From e1fcf8b09b8115156313c61bdd42186a7c7fb7be Mon Sep 17 00:00:00 2001 From: Bartowski <3266127+bartowski1182@users.noreply.github.com> Date: Fri, 14 Nov 2025 07:54:10 -0500 Subject: [PATCH] model : add AfmoeForCausalLM support (#16477) MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit * Add AFMOE model support * Update to vocab * Add model sizing * Undo Rope change for ARCEE model * Address review comments * Update modeling code is_sliding -> use_rope, replace hard-coded logic * Fix AFMOE tokenizer * Update convert_hf_to_gguf.py Co-authored-by: Sigbjørn Skjæret * Update convert_hf_to_gguf.py Co-authored-by: Sigbjørn Skjæret * Update AFMoE tokenizer class identification to be more unique --------- Co-authored-by: Sigbjørn Skjæret --- convert_hf_to_gguf.py | 78 ++++++++++++++ convert_hf_to_gguf_update.py | 1 + gguf-py/gguf/constants.py | 31 ++++++ gguf-py/gguf/tensor_mapping.py | 9 +- src/CMakeLists.txt | 1 + src/llama-arch.cpp | 32 ++++++ src/llama-arch.h | 2 + src/llama-model.cpp | 102 ++++++++++++++++++ src/llama-model.h | 2 + src/llama-vocab.cpp | 15 +++ src/llama-vocab.h | 1 + src/models/afmoe.cpp | 187 +++++++++++++++++++++++++++++++++ src/models/models.h | 4 + src/unicode.cpp | 77 ++++++++++++++ 14 files changed, 541 insertions(+), 1 deletion(-) create mode 100644 src/models/afmoe.cpp diff --git a/convert_hf_to_gguf.py b/convert_hf_to_gguf.py index cc77a3db27..2b08013e1e 100755 --- a/convert_hf_to_gguf.py +++ b/convert_hf_to_gguf.py @@ -1124,6 +1124,9 @@ class TextModel(ModelBase): if chkhsh == "a1e163ecab2e718a4c829d1148b6e86824ec36163bb71941c3dca9cd5ac25756": # ref: https://huggingface.co/JetBrains/Mellum-4b-base res = "mellum" + if chkhsh == "49fc0303c9e0d2c2c565c510f64b2d9b271276acdcdadff733249eda9f7d59df": + # ref: https://huggingface.co/arcee-ai/Trinity-Tokenizer + res = "afmoe" if chkhsh == "9b1be57e70d20d9501b2b3186e792d81181ae36ada3903c26f9fea418cf87206": # ref: https://huggingface.co/inclusionAI/Ling-mini-base-2.0 res = "bailingmoe2" @@ -2533,6 +2536,81 @@ class ArceeModel(LlamaModel): self.gguf_writer.add_rope_scaling_orig_ctx_len(rope_scaling["original_max_position_embeddings"]) +@ModelBase.register("AfmoeForCausalLM") +class AfmoeModel(LlamaModel): + model_arch = gguf.MODEL_ARCH.AFMOE + + def set_gguf_parameters(self): + super().set_gguf_parameters() + + # MoE parameters + if (n_experts := self.hparams.get("num_experts")) is not None: + self.gguf_writer.add_expert_count(n_experts) + if (n_shared_experts := self.hparams.get("num_shared_experts")) is not None: + self.gguf_writer.add_expert_shared_count(n_shared_experts) + if (moe_intermediate_size := self.hparams.get("moe_intermediate_size")) is not None: + self.gguf_writer.add_expert_feed_forward_length(moe_intermediate_size) + if (n_dense_layers := self.hparams.get("num_dense_layers")) is not None: + self.gguf_writer.add_leading_dense_block_count(n_dense_layers) + + # Expert Gating Function + score_func = self.hparams.get("score_func") + if score_func == "sigmoid": + self.gguf_writer.add_expert_gating_func(gguf.ExpertGatingFuncType.SIGMOID) + elif score_func == "softmax": + self.gguf_writer.add_expert_gating_func(gguf.ExpertGatingFuncType.SOFTMAX) + elif score_func is not None: + raise ValueError(f"Unsupported score_function value: {score_func}") + + # Route normalization and scaling + if (route_norm := self.hparams.get("route_norm")) is not None: + self.gguf_writer.add_expert_weights_norm(route_norm) + if (route_scale := self.hparams.get("route_scale")) is not None: + self.gguf_writer.add_expert_weights_scale(route_scale) + + # Sliding window attention + if (sliding_window := self.hparams.get("sliding_window")) is not None: + self.gguf_writer.add_sliding_window(sliding_window) + + def