From 632861e6c1c21d28b362e94f90d2ab30b920989d Mon Sep 17 00:00:00 2001 From: younesbelkada Date: Mon, 7 Jul 2025 17:27:34 +0400 Subject: [PATCH] some cleanups --- convert_hf_to_gguf.py | 2 -- src/llama-arch.cpp | 2 -- src/llama-arch.h | 2 -- src/llama-hparams.h | 2 -- src/llama-model.cpp | 2 -- 5 files changed, 10 deletions(-) diff --git a/convert_hf_to_gguf.py b/convert_hf_to_gguf.py index f9e279b8f3..09f6cd173c 100755 --- a/convert_hf_to_gguf.py +++ b/convert_hf_to_gguf.py @@ -6682,8 +6682,6 @@ class FalconH1Model(Mamba2Model): # Add any other Falcon Mamba2 specific configuration - 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_rms_norm", self.find_hparam(["mamba_rms_norm"], optional=True)) self.gguf_writer.add_rope_freq_base(self.find_hparam(["rope_theta"])) diff --git a/src/llama-arch.cpp b/src/llama-arch.cpp index d78d407a03..fce25a5064 100644 --- a/src/llama-arch.cpp +++ b/src/llama-arch.cpp @@ -227,8 +227,6 @@ static const std::map LLM_KV_NAMES = { { LLM_KV_SSM_HEAD_DIM, "%s.ssm.head_dim" }, { LLM_KV_MAMBA_D_SSM, "%s.ssm.mamba_d_ssm" }, - { LLM_KV_FALCON_H1_USE_MLP, "%s.mamba_use_mlp" }, - { 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_ADAPTER_TYPE, "adapter.type" }, diff --git a/src/llama-arch.h b/src/llama-arch.h index 1dcd135ff7..ac7bc39166 100644 --- a/src/llama-arch.h +++ b/src/llama-arch.h @@ -162,8 +162,6 @@ enum llm_kv { LLM_KV_SSM_HEAD_DIM, LLM_KV_MAMBA_D_SSM, LLM_KV_N_LAYER, - LLM_KV_FALCON_H1_USE_MLP, - LLM_KV_FALCON_H1_MAMBA_NORM_BEFORE_GATE, LLM_KV_FALCON_H1_MAMBA_RMS_NORM, LLM_KV_ROPE_DIMENSION_COUNT, diff --git a/src/llama-hparams.h b/src/llama-hparams.h index 429eaf0482..556c8a056f 100644 --- a/src/llama-hparams.h +++ b/src/llama-hparams.h @@ -119,8 +119,6 @@ struct llama_hparams { uint32_t ssm_mamba_d_ssm = 0; uint32_t attn_head_dim = 0; - bool mamba_use_mlp = false; - bool mamba_norm_before_gate = false; bool mamba_rms_norm = false; double rope_theta = 10000.0; uint32_t vocab_size = 0; diff --git a/src/llama-model.cpp b/src/llama-model.cpp index ee3b2a2c5f..0e9da3c410 100644 --- a/src/llama-model.cpp +++ b/src/llama-model.cpp @@ -1566,8 +1566,6 @@ void llama_model::load_hparams(llama_model_loader & ml) { // Falcon-H1 parameters ml.get_key(LLM_KV_ATTN_HEAD_DIM, hparams.attn_head_dim); - ml.get_key(LLM_KV_FALCON_H1_USE_MLP, hparams.mamba_use_mlp); - 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); std::fill(hparams.recurrent_layer_arr.begin(), hparams.recurrent_layer_arr.end(), true);