happyz
/
llama.cpp
mirror of https://github.com/ggerganov/llama.cpp.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

models : fix the attn_factor for mistral3 graphs + improve consistency (#17945) 

* models : fix the attn_factor for mistral3 graphs

* cont : rework attn_factor correction logic

* cont : make deepseek2 consistent

* cont : add TODO

* cont : special-case DSv2

* cont : revert Mistral 3 Large changes

* cont : fix DS2 to use the original attn_factor

* cont : minor comments
This commit is contained in:
Georgi Gerganov 2025-12-12 17:12:40 +02:00 committed by GitHub
parent dcb7d17758
commit 7bed317f53
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
7 changed files with 78 additions and 33 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

8
convert_hf_to_gguf.py
View File

@ -7286,6 +7286,10 @@ class DeepseekV2Model(TextModel):
self.gguf_writer.add_rope_scaling_type(gguf.RopeScalingType.YARN)
self.gguf_writer.add_rope_scaling_factor(rope_scaling["factor"])
self.gguf_writer.add_rope_scaling_orig_ctx_len(rope_scaling["original_max_position_embeddings"])
# [TAG_DEEPSEEK2_YARN_LOG_MUL_FIX]
# note: for legacy reasons, this is not consistent with the other usages of self.gguf_writer.add_rope_scaling_yarn_log_mul
# ref https://github.com/ggml-org/llama.cpp/pull/17945
self.gguf_writer.add_rope_scaling_yarn_log_mul(0.1 * rope_scaling["mscale_all_dim"])
_experts: list[dict[str, Tensor]] | None = None
@ -10041,6 +10045,10 @@ class MistralMoeModel(DeepseekV2Model):
MistralModel.set_mistral_config(self.gguf_writer, self.hparams)
yarn_params = self.hparams["yarn"]
self.gguf_writer.add_attn_temperature_length(yarn_params["original_max_position_embeddings"])
# [TAG_DEEPSEEK2_YARN_LOG_MUL_FIX]
# note: for legacy reasons, this is not consistent with the other usages of self.gguf_writer.add_rope_scaling_yarn_log_mul
# ref https://github.com/ggml-org/llama.cpp/pull/17945
self.gguf_writer.add_rope_scaling_yarn_log_mul(0.1) # mscale_all_dim * 0.1
def modify_tensors(self, data_torch: Tensor, name: str, bid: int | None):

2
src/llama-graph.cpp
View File

@ -574,7 +574,7 @@ llm_graph_context::llm_graph_context(const llm_graph_params & params) :
freq_base (cparams.rope_freq_base),
freq_scale (cparams.rope_freq_scale),
ext_factor (cparams.yarn_ext_factor),
attn_factor (cparams.yarn_attn_factor),
attn_factor (llama_hparams::yarn_attn_factor_adjust(cparams.yarn_attn_factor, cparams.rope_freq_scale, cparams.yarn_ext_factor)),
beta_fast (cparams.yarn_beta_fast),
beta_slow (cparams.yarn_beta_slow),
norm_eps (hparams.f_norm_eps),

12
src/llama-hparams.cpp
View File

@ -1,7 +1,9 @@
#include "llama-hparams.h"
#include "ggml.h"
#include <cassert>
#include <cmath>
void llama_hparams::set_swa_pattern(uint32_t n_pattern, bool dense_first) {
if (dense_first) {
@ -229,3 +231,13 @@ bool llama_hparams::is_masked_swa(uint32_t n_swa, llama_swa_type swa_type, llama
return false;
}
float llama_hparams::yarn_attn_factor_adjust(float attn_factor, float freq_scale, float ext_factor) {
GGML_ASSERT(ext_factor >= 0.0f);
if (ext_factor != 0.0f) {
attn_factor *= 1.0f / (1.0f + 0.1f * logf(1.0f / freq_scale));
}
return attn_factor;
}

9
src/llama-hparams.h
View File

@ -107,6 +107,7 @@ struct llama_hparams {
float rope_freq_base_train_swa;
float rope_freq_scale_train;
float rope_freq_scale_train_swa;
uint32_t n_ctx_orig_yarn;
float rope_yarn_log_mul = 0.0f;
@ -267,7 +268,13 @@ struct llama_hparams {
// TODO: think of a better place for this function
// TODO: pack the SWA params in a struct?
static bool is_masked_swa(uint32_t n_swa, llama_swa_type swa_type, llama_pos p0, llama_pos p1);
// when YARN is applied with yarn_ext_factor != 0.0f, we need to cancel this factor:
// https://github.com/ggml-org/llama.cpp/blob/a81a569577cc38b32558958b048228150be63eae/ggml/src/ggml-cpu/ops.cpp#L5541-L5544
//
// ref: https://github.com/ggml-org/llama.cpp/discussions/7416
// https://github.com/ggml-org/llama.cpp/pull/17945
static float yarn_attn_factor_adjust(float attn_factor, float freq_scale, float ext_factor);
};
static_assert(std::is_trivially_copyable<llama_hparams>::value, "llama_hparams must be trivially copyable");