modify_tensors(self, data_torch: Tensor, name: str, bid: int | None) -> Iterable[tuple[str, Tensor]]: + # Handle expert weights - they're already merged in the HF format + # process the experts separately + if name.find("mlp.experts") != -1: + n_experts = self.hparams["num_experts"] + assert bid is not None + + if self._experts is None: + self._experts = [{} for _ in range(self.block_count)] + + self._experts[bid][name] = data_torch + + if len(self._experts[bid]) >= n_experts * 3: + tensors: list[tuple[str, Tensor]] = [] + + # merge the experts into a single 3d tensor + for w_name in ["gate_proj", "up_proj", "down_proj"]: + datas: list[Tensor] = [] + + for xid in range(n_experts): + ename_to_retrieve = f"model.layers.{bid}.mlp.experts.{xid}.{w_name}.weight" + datas.append(self._experts[bid][ename_to_retrieve]) + del self._experts[bid][ename_to_retrieve] + + data_torch = torch.stack(datas, dim=0) + merged_name = f"model.layers.{bid}.mlp.experts.{w_name}.weight" + new_name = self.map_tensor_name(merged_name) + tensors.append((new_name, data_torch)) + + return tensors + else: + return [] + + if name.endswith(".expert_bias"): + name = name.replace(".expert_bias", ".expert_bias.bias") + + return [(self.map_tensor_name(name), data_torch)] + + @ModelBase.register( "LlavaForConditionalGeneration", # pixtral "Mistral3ForConditionalGeneration", # mistral small 3.1 diff --git a/convert_hf_to_gguf_update.py b/convert_hf_to_gguf_update.py index 7df96eb083..b8f694e86c 100755 --- a/convert_hf_to_gguf_update.py +++ b/convert_hf_to_gguf_update.py @@ -139,6 +139,7 @@ models = [ {"name": "lfm2", "tokt": TOKENIZER_TYPE.BPE, "repo": "https://huggingface.co/LiquidAI/LFM2-Tokenizer"}, {"name": "exaone4", "tokt": TOKENIZER_TYPE.BPE, "repo": "https://huggingface.co/LGAI-EXAONE/EXAONE-4.0-32B", }, {"name": "mellum", "tokt": TOKENIZER_TYPE.BPE, "repo": "https://huggingface.co/JetBrains/Mellum-4b-base", }, + {"name": "afmoe", "tokt": TOKENIZER_TYPE.BPE, "repo": "https://huggingface.co/arcee-ai/Trinity-Tokenizer", }, {"name": "bailingmoe2", "tokt": TOKENIZER_TYPE.BPE, "repo": "https://huggingface.co/inclusionAI/Ling-mini-base-2.0", }, {"name": "granite-docling", "tokt": TOKENIZER_TYPE.BPE, "repo": "https://huggingface.co/ibm-granite/granite-docling-258M", }, {"name": "minimax-m2", "tokt": TOKENIZER_TYPE.BPE, "repo": "https://huggingface.co/MiniMaxAI/MiniMax-M2", }, diff --git a/gguf-py/gguf/constants.py b/gguf-py/gguf/constants.py index 6b4b6c5ab0..1cd0efad4a 100644 --- a/gguf-py/gguf/constants.py +++ b/gguf-py/gguf/constants.py @@ -409,6 +409,7 @@ class MODEL_ARCH(IntEnum): BAILINGMOE2 = auto() DOTS1 = auto() ARCEE = auto() + AFMOE = auto() ERNIE4_5 = auto() ERNIE4_5_MOE = auto() HUNYUAN_MOE = auto() @@ -464,6 +465,7 @@ class MODEL_TENSOR(IntEnum): ATTN_POST_NORM = auto() ATTN_ROT_EMBD = auto() ATTN_SINKS = auto() + ATTN_GATE = auto() FFN_GATE_INP = auto() FFN_GATE_INP_SHEXP = auto() FFN_NORM = auto() @@ -776,6 +778,7 @@ MODEL_ARCH_NAMES: dict[MODEL_ARCH, str] = { MODEL_ARCH.BAILINGMOE2: "bailingmoe2", MODEL_ARCH.DOTS1: "dots1", MODEL_ARCH.ARCEE: "arcee", + MODEL_ARCH.AFMOE: "afmoe", MODEL_ARCH.ERNIE4_5: "ernie4_5", MODEL_ARCH.ERNIE4_5_MOE: "ernie4_5-moe", MODEL_ARCH.FALCON_H1: "falcon-h1", @@ -828,6 +831,7 @@ TENSOR_NAMES: dict[MODEL_TENSOR, str] = { MODEL_TENSOR.ATTN_OUT: "blk.{bid}.attn_output", MODEL_TENSOR.ATTN_ROT_EMBD: "blk.{bid}.attn_rot_embd", MODEL_TENSOR.ATTN_SINKS: "blk.{bid}.attn_sinks", + MODEL_TENSOR.ATTN_GATE: "blk.{bid}.attn_gate", MODEL_TENSOR.ATTN_Q_NORM: "blk.{bid}.attn_q_norm", MODEL_TENSOR.ATTN_K_NORM: "blk.{bid}.attn_k_norm", MODEL_TENSOR.ATTN_OUT_NORM: "blk.