mirror of
https://github.com/LostRuins/koboldcpp.git
synced 2026-09-18 16:55:14 +02:00
Merge branch 'upstream' into concedo_experimental
# Conflicts: # .github/workflows/build.yml # AUTHORS # ci/run.sh # docs/backend/SYCL.md # docs/build.md # docs/multimodal/minicpmo2.6.md # docs/multimodal/minicpmo4.0.md # docs/multimodal/minicpmv2.5.md # docs/multimodal/minicpmv2.6.md # docs/multimodal/minicpmv4.0.md # docs/multimodal/minicpmv4.5.md # docs/ops.md # docs/ops/SYCL.csv # docs/speculative.md # examples/deprecation-warning/README.md # examples/deprecation-warning/deprecation-warning.cpp # examples/model-conversion/Makefile # examples/model-conversion/scripts/causal/convert-model.sh # ggml/include/ggml-cann.h # ggml/src/ggml-cann/acl_tensor.cpp # ggml/src/ggml-cann/acl_tensor.h # ggml/src/ggml-cann/aclnn_ops.cpp # ggml/src/ggml-cann/aclnn_ops.h # ggml/src/ggml-cann/common.h # ggml/src/ggml-cann/ggml-cann.cpp # ggml/src/ggml-metal/CMakeLists.txt # ggml/src/ggml-opencl/ggml-opencl.cpp # ggml/src/ggml-opencl/kernels/concat.cl # ggml/src/ggml-opencl/kernels/repeat.cl # ggml/src/ggml-opencl/kernels/scale.cl # ggml/src/ggml-opencl/kernels/tanh.cl # ggml/src/ggml-sycl/CMakeLists.txt # ggml/src/ggml-sycl/dpct/helper.hpp # ggml/src/ggml-sycl/ggml-sycl.cpp # ggml/src/ggml-sycl/outprod.cpp # ggml/src/ggml-sycl/rope.cpp # ggml/src/ggml-sycl/wkv.cpp # src/llama-vocab.cpp # tests/test-autorelease.cpp # tests/test-backend-ops.cpp # tools/cvector-generator/pca.hpp # tools/export-lora/export-lora.cpp # tools/perplexity/README.md
This commit is contained in:
Concedo
+1
-1
@@ -233,7 +233,7 @@ int32_t llm_chat_apply_template(
|
||||
llm_chat_template tmpl,
|
||||
const std::vector<const llama_chat_message *> & chat,
|
||||
std::string & dest, bool add_ass) {
|
||||
// Taken from the research: https://github.com/ggerganov/llama.cpp/issues/5527
|
||||
// Taken from the research: https://github.com/ggml-org/llama.cpp/issues/5527
|
||||
std::stringstream ss;
|
||||
if (tmpl == LLM_CHAT_TEMPLATE_CHATML) {
|
||||
// chatml template
|
||||
|
||||
@@ -325,6 +325,7 @@ llama_context::llama_context(
|
||||
auto dev_type = ggml_backend_dev_type(ggml_backend_get_device(backend.get()));
|
||||
if (dev_type == GGML_BACKEND_DEVICE_TYPE_CPU) {
|
||||
// ignore CPU backend
|
||||
// TODO: should we ignore ACCEL types too?
|
||||
continue;
|
||||
}
|
||||
auto * dev = ggml_backend_get_device(backend.get());
|
||||
|
||||
+1
-1
@@ -195,7 +195,7 @@ struct llama_hparams {
|
||||
uint32_t n_deepstack_layers = 0;
|
||||
|
||||
// needed by encoder-decoder models (e.g. T5, FLAN-T5)
|
||||
// ref: https://github.com/ggerganov/llama.cpp/pull/8141
|
||||
// ref: https://github.com/ggml-org/llama.cpp/pull/8141
|
||||
llama_token dec_start_token_id = LLAMA_TOKEN_NULL;
|
||||
uint32_t dec_n_layer = 0;
|
||||
|
||||
|
||||
+7
-4
@@ -94,7 +94,7 @@ static_assert(std::is_trivially_copyable<llm_symbol>::value, "llm_symbol is not
|
||||
//
|
||||
// SPM tokenizer
|
||||
// original implementation:
|
||||
// https://github.com/ggerganov/llama.cpp/commit/074bea2eb1f1349a0118239c4152914aecaa1be4
|
||||
// https://github.com/ggml-org/llama.cpp/commit/074bea2eb1f1349a0118239c4152914aecaa1be4
|
||||
//
|
||||
|
||||
struct llm_bigram_spm {
|
||||
@@ -510,7 +510,7 @@ struct llm_tokenizer_bpe : llm_tokenizer {
|
||||
// original regex from tokenizer.json
|
||||
//"(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?\\p{L}+|\\p{N}{1,3}| ?[^\\s\\p{L}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