7
src/llama-kv-cache.cpp
View File

@ -1372,6 +1372,7 @@ ggml_tensor * llama_kv_cache::build_rope_shift(
const auto & yarn_ext_factor = cparams.yarn_ext_factor;
const auto & yarn_beta_fast = cparams.yarn_beta_fast;
const auto & yarn_beta_slow = cparams.yarn_beta_slow;
const auto & yarn_attn_factor = llama_hparams::yarn_attn_factor_adjust(cparams.yarn_attn_factor, cparams.rope_freq_scale, cparams.yarn_ext_factor);
const auto & n_rot = hparams.n_rot;
const auto & rope_type = hparams.rope_type == LLAMA_ROPE_TYPE_MROPE || hparams.rope_type == LLAMA_ROPE_TYPE_IMROPE
@ -1382,12 +1383,6 @@ ggml_tensor * llama_kv_cache::build_rope_shift(
? LLAMA_ROPE_TYPE_NEOX
: hparams.rope_type;
// See llm_build_deepseek2() for why attn_factor has to be scaled for YaRN RoPE to work correctly.
// See https://github.com/ggerganov/llama.cpp/discussions/7416 for detailed explanation.
const float yarn_attn_factor = model.arch == LLM_ARCH_DEEPSEEK2
? 1.0f / (1.0f + 0.1f * logf(1.0f / freq_scale))
: cparams.yarn_attn_factor;
ggml_tensor * tmp;
if (ggml_is_quantized(cur->type)) {