{bid}.attn_output_norm", @@ -2693,6 +2697,33 @@ MODEL_TENSORS: dict[MODEL_ARCH, list[MODEL_TENSOR]] = { MODEL_TENSOR.FFN_DOWN, MODEL_TENSOR.FFN_UP, ], + MODEL_ARCH.AFMOE: [ + MODEL_TENSOR.TOKEN_EMBD, + MODEL_TENSOR.OUTPUT_NORM, + MODEL_TENSOR.OUTPUT, + MODEL_TENSOR.ATTN_NORM, + MODEL_TENSOR.ATTN_POST_NORM, + MODEL_TENSOR.ATTN_Q, + MODEL_TENSOR.ATTN_K, + MODEL_TENSOR.ATTN_V, + MODEL_TENSOR.ATTN_OUT, + MODEL_TENSOR.ATTN_Q_NORM, + MODEL_TENSOR.ATTN_K_NORM, + MODEL_TENSOR.ATTN_GATE, + MODEL_TENSOR.FFN_GATE, + MODEL_TENSOR.FFN_DOWN, + MODEL_TENSOR.FFN_UP, + MODEL_TENSOR.FFN_GATE_INP, + MODEL_TENSOR.FFN_GATE_EXP, + MODEL_TENSOR.FFN_DOWN_EXP, + MODEL_TENSOR.FFN_UP_EXP, + MODEL_TENSOR.FFN_GATE_SHEXP, + MODEL_TENSOR.FFN_UP_SHEXP, + MODEL_TENSOR.FFN_DOWN_SHEXP, + MODEL_TENSOR.FFN_PRE_NORM, + MODEL_TENSOR.FFN_POST_NORM, + MODEL_TENSOR.FFN_EXP_PROBS_B, + ], MODEL_ARCH.ERNIE4_5: [ MODEL_TENSOR.TOKEN_EMBD, MODEL_TENSOR.OUTPUT_NORM, diff --git a/gguf-py/gguf/tensor_mapping.py b/gguf-py/gguf/tensor_mapping.py index 9294066876..8c7ed10f2e 100644 --- a/gguf-py/gguf/tensor_mapping.py +++ b/gguf-py/gguf/tensor_mapping.py @@ -314,6 +314,10 @@ class TensorNameMap: "model.layers.{bid}.self_attn.sinks", # openai-moe ), + MODEL_TENSOR.ATTN_GATE: ( + "model.layers.{bid}.self_attn.gate_proj", # afmoe + ), + # Feed-forward norm MODEL_TENSOR.FFN_NORM: ( "gpt_neox.layers.{bid}.post_attention_layernorm", # gptneox @@ -340,11 +344,12 @@ class TensorNameMap: "model.layers.{bid}.feedforward_layernorm", # apertus ), - # Post feed-forward norm + # Pre feed-forward norm MODEL_TENSOR.FFN_PRE_NORM: ( "model.layers.{bid}.pre_feedforward_layernorm", # gemma2 "layers.{bid}.pre_feedforward_layernorm", # embeddinggemma "model.layers.{bid}.pre_ff_layernorm.weight", + "model.layers.{bid}.pre_mlp_layernorm", # afmoe ), # Post feed-forward norm @@ -370,6 +375,7 @@ class TensorNameMap: "model.layers.{bid}.mlp.gate.wg", # hunyuan "model.layers.{bid}.block_sparse_moe.primary_router", # smallthinker "model.layers.{bid}.feed_forward.gate", # lfm2moe + "model.layers.{bid}.mlp.router.gate", # afmoe ), MODEL_TENSOR.FFN_GATE_INP_SHEXP: ( @@ -380,6 +386,7 @@ class TensorNameMap: "model.layers.{bid}.mlp.gate.e_score_correction", # deepseek-v3 dots1 "model.layers.{bid}.mlp.moe_statics.e_score_correction", # ernie4.5-moe "model.layers.{bid}.mlp.gate.expert_bias", # bailingmoe2 + "model.layers.{bid}.mlp.expert_bias", # afmoe "model.layers.{bid}.feed_forward.expert_bias", # lfm2moe "model.layers.{bid}.block_sparse_moe.e_score_correction", # minimax-m2 ), diff --git a/src/CMakeLists.txt b/src/CMakeLists.txt index 6fc5b00101..8ec95ee176 100644 --- a/src/CMakeLists.txt +++ b/src/CMakeLists.txt @@ -35,6 +35,7 @@ add_library(llama unicode-data.cpp unicode.cpp unicode.h + models/afmoe.cpp models/apertus.cpp models/arcee.cpp models/arctic.cpp diff --git a/src/llama-arch.cpp b/src/llama-arch.cpp index b7642b568d..b2eb2477f9 100644 --- a/src/llama-arch.cpp +++ b/src/llama-arch.cpp @@ -90,6 +90,7 @@ static const std::map LLM_ARCH_NAMES = { { LLM_ARCH_BAILINGMOE2, "bailingmoe2" }, { LLM_ARCH_DOTS1, "dots1" }, { LLM_ARCH_ARCEE, "arcee" }, + { LLM_ARCH_AFMOE, "afmoe" }, { LLM_ARCH_ERNIE4_5, "ernie4_5" }, { LLM_ARCH_ERNIE4_5_MOE, "ernie4_5-moe" }, { LLM_ARCH_HUNYUAN_MOE, "hunyuan-moe" }, @@ -333,6 +334,36 @@ static const std::map> LLM_TENSOR_N { LLM_TENSOR_FFN_UP, "blk.%d.ffn_up" }, }, }, + { + LLM_ARCH_AFMOE, + { + { LLM_TENSOR_TOKEN_EMBD, "token_embd" }, + { LLM_TENSOR_OUTPUT_NORM, "output_norm" }, + { LLM_TENSOR_OUTPUT, "output" }, + { LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" }, + { LLM_TENSOR_ATTN_POST_NORM, "blk.%d.post_attention_norm" }, + { LLM_TENSOR_ATTN_Q, "blk.%d.attn_q" }, + { LLM_TENSOR_ATTN_K, "blk.%d.attn_k" }, + { LLM_TENSOR_ATTN_V, "blk.%d.attn_v" }, + { LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" }, + { LLM_TENSOR_ATTN_Q_NORM, "blk.