|
||||
|
||||
// adapted: https://github.com/ggerganov/llama.cpp/pull/6920#issuecomment-2080233989
|
||||
// adapted: https://github.com/ggml-org/llama.cpp/pull/6920#issuecomment-2080233989
|
||||
"(?:'[sS]|'[tT]|'[rR][eE]|'[vV][eE]|'[mM]|'[lL][lL]|'[dD])|[^\\r\\n\\p{L}\\p{N}]?\\p{L}+|\\p{N}{1,3}| ?[^\\s\\p{L}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
|
||||
};
|
||||
break;
|
||||
@@ -2501,6 +2501,7 @@ void llama_vocab::impl::load(llama_model_loader & ml, const LLM_KV & kv) {
|
||||
|| t.first == "<PRE>"
|
||||
|| t.first == "▁<PRE>" // CodeLlama
|
||||
|| t.first == "<|code_prefix|>" // GLM-4.5
|
||||
|| t.first == "<|prefix|>" // Falcon-H1-Tiny-Coder
|
||||
) {
|
||||
special_fim_pre_id = t.second;
|
||||
if ((attr & LLAMA_TOKEN_ATTR_CONTROL) == 0) {
|
||||
@@ -2521,6 +2522,7 @@ void llama_vocab::impl::load(llama_model_loader & ml, const LLM_KV & kv) {
|
||||
|| t.first == "<SUF>"
|
||||
|| t.first == "▁<SUF>" // CodeLlama
|
||||
|| t.first == "<|code_suffix|>" // GLM-4.5
|
||||
|| t.first == "<|suffix|>" // Falcon-H1-Tiny-Coder
|
||||
) {
|
||||
special_fim_suf_id = t.second;
|
||||
if ((attr & LLAMA_TOKEN_ATTR_CONTROL) == 0) {
|
||||
@@ -2541,6 +2543,7 @@ void llama_vocab::impl::load(llama_model_loader & ml, const LLM_KV & kv) {
|
||||
|| t.first == "<MID>"
|
||||
|| t.first == "▁<MID>" // CodeLlama
|
||||
|| t.first == "<|code_middle|>" // GLM-4.5
|
||||
|| t.first == "<|middle|>" // Falcon-H1-Tiny-Coder
|
||||
) {
|
||||
special_fim_mid_id = t.second;
|
||||
if ((attr & LLAMA_TOKEN_ATTR_CONTROL) == 0) {
|
||||
@@ -2629,7 +2632,7 @@ void llama_vocab::impl::load(llama_model_loader & ml, const LLM_KV & kv) {
|
||||
|
||||
// maintain a list of tokens that cause end-of-generation
|
||||
// this is currently determined based on the token text, which is obviously not ideal
|
||||
// ref: https://github.com/ggerganov/llama.cpp/issues/9606
|
||||
// ref: https://github.com/ggml-org/llama.cpp/issues/9606
|
||||
special_eog_ids.clear();
|
||||
|
||||
if (special_fim_pad_id != LLAMA_TOKEN_NULL && special_eog_ids.count(special_fim_pad_id) == 0) {
|
||||
@@ -3355,7 +3358,7 @@ int32_t llama_vocab::impl::token_to_piece(llama_token token, char * buf, int32_t
|
||||
{
|
||||
return llama_token_to_piece_old(vocab, token, buf, length);
|
||||
}
|
||||
// ref: https://github.com/ggerganov/llama.cpp/pull/7587#discussion_r1620983843
|
||||
// ref: https://github.com/ggml-org/llama.cpp/pull/7587#discussion_r1620983843
|
||||
static const int attr_special = LLAMA_TOKEN_ATTR_UNKNOWN | LLAMA_TOKEN_ATTR_CONTROL;
|
||||
const llama_token_attr attr = token_get_attr(token);
|
||||
if (!special && (attr & attr_special)) {
|
||||
|
||||
@@ -14,7 +14,7 @@ llm_build_deepseek2::llm_build_deepseek2(const llama_model & model, const llm_gr
|
||||
const uint32_t kv_lora_rank = hparams.n_lora_kv;
|
||||
|
||||
// We have to pre-scale kq_scale and attn_factor to make the YaRN RoPE work correctly.
|
||||
// See https://github.com/ggerganov/llama.cpp/discussions/7416 for detailed explanation.
|
||||
// See https://github.com/ggml-org/llama.cpp/discussions/7416 for detailed explanation.
|
||||
// And also: https://github.com/ggml-org/llama.cpp/pull/17945 [TAG_DEEPSEEK2_YARN_LOG_MUL_FIX]
|
||||
|
||||
// first cancel the adjustment from llama_hparams::yarn_attn_factor_adjust to get the original attn_factor
|
||||
|
||||
Reference in New Issue
Block a user