49
src/llama-model.cpp
View File

@ -1635,7 +1635,12 @@ void llama_model::load_hparams(llama_model_loader & ml) {
// that have no expert_gating_func model parameter set
hparams.expert_gating_func = LLAMA_EXPERT_GATING_FUNC_TYPE_SOFTMAX;
}
ml.get_key(LLM_KV_ROPE_SCALING_YARN_LOG_MUL, hparams.rope_yarn_log_mul, false);
if (ml.get_key(LLM_KV_ROPE_SCALING_YARN_LOG_MUL, hparams.rope_yarn_log_mul, 0.0f)) {
// [TAG_DEEPSEEK2_YARN_LOG_MUL_FIX]
// cancel the factor from the convert script
hparams.rope_yarn_log_mul /= 0.1f;
}
// (optional) temperature tuning - used by mistral-large
ml.get_key(LLM_KV_ATTENTION_TEMPERATURE_SCALE, hparams.f_attn_temp_scale, false);
@ -2269,7 +2274,7 @@ void llama_model::load_hparams(llama_model_loader & ml) {
ml.get_key(LLM_KV_ROPE_SCALING_YARN_BETA_FAST, hparams.yarn_beta_fast, false);
ml.get_key(LLM_KV_ROPE_SCALING_YARN_BETA_SLOW, hparams.yarn_beta_slow, false);
ml.get_key(LLM_KV_ROPE_SCALING_YARN_LOG_MUL, hparams.rope_yarn_log_mul, false);
ml.get_key(LLM_KV_ROPE_SCALING_YARN_LOG_MUL, hparams.rope_yarn_log_mul, 0.0f);
// TODO: maybe add n_attn_temp_floor_scale as a separate KV?
if (hparams.f_attn_temp_scale != 0.0f) {
@ -2279,18 +2284,6 @@ void llama_model::load_hparams(llama_model_loader & ml) {
}
}
// TODO: this seems to be correct with the case of mscale == mscale_all_dims == 1.0f
// but may need further verification with other values
if (hparams.rope_yarn_log_mul != 0.0f) {
float factor = 1.0f / hparams.rope_freq_scale_train;
float mscale = 1.0f;
float mscale_all_dims = hparams.rope_yarn_log_mul;
static auto get_mscale = [](float scale, float mscale) {
return scale <= 1.0f ? 1.0f : (0.1f * mscale * logf(scale) + 1.0f);
};
hparams.yarn_attn_factor = get_mscale(factor, mscale) / get_mscale(factor, mscale_all_dims);
}
switch (hparams.n_layer) {
case 26: type = LLM_TYPE_3B; break;
case 34: type = LLM_TYPE_8B; break;
@ -2301,6 +2294,32 @@ void llama_model::load_hparams(llama_model_loader & ml) {
default: throw std::runtime_error("unsupported model architecture");
}
// ref: https://github.com/huggingface/transformers/blob/6d00f6b0a5679c36510f203e4226e36f517c3032/src/transformers/modeling_rope_utils.py#L336-L348
if (hparams.rope_yarn_log_mul != 0.0f) {
const float factor = 1.0f / hparams.rope_freq_scale_train;
// note: here we assume `mscale == 1.0f`
// TODO: start reading the actual value of mscale and handle the case where it is not 1.0f
float mscale = 1.0f;
const float mscale_all_dims = hparams.rope_yarn_log_mul;
// [TAG_DEEPSEEK2_YARN_LOG_MUL_FIX]
// special-case DEEPSEEK v2:
// https://huggingface.co/deepseek-ai/DeepSeek-V2-Lite-Chat/blob/main/config.json#L42-L43
if (arch == LLM_ARCH_DEEPSEEK2 && mscale_all_dims != 1.0f) {
mscale = mscale_all_dims;
}
static auto get_mscale = [](float scale, float mscale) {
return scale <= 1.0f ? 1.0f : (0.1f * mscale * logf(scale) + 1.0f);
};
hparams.yarn_attn_factor = get_mscale(factor, mscale) / get_mscale(factor, mscale_all_dims);
LLAMA_LOG_WARN("%s: setting new yarn_attn_factor = %.4f (mscale == %.1f, mscale_all_dim = %.1f)\n",
__func__, hparams.yarn_attn_factor, mscale, mscale_all_dims);
}
pimpl->n_bytes = ml.n_bytes;
pimpl->desc_str = arch_name() + " " + type_name() + " " + ml.ftype_name();
@ -6806,6 +6825,7 @@ void llama_model::print_info() const {
LLAMA_LOG_INFO("%s: freq_base_train = %.1f\n", __func__, hparams.rope_freq_base_train);
LLAMA_LOG_INFO("%s: freq_scale_train = %g\n", __func__, hparams.rope_freq_scale_train);
LLAMA_LOG_INFO("%s: n_ctx_orig_yarn = %u\n", __func__, hparams.n_ctx_orig_yarn);
LLAMA_LOG_INFO("%s: rope_yarn_log_mul= %.4f\n", __func__, hparams.rope_yarn_log_mul);
LLAMA_LOG_INFO("%s: rope_finetuned = %s\n", __func__, hparams.rope_finetuned ? "yes" : "unknown");
// MRoPE (Multi-axis Rotary Position Embedding) sections
if (const auto & s = hparams.rope_sections; s[0] || s[1] || s[2] || s[3]) {
@ -6869,7 +6889,6 @@ void llama_model::print_info() const {
LLAMA_LOG_INFO("%s: expert_weights_scale = %.1f\n", __func__, hparams.expert_weights_scale);
LLAMA_LOG_INFO("%s: expert_weights_norm = %d\n", __func__, hparams.expert_weights_norm);
LLAMA_LOG_INFO("%s: expert_gating_func = %s\n", __func__, llama_expert_gating_func_name((llama_expert_gating_func_type) hparams.expert_gating_func));
LLAMA_LOG_INFO("%s: rope_yarn_log_mul = %.4f\n", __func__, hparams.rope_yarn_log_mul);
}
if (arch == LLM_ARCH_QWEN2MOE) {

12
src/models/deepseek2.cpp
View File

@ -1,7 +1,5 @@
#include "models.h"
llm_build_deepseek2::llm_build_deepseek2(const llama_model & model, const llm_graph_params & params) :
llm_graph_context(params) {
// lite variants include DeepSeek-V2-Lite, GigaChat3-10B-A1.8B
@ -20,9 +18,15 @@ llm_build_deepseek2::llm_build_deepseek2(const llama_model & model, const llm_gr
// 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.
const float mscale = attn_factor * (1.0f + hparams.rope_yarn_log_mul * logf(1.0f / freq_scale));
// 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
GGML_ASSERT(ext_factor >= 0.0f);
const float attn_factor_org = attn_factor * (1.0f + 0.1f * logf(1.0f / freq_scale));
// use the original attn_factor to pre-scale the kq_scale
const float mscale = attn_factor_org * (1.0f + 0.1f * hparams.rope_yarn_log_mul * logf(1.0f / freq_scale));
const float kq_scale = 1.0f * mscale * mscale / sqrtf(float(n_embd_head_k));
const float attn_factor = 1.0f / (1.0f + 0.1f * logf(1.0f / freq_scale));
ggml_tensor * cur;
ggml_tensor * inpL;