%d.attn_q_norm" }, + { LLM_TENSOR_ATTN_K_NORM, "blk.%d.attn_k_norm" }, + { LLM_TENSOR_ATTN_GATE, "blk.%d.attn_gate" }, + { LLM_TENSOR_FFN_NORM, "blk.%d.ffn_norm" }, + { LLM_TENSOR_FFN_POST_NORM, "blk.%d.post_ffw_norm" }, + { LLM_TENSOR_FFN_GATE_INP, "blk.%d.ffn_gate_inp" }, + { LLM_TENSOR_FFN_GATE, "blk.%d.ffn_gate" }, + { LLM_TENSOR_FFN_DOWN, "blk.%d.ffn_down" }, + { LLM_TENSOR_FFN_UP, "blk.%d.ffn_up" }, + { LLM_TENSOR_FFN_GATE_EXPS, "blk.%d.ffn_gate_exps" }, + { LLM_TENSOR_FFN_DOWN_EXPS, "blk.%d.ffn_down_exps" }, + { LLM_TENSOR_FFN_UP_EXPS, "blk.%d.ffn_up_exps" }, + { LLM_TENSOR_FFN_GATE_SHEXP, "blk.%d.ffn_gate_shexp" }, + { LLM_TENSOR_FFN_UP_SHEXP, "blk.%d.ffn_up_shexp" }, + { LLM_TENSOR_FFN_DOWN_SHEXP, "blk.%d.ffn_down_shexp" }, + { LLM_TENSOR_FFN_EXP_PROBS_B, "blk.%d.exp_probs_b" }, + }, + }, { LLM_ARCH_LLAMA4, { @@ -2444,6 +2475,7 @@ static const std::map LLM_TENSOR_INFOS = { {LLM_TENSOR_ATTN_V, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}}, {LLM_TENSOR_ATTN_QKV, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}}, {LLM_TENSOR_ATTN_OUT, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}}, + {LLM_TENSOR_ATTN_GATE, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}}, {LLM_TENSOR_FFN_GATE, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}}, {LLM_TENSOR_FFN_DOWN, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}}, {LLM_TENSOR_FFN_UP, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}}, diff --git a/src/llama-arch.h b/src/llama-arch.h index a769dd1e85..ae7fa222ac 100644 --- a/src/llama-arch.h +++ b/src/llama-arch.h @@ -94,6 +94,7 @@ enum llm_arch { LLM_ARCH_BAILINGMOE2, LLM_ARCH_DOTS1, LLM_ARCH_ARCEE, + LLM_ARCH_AFMOE, LLM_ARCH_ERNIE4_5, LLM_ARCH_ERNIE4_5_MOE, LLM_ARCH_HUNYUAN_MOE, @@ -312,6 +313,7 @@ enum llm_tensor { LLM_TENSOR_ATTN_POST_NORM, LLM_TENSOR_ATTN_ROT_EMBD, LLM_TENSOR_ATTN_SINKS, + LLM_TENSOR_ATTN_GATE, LLM_TENSOR_FFN_GATE_INP, LLM_TENSOR_FFN_GATE_INP_SHEXP, LLM_TENSOR_FFN_NORM, diff --git a/src/llama-model.cpp b/src/llama-model.cpp index 829f1e3c14..e703181a19 100644 --- a/src/llama-model.cpp +++ b/src/llama-model.cpp @@ -84,6 +84,7 @@ const char * llm_type_name(llm_type type) { case LLM_TYPE_15B: return "15B"; case LLM_TYPE_16B: return "16B"; case LLM_TYPE_20B: return "20B"; + case LLM_TYPE_26B: return "26B"; case LLM_TYPE_27B: return "27B"; case LLM_TYPE_30B: return "30B"; case LLM_TYPE_32B: return "32B"; @@ -695,6 +696,37 @@ void llama_model::load_hparams(llama_model_loader & ml) { default: type = LLM_TYPE_UNKNOWN; } } break; + case LLM_ARCH_AFMOE: + { + ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps); + ml.get_key(LLM_KV_LEADING_DENSE_BLOCK_COUNT, hparams.n_layer_dense_lead); + ml.get_key(LLM_KV_EXPERT_FEED_FORWARD_LENGTH, hparams.n_ff_exp); + ml.get_key(LLM_KV_EXPERT_SHARED_COUNT, hparams.n_expert_shared); + ml.get_key(LLM_KV_EXPERT_GATING_FUNC, hparams.expert_gating_func, false); + ml.get_key(LLM_KV_EXPERT_WEIGHTS_SCALE, hparams.expert_weights_scale, false); + ml.get_key(LLM_KV_EXPERT_WEIGHTS_NORM, hparams.expert_weights_norm, false); + ml.get_key(LLM_KV_ATTENTION_SLIDING_WINDOW, hparams.n_swa, false); + + // Set up interleaved sliding window attention (ISWA) + // Pattern: 3 sliding - 1 full (global_attn_every_n_layers = 4) + if (hparams.n_swa > 0) { + hparams.swa_type = LLAMA_SWA_TYPE_STANDARD; + hparams.set_swa_pattern(4); + } else { + hparams.swa_type = LLAMA_SWA_TYPE_NONE; + } + + // Default to sigmoid if not set + if (hparams.expert_gating_func == LLAMA_EXPERT_GATING_FUNC_TYPE_NONE) { + hparams.expert_gating_func = LLAMA_EXPERT_GATING_FUNC_TYPE_SIGMOID; + } + + switch (hparams.n_layer) { + case 56: type = LLM_TYPE_6B; break; + case 32: type = LLM_TYPE_26B; break; + default: type = LLM_TYPE_UNKNOWN; + } + } break; case