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

fix rope_theta

This commit is contained in:
ibrahimkhadraoui 2025-07-07 14:36:51 +04:00
parent 97011d7a1f
commit 286e1fa569
4 changed files with 1 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
convert_hf_to_gguf.py
View File

@ -6675,7 +6675,7 @@ class FalconH1Model(Mamba2Model):
self.gguf_writer.add_bool("falcon_h1.mamba_use_mlp", self.find_hparam(["mamba_use_mlp"], optional=True)) self.gguf_writer.add_bool("falcon_h1.mamba_use_mlp", self.find_hparam(["mamba_use_mlp"], optional=True))
self.gguf_writer.add_bool("falcon_h1.mamba_norm_before_gate", self.find_hparam(["mamba_norm_before_gate"], optional=True)) self.gguf_writer.add_bool("falcon_h1.mamba_norm_before_gate", self.find_hparam(["mamba_norm_before_gate"], optional=True))
self.gguf_writer.add_bool("falcon_h1.mamba_rms_norm", self.find_hparam(["mamba_rms_norm"], optional=True)) self.gguf_writer.add_bool("falcon_h1.mamba_rms_norm", self.find_hparam(["mamba_rms_norm"], optional=True))
self.gguf_writer.add_float64("falcon_h1.rope_theta", self.find_hparam(["rope_theta"], optional=True)) self.gguf_writer.add_rope_freq_base(self.find_hparam(["rope_theta"]))
###### CONVERSION LOGIC ###### ###### CONVERSION LOGIC ######

1
src/llama-arch.cpp
View File

@ -237,7 +237,6 @@ static const std::map<llm_kv, const char *> LLM_KV_NAMES = {
{ LLM_KV_FALCON_H1_SSM_HAS_MUP, "%s.ssm.has_mup" }, { LLM_KV_FALCON_H1_SSM_HAS_MUP, "%s.ssm.has_mup" },
{ LLM_KV_FALCON_H1_MAMBA_NORM_BEFORE_GATE, "%s.mamba_norm_before_gate" }, { LLM_KV_FALCON_H1_MAMBA_NORM_BEFORE_GATE, "%s.mamba_norm_before_gate" },
{ LLM_KV_FALCON_H1_MAMBA_RMS_NORM, "%s.mamba_rms_norm" }, { LLM_KV_FALCON_H1_MAMBA_RMS_NORM, "%s.mamba_rms_norm" },
{ LLM_KV_FALCON_H1_ROPE_THETA, "%s.rope_theta" },
{ LLM_KV_FALCON_H1_KEY_MULTIPLIER, "%s.key_multiplier" }, { LLM_KV_FALCON_H1_KEY_MULTIPLIER, "%s.key_multiplier" },
{ LLM_KV_FALCON_H1_LM_HEAD_MULTIPLIER, "%s.lm_head_multiplier" }, { LLM_KV_FALCON_H1_LM_HEAD_MULTIPLIER, "%s.lm_head_multiplier" },
{ LLM_KV_FALCON_H1_EMBEDDING_MULTIPLIER, "%s.embedding_multiplier" }, { LLM_KV_FALCON_H1_EMBEDDING_MULTIPLIER, "%s.embedding_multiplier" },

1
src/llama-arch.h
View File

@ -172,7 +172,6 @@ enum llm_kv {
LLM_KV_FALCON_H1_SSM_HAS_MUP, LLM_KV_FALCON_H1_SSM_HAS_MUP,
LLM_KV_FALCON_H1_MAMBA_NORM_BEFORE_GATE, LLM_KV_FALCON_H1_MAMBA_NORM_BEFORE_GATE,
LLM_KV_FALCON_H1_MAMBA_RMS_NORM, LLM_KV_FALCON_H1_MAMBA_RMS_NORM,
LLM_KV_FALCON_H1_ROPE_THETA,
LLM_KV_FALCON_H1_KEY_MULTIPLIER, LLM_KV_FALCON_H1_KEY_MULTIPLIER,
LLM_KV_FALCON_H1_LM_HEAD_MULTIPLIER, LLM_KV_FALCON_H1_LM_HEAD_MULTIPLIER,
LLM_KV_FALCON_H1_EMBEDDING_MULTIPLIER, LLM_KV_FALCON_H1_EMBEDDING_MULTIPLIER,

1
src/llama-model.cpp
View File

@ -1577,7 +1577,6 @@ void llama_model::load_hparams(llama_model_loader & ml) {
ml.get_key(LLM_KV_FALCON_H1_SSM_HAS_MUP, hparams.ssm_has_mup); ml.get_key(LLM_KV_FALCON_H1_SSM_HAS_MUP, hparams.ssm_has_mup);
ml.get_key(LLM_KV_FALCON_H1_MAMBA_NORM_BEFORE_GATE, hparams.mamba_norm_before_gate); ml.get_key(LLM_KV_FALCON_H1_MAMBA_NORM_BEFORE_GATE, hparams.mamba_norm_before_gate);
ml.get_key(LLM_KV_FALCON_H1_MAMBA_RMS_NORM, hparams.mamba_rms_norm); ml.get_key(LLM_KV_FALCON_H1_MAMBA_RMS_NORM, hparams.mamba_rms_norm);
ml.get_key(LLM_KV_FALCON_H1_ROPE_THETA, hparams.rope_theta);
ml.get_key(LLM_KV_FALCON_H1_KEY_MULTIPLIER, hparams.key_multiplier); ml.get_key(LLM_KV_FALCON_H1_KEY_MULTIPLIER, hparams.key_multiplier);
ml.get_key(LLM_KV_FALCON_H1_LM_HEAD_MULTIPLIER, hparams.lm_head_multiplier); ml.get_key(LLM_KV_FALCON_H1_LM_HEAD_MULTIPLIER, hparams.lm_head_multiplier);
ml.get_key(LLM_KV_FALCON_H1_EMBEDDING_MULTIPLIER, hparams.embedding_multiplier); ml.get_key(LLM_KV_FALCON_H1_EMBEDDING_MULTIPLIER, hparams.embedding_multiplier);