+++ b/src/llama-model.h
@@ -76,6 +76,7 @@ enum llm_type {
@@ -79,6 +79,7 @@ enum llm_type {
LLM_TYPE_15B,
LLM_TYPE_16B,
LLM_TYPE_20B,
@@ -264,7 +264,7 @@ index c6eb95318..b378b23ec 100644
LLM_TYPE_26B,
LLM_TYPE_27B,
LLM_TYPE_30B,
@@ -405,6 +406,8 @@ struct llama_layer {
@@ -409,6 +410,8 @@ struct llama_layer {
struct ggml_tensor * ffn_act_beta = nullptr;
struct ggml_tensor * ffn_act_eps = nullptr;
@@ -274,10 +274,10 @@ index c6eb95318..b378b23ec 100644
struct llama_layer_convnext convnext;
diff --git a/src/models/models.h b/src/models/models.h
index ffb36acc6..6d84a185d 100644
index dd0e286ed..40f61b59d 100644
--- a/src/models/models.h
+++ b/src/models/models.h
@@ -515,6 +515,11 @@ struct llm_build_smollm3 : public llm_graph_context {
@@ -525,6 +525,11 @@ struct llm_build_smollm3 : public llm_graph_context {
llm_build_smollm3(const llama_model & model, const llm_graph_params & params);
};