LLM_ARCH_DECI: { ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps); @@ -5749,6 +5781,71 @@ bool llama_model::load_tensors(llama_model_loader & ml) { layer.ffn_up = create_tensor(tn(LLM_TENSOR_FFN_UP, "weight", i), {n_embd, n_ff}, 0); } } break; + case LLM_ARCH_AFMOE: + { + tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0); + + // output + output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0); + output = create_tensor(tn(LLM_TENSOR_OUTPUT, "weight"), {n_embd, n_vocab}, TENSOR_NOT_REQUIRED); + + // if output is NULL, init from the input tok embed + if (output == NULL) { + output = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, TENSOR_DUPLICATED); + } + + const int64_t n_ff_exp = hparams.n_ff_exp; + const int64_t n_expert_shared = hparams.n_expert_shared; + + for (int i = 0; i < n_layer; ++i) { + auto & layer = layers[i]; + + // dual attention normalization + layer.attn_norm = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, 0); + layer.attn_post_norm = create_tensor(tn(LLM_TENSOR_ATTN_POST_NORM, "weight", i), {n_embd}, 0); + + // attention projections + layer.wq = create_tensor(tn(LLM_TENSOR_ATTN_Q, "weight", i), {n_embd, n_embd_head_k * n_head}, 0); + layer.wk = create_tensor(tn(LLM_TENSOR_ATTN_K, "weight", i), {n_embd, n_embd_k_gqa}, 0); + layer.wv = create_tensor(tn(LLM_TENSOR_ATTN_V, "weight", i), {n_embd, n_embd_v_gqa}, 0); + layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd_head_k * n_head, n_embd}, 0); + + // Q/K normalization + layer.attn_q_norm = create_tensor(tn(LLM_TENSOR_ATTN_Q_NORM, "weight", i), {n_embd_head_k}, 0); + layer.attn_k_norm = create_tensor(tn(LLM_TENSOR_ATTN_K_NORM, "weight", i), {n_embd_head_k}, 0); + + // attention gating + layer.wqkv_gate = create_tensor(tn(LLM_TENSOR_ATTN_GATE, "weight", i), {n_embd, n_embd_head_k * n_head}, 0); + + // dual ffn normalization + layer.ffn_norm = create_tensor(tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, 0); + layer.ffn_post_norm = create_tensor(tn(LLM_TENSOR_FFN_POST_NORM, "weight", i), {n_embd}, 0); + + if (static_cast(i) >= hparams.n_layer_dense_lead) { + // MoE layers + layer.ffn_gate_inp = create_tensor(tn(LLM_TENSOR_FFN_GATE_INP, "weight", i), {n_embd, n_expert}, 0); + layer.ffn_exp_probs_b = create_tensor(tn(LLM_TENSOR_FFN_EXP_PROBS_B, "bias", i), {n_expert}, 0); + + // grouped expert weights + layer.ffn_gate_exps = create_tensor(tn(LLM_TENSOR_FFN_GATE_EXPS, "weight", i), {n_embd, n_ff_exp, n_expert}, 0); + layer.ffn_down_exps = create_tensor(tn(LLM_TENSOR_FFN_DOWN_EXPS, "weight", i), {n_ff_exp, n_embd, n_expert}, 0); + layer.ffn_up_exps = create_tensor(tn(LLM_TENSOR_FFN_UP_EXPS, "weight", i), {n_embd, n_ff_exp, n_expert}, 0); + + // shared expert + if (n_expert_shared > 0) { + const int64_t n_ff_shexp = n_ff_exp * n_expert_shared; + layer.ffn_gate_shexp = create_tensor(tn(LLM_TENSOR_FFN_GATE_SHEXP, "weight", i), {n_embd, n_ff_shexp}, 0); + layer.ffn_down_shexp = create_tensor(tn(LLM_TENSOR_FFN_DOWN_SHEXP, "weight", i), {n_ff_shexp, n_embd}, 0); + layer.ffn_up_shexp = create_tensor(tn(LLM_TENSOR_FFN_UP_SHEXP, "weight", i), {n_embd, n_ff_shexp}, 0); + } + } else { + // Dense layers + 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); + } + } + } break; case LLM_ARCH_ERNIE4_5: case LLM_ARCH_ERNIE4_5_MOE: { @@ -7243,6 +7340,10 @@ ggml_cgraph * llama_model::build_graph(const llm_graph_params & params) const { { llm = std::make_unique(*this, params); } break; + case LLM_ARCH_AFMOE: + { + llm = std::make_unique(*this, params); + } break; case LLM_ARCH_ERNIE4_5: { llm = std::make_unique(*this, params); @@ -7528,6 +7629,7 @@ llama_rope_type llama_model_rope_type(const llama_model * model) { case LLM_ARCH_MINIMAX_M2: case LLM_ARCH_COGVLM: case LLM_ARCH_PANGU_EMBED: + case LLM_ARCH_AFMOE: return LLAMA_ROPE_TYPE_NEOX; case LLM_ARCH_QWEN2VL: diff --git a/src/llama-model.h b/src/llama-model.h index 71ff148e07..f730c49540 100644 --- a/src/llama-model.h +++ b/src/llama-model.h @@ -76,6 +76,7 @@ enum llm_type { LLM_TYPE_15B, LLM_TYPE_16B, LLM_TYPE_20B, + LLM_TYPE_26B, LLM_TYPE_27B, LLM_TYPE_30B, LLM_TYPE_32B, @@ -234,6 +235,7 @@ struct llama_layer { struct ggml_tensor * wk_enc = nullptr; struct ggml_tensor * wv_enc = nullptr; struct ggml_tensor * wo_enc = nullptr; + struct ggml_tensor * wqkv_gate = nullptr; // attention bias struct ggml_tensor * bq = nullptr; diff --git a/src/llama-vocab.cpp b/src/llama-vocab.cpp index 97f374eac9..29e31cecd1 100644 --- a/src/llama-vocab.cpp +++ b/src/llama-vocab.cpp @@ -443,6 +443,17 @@ struct llm_tokenizer_bpe : llm_tokenizer { "(?:'[sS]|'[tT]|'[rR][eE]|'[vV][eE]|'[mM]|'[lL][lL]|'[dD])|[^\\r\\n\\p{L}\\p{N}]?\\p{L}+|\\p{N}| ?[^\\s\\p{L}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+", }; break; + case LLAMA_VOCAB_PRE_TYPE_AFMOE: + regex_exprs = { + // Digit handling - uses custom implementation in unicode.cpp + // Groups digits with leading 1-2 based on total length modulo 3 + "\\p{AFMoE_digits}", + // CJK and Asian scripts (using direct Unicode literals) + "[一-鿿㐀-䶿豈-﫿぀-ゟ゠-ヿ･-ﾟ⼀-⿟เ-๿຀-໿ក-៿က-႟ꩠ-ꩿꧠ-꧿가-힯ᄀ-ᇿ]+", + // Main BPE pattern + "[!\"#$%&'()*+,\\-./:;<=>?@\\[\\\\\\]^_`{|}~][A-Za-z]+|[^\\r\\n\\p{L}\\p{P}\\p{S}]?[\\p{L}\\p{M}]+| ?[\\p{P}\\p{S}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+", + }; + break; default: // default regex for BPE tokenization pre-processing regex_exprs = { @@ -1993,6 +2004,10 @@ void llama_vocab::impl::load(llama_model_loader & ml, const LLM_KV & kv) { tokenizer_pre == "grok-2") { pre_type = LLAMA_VOCAB_PRE_TYPE_GROK_2; clean_spaces = false; + } else if ( + tokenizer_pre == "afmoe") { + pre_type = LLAMA_VOCAB_PRE_TYPE_AFMOE; + clean_spaces = false; } else if ( tokenizer_pre == "minimax-m2") { pre_type = LLAMA_VOCAB_PRE_TYPE_MINIMAX_M2; diff --git a/src/llama-vocab.h b/src/llama-vocab.h index 1194ec473d..55f8f3923c 100644 --- a/src/llama-vocab.h +++ b/src/llama-vocab.h @@ -50,6 +50,7 @@ enum llama_vocab_pre_type { LLAMA_VOCAB_PRE_TYPE_GROK_2 = 39, LLAMA_VOCAB_PRE_TYPE_GRANITE_DOCLING = 40, LLAMA_VOCAB_PRE_TYPE_MINIMAX_M2 = 41, + LLAMA_VOCAB_PRE_TYPE_AFMOE = 42, }; struct LLM_KV; diff --git a/src/models/afmoe.cpp b/src/models/afmoe.cpp new file mode 100644 index 0000000000..0192e344ca --- /dev/null +++ b/src/models/afmoe.cpp @@ -0,0 +1,187 @@ +#include "models.h" + +llm_build_afmoe::llm_build_afmoe(const llama_model & model, const llm_graph_params & params) : llm_graph_context(params) { + const int64_t n_embd_head = hparams.n_embd_head_v; + GGML_ASSERT(n_embd_head == hparams.n_embd_head_k); + + ggml_tensor * cur; + ggml_tensor * inpL; + + inpL = build_inp_embd(model.tok_embd); + + // MuP scaling: embeddings * sqrt(hidden_size) + // mup_enabled = true, hidden_size = 1024, scale = 32.0 + inpL = ggml_scale(ctx0, inpL, sqrtf(float(n_embd))); + cb(inpL, "inp_embd_scaled", -1); + + // inp_pos - contains the positions + ggml_tensor * inp_pos = build_inp_pos(); + auto * inp_attn = build_attn_inp_kv_iswa(); + ggml_tensor * inp_out_ids = build_inp_out_ids(); + + const float kq_scale = 1.0f/sqrtf(float(n_embd_head)); + + for (int il = 0; il < n_layer; ++il) { + ggml_tensor * inpSA = inpL; + + // dual attention normalization (pre) + cur = build_norm(inpL, + model.layers[il].attn_norm, NULL, + LLM_NORM_RMS, il); + cb(cur, "attn_norm", il); + + // self-attention + { + ggml_tensor * attn_inp = cur; // save input for gate computation + + ggml_tensor * Qcur = build_lora_mm(model.layers[il].wq, cur); + cb(Qcur, "Qcur", il); + + ggml_tensor * Kcur = build_lora_mm(model.layers[il].wk, cur); + cb(Kcur, "Kcur", il); + + ggml_tensor * Vcur = build_lora_mm(model.layers[il].wv, cur); + cb(Vcur, "Vcur", il); + + // compute gate from input + ggml_tensor * gate = build_lora_mm(model.layers[il].wqkv_gate, attn_inp); + cb(gate, "attn_gate_proj", il); + + Qcur = ggml_reshape_3d(ctx0, Qcur, n_embd_head, n_head, n_tokens); + Kcur = ggml_reshape_3d(ctx0, Kcur, n_embd_head, n_head_kv, n_tokens); + + // Q/K normalization + Qcur = build_norm(Qcur, model.layers[il].attn_q_norm, NULL, LLM_NORM_RMS, il); + Kcur = build_norm(Kcur, model.layers[il].attn_k_norm, NULL, LLM_NORM_RMS, il); + cb(Qcur, "Qcur_normed", il); + cb(Kcur, "Kcur_normed", il); + + // RoPE only for sliding_attention layers + const bool use_rope = hparams.n_no_rope_layer_step > 0 && + ((il + 1) % hparams.n_no_rope_layer_step) != 0; + if (use_rope) { + Qcur = ggml_rope_ext( + ctx0, Qcur, inp_pos, nullptr, + n_rot, rope_type, n_ctx_orig, freq_base, freq_scale, + ext_factor, attn_factor, beta_fast, beta_slow); + cb(Qcur, "Qcur_rope", il); + + Kcur = ggml_rope_ext( + ctx0, Kcur, inp_pos, nullptr, + n_rot, rope_type, n_ctx_orig, freq_base, freq_scale, + ext_factor, attn_factor, beta_fast, beta_slow); + cb(Kcur, "Kcur_rope", il); + } + + Vcur = ggml_reshape_3d(ctx0, Vcur, n_embd_head, n_head_kv, n_tokens); + + cur = build_attn(inp_attn, + NULL, NULL, // wo will be applied after gating + Qcur, Kcur, Vcur, nullptr, nullptr, nullptr, kq_scale, il); + cb(cur, "attn_out", il); + + // attention gating: attn_out * sigmoid(gate) BEFORE o_proj + gate = ggml_sigmoid(ctx0, gate); + cb(gate, "attn_gate_sig", il); + cur = ggml_mul(ctx0, cur, gate); + cb(cur, "attn_gated", il); + + // now apply output projection + cur = build_lora_mm(model.layers[il].wo, cur); + cb(cur, "attn_o_proj", il); + } + + // dual attention normalization (post) + cur = build_norm(cur, + model.layers[il].attn_post_norm, NULL, + LLM_NORM_RMS, il); + cb(cur, "attn_post_norm", il); + + if (il == n_layer - 1 && inp_out_ids) { + cur = ggml_get_rows(ctx0, cur, inp_out_ids); + inpSA = ggml_get_rows(ctx0, inpSA, inp_out_ids); + } + + ggml_tensor * ffn_inp = ggml_add(ctx0, cur, inpSA); + cb(ffn_inp, "ffn_inp", il); + + // dual ffn normalization (pre) + cur = build_norm(ffn_inp, + model.layers[il].ffn_norm, NULL, + LLM_NORM_RMS, il); + cb(cur, "ffn_norm", il); + + // MoE or dense FFN + if ((uint32_t)il >= hparams.n_layer_dense_lead) { + // MoE layer with sigmoid routing, normalization, and scaling + ggml_tensor * moe_out = build_moe_ffn(cur, + model.layers[il].ffn_gate_inp, + model.layers[il].ffn_up_exps, + model.layers[il].ffn_gate_exps, + model.layers[il].ffn_down_exps, + model.layers[il].ffn_exp_probs_b, + n_expert, n_expert_used, + LLM_FFN_SILU, + hparams.expert_weights_norm, // norm_w (route_norm=True) + hparams.expert_weights_scale, // scale_w + hparams.expert_weights_scale, // w_scale (route_scale=2.826) + (llama_expert_gating_func_type) hparams.expert_gating_func, + il); + cb(moe_out, "ffn_moe_out", il); + + // shared expert + if (hparams.n_expert_shared > 0) { + ggml_tensor * ffn_shexp = build_ffn(cur, + model.layers[il].ffn_up_shexp, NULL, NULL, + model.layers[il].ffn_gate_shexp, NULL, NULL, + model.layers[il].ffn_down_shexp, NULL, NULL, + NULL, + LLM_FFN_SILU, LLM_FFN_PAR, il); + cb(ffn_shexp, "ffn_shexp", il); + + cur = ggml_add(ctx0, moe_out, ffn_shexp); + cb(cur, "ffn_out", il); + } else { + cur = moe_out; + } + } else { + // dense layer + cur = build_ffn(cur, + model.layers[il].ffn_up, NULL, NULL, + model.layers[il].ffn_gate, NULL, NULL, + model.layers[il].ffn_down, NULL, NULL, + NULL, + LLM_FFN_SILU, LLM_FFN_PAR, il); + cb(cur, "ffn_out", il); + } + + // dual ffn normalization (post) + cur = build_norm(cur, + model.layers[il].ffn_post_norm, NULL, + LLM_NORM_RMS, il); + cb(cur, "ffn_post_norm", il); + + cur = ggml_add(ctx0, cur, ffn_inp); + cur = build_cvec(cur, il); + cb(cur, "l_out", il); + + // input for next layer + inpL = cur; + } + + cur = inpL; + + cur = build_norm(cur, + model.output_norm, NULL, + LLM_NORM_RMS, -1); + cb(cur, "result_norm", -1); + + res->t_embd = cur; + + // lm_head + cur = build_lora_mm(model.output, cur); + cb(cur, "result_output", -1); + res->t_logits = cur; + + ggml_build_forward_expand(gf, cur); +} diff --git a/src/models/models.h b/src/models/models.h index 2fffb382df..4d7aeb4f42 100644 --- a/src/models/models.h +++ b/src/models/models.h @@ -57,6 +57,10 @@ struct llm_build_rwkv7_base : public llm_graph_context { int il) const; }; +struct llm_build_afmoe : public llm_graph_context { + llm_build_afmoe(const llama_model & model, const llm_graph_params & params); +}; + struct llm_build_apertus : public llm_graph_context { llm_build_apertus(const llama_model & model, const llm_graph_params & params); }; diff --git a/src/unicode.cpp b/src/unicode.cpp index 65f3665171..77ba4fc46b 100644 --- a/src/unicode.cpp +++ b/src/unicode.cpp @@ -729,6 +729,80 @@ static std::vector unicode_regex_split_custom_kimi_k2(const std::string return bpe_offsets; } +// AFMOE digit handling: splits digits with leading 1-2 based on total length modulo 3 +static std::vector unicode_regex_split_custom_afmoe(const std::string & text, const std::vector & offsets) { + std::vector bpe_offsets; + bpe_offsets.reserve(offsets.size()); + + const auto cpts = unicode_cpts_from_utf8(text); + + size_t start = 0; + for (auto offset : offsets) { + const size_t offset_ini = start; + const size_t offset_end = start + offset; + assert(offset_end <= cpts.size()); + start = offset_end; + + auto _get_flags = [&] (const size_t pos) -> unicode_cpt_flags { + return (offset_ini <= pos && pos < offset_end) ? unicode_cpt_flags_from_cpt(cpts[pos]) : unicode_cpt_flags{}; + }; + + size_t _prev_end = offset_ini; + auto _add_token = [&] (const size_t end) -> size_t { + assert(_prev_end <= end && end <= offset_end); + size_t len = end - _prev_end; + if (len > 0) { + bpe_offsets.push_back(len); + } + _prev_end = end; + return len; + }; + + for (size_t pos = offset_ini; pos < offset_end; ) { + const auto flags = _get_flags(pos); + + // Handle digit sequences with special splitting logic + if (flags.is_number) { + size_t digit_start = pos; + size_t digit_count = 0; + + // Count consecutive digits + while (_get_flags(pos).is_number && pos < offset_end) { + digit_count++; + pos++; + } + + // Split based on total length modulo 3 + size_t remainder = digit_count % 3; + size_t current = digit_start; + + // Emit leading 1-2 digits if needed + if (remainder > 0) { + _add_token(current + remainder); + current += remainder; + } + + // Emit groups of 3 + while (current < digit_start + digit_count) { + _add_token(current + 3); + current += 3; + } + continue; + } + + // For non-digits, just move forward + pos++; + } + + // Add any remaining content + if (_prev_end < offset_end) { + _add_token(offset_end); + } + } + + return bpe_offsets; +} + static std::vector unicode_regex_split_custom(const std::string & text, const std::string & regex_expr, const std::vector & offsets) { std::vector bpe_offsets; @@ -742,6 +816,9 @@ static std::vector unicode_regex_split_custom(const std::string & text, } else if (regex_expr == "\\p{Han}+") { // K2's first pattern - handle all K2 patterns together bpe_offsets = unicode_regex_split_custom_kimi_k2(text, offsets); + } else if (regex_expr == "\\p{AFMoE_digits}") { + // AFMOE digit pattern - use custom implementation for proper splitting + bpe_offsets = unicode_regex_split_custom_afmoe(text, offsets); } return bpe_offsets;