bigdl-core-cpp 2.6.0b20250227__py3-none-win_amd64.whl → 2.6.0b20250231__py3-none-win_amd64.whl

bigdl/cpp/gguf-py/gguf/constants.py

@@ -90,6 +90,7 @@ class Keys:
         VOCAB_SIZE                        = "{arch}.vocab_size"
         CONTEXT_LENGTH                    = "{arch}.context_length"
         EMBEDDING_LENGTH                  = "{arch}.embedding_length"
+        FEATURES_LENGTH                   = "{arch}.features_length"
         BLOCK_COUNT                       = "{arch}.block_count"
         LEADING_DENSE_BLOCK_COUNT         = "{arch}.leading_dense_block_count"
         FEED_FORWARD_LENGTH               = "{arch}.feed_forward_length"
@@ -101,6 +102,8 @@ class Keys:
         EXPERT_USED_COUNT                 = "{arch}.expert_used_count"
         EXPERT_SHARED_COUNT               = "{arch}.expert_shared_count"
         EXPERT_WEIGHTS_SCALE              = "{arch}.expert_weights_scale"
+        EXPERT_WEIGHTS_NORM               = "{arch}.expert_weights_norm"
+        EXPERT_GATING_FUNC                = "{arch}.expert_gating_func"
         POOLING_TYPE                      = "{arch}.pooling_type"
         LOGIT_SCALE                       = "{arch}.logit_scale"
         DECODER_START_TOKEN_ID            = "{arch}.decoder_start_token_id"
@@ -112,6 +115,7 @@ class Keys:
         TIME_DECAY_EXTRA_DIM              = "{arch}.time_decay_extra_dim"
         RESIDUAL_SCALE                    = "{arch}.residual_scale"
         EMBEDDING_SCALE                   = "{arch}.embedding_scale"
+        TOKEN_SHIFT_COUNT                 = "{arch}.token_shift_count"
 
     class Attention:
         HEAD_COUNT        = "{arch}.attention.head_count"
@@ -122,6 +126,8 @@ class Keys:
         VALUE_LENGTH      = "{arch}.attention.value_length"
         LAYERNORM_EPS     = "{arch}.attention.layer_norm_epsilon"
         LAYERNORM_RMS_EPS = "{arch}.attention.layer_norm_rms_epsilon"
+        GROUPNORM_EPS     = "{arch}.attention.group_norm_epsilon"
+        GROUPNORM_GROUPS  = "{arch}.attention.group_norm_groups"
         CAUSAL            = "{arch}.attention.causal"
         Q_LORA_RANK       = "{arch}.attention.q_lora_rank"
         KV_LORA_RANK      = "{arch}.attention.kv_lora_rank"
@@ -155,6 +161,14 @@ class Keys:
     class WKV:
         HEAD_SIZE = "{arch}.wkv.head_size"
 
+    class PosNet:
+        EMBEDDING_LENGTH = "{arch}.posnet.embedding_length"
+        BLOCK_COUNT      = "{arch}.posnet.block_count"
+
+    class ConvNext:
+        EMBEDDING_LENGTH = "{arch}.convnext.embedding_length"
+        BLOCK_COUNT      = "{arch}.convnext.block_count"
+
     class Tokenizer:
         MODEL                = "tokenizer.ggml.model"
         PRE                  = "tokenizer.ggml.pre"
@@ -170,7 +184,6 @@ class Keys:
         UNK_ID               = "tokenizer.ggml.unknown_token_id"
         SEP_ID               = "tokenizer.ggml.seperator_token_id"
         PAD_ID               = "tokenizer.ggml.padding_token_id"
-        CLS_ID               = "tokenizer.ggml.cls_token_id"
         MASK_ID              = "tokenizer.ggml.mask_token_id"
         ADD_BOS              = "tokenizer.ggml.add_bos_token"
         ADD_EOS              = "tokenizer.ggml.add_eos_token"
@@ -209,57 +222,63 @@ class GGUFType:
 
 
 class MODEL_ARCH(IntEnum):
-    LLAMA        = auto()
-    FALCON       = auto()
-    BAICHUAN     = auto()
-    GROK         = auto()
-    GPT2         = auto()
-    GPTJ         = auto()
-    GPTNEOX      = auto()
-    MPT          = auto()
-    STARCODER    = auto()
-    REFACT       = auto()
-    BERT         = auto()
-    NOMIC_BERT   = auto()
-    JINA_BERT_V2 = auto()
-    BLOOM        = auto()
-    STABLELM     = auto()
-    QWEN         = auto()
-    QWEN2        = auto()
-    QWEN2MOE     = auto()
-    QWEN2VL      = auto()
-    PHI2         = auto()
-    PHI3         = auto()
-    PLAMO        = auto()
-    CODESHELL    = auto()
-    ORION        = auto()
-    INTERNLM2    = auto()
-    MINICPM      = auto()
-    MINICPM3     = auto()
-    GEMMA        = auto()
-    GEMMA2       = auto()
-    STARCODER2   = auto()
-    RWKV6        = auto()
-    MAMBA        = auto()
-    XVERSE       = auto()
-    COMMAND_R    = auto()
-    DBRX         = auto()
-    OLMO         = auto()
-    OLMO2        = auto()
-    OLMOE        = auto()
-    OPENELM      = auto()
-    ARCTIC       = auto()
-    DEEPSEEK2    = auto()
-    CHATGLM      = auto()
-    BITNET       = auto()
-    T5           = auto()
-    T5ENCODER    = auto()
-    JAIS         = auto()
-    NEMOTRON     = auto()
-    EXAONE       = auto()
-    GRANITE      = auto()
-    GRANITE_MOE  = auto()
-    CHAMELEON    = auto()
+    LLAMA            = auto()
+    DECI             = auto()
+    FALCON           = auto()
+    BAICHUAN         = auto()
+    GROK             = auto()
+    GPT2             = auto()
+    GPTJ             = auto()
+    GPTNEOX          = auto()
+    MPT              = auto()
+    STARCODER        = auto()
+    REFACT           = auto()
+    BERT             = auto()
+    NOMIC_BERT       = auto()
+    JINA_BERT_V2     = auto()
+    BLOOM            = auto()
+    STABLELM         = auto()
+    QWEN             = auto()
+    QWEN2            = auto()
+    QWEN2MOE         = auto()
+    QWEN2VL          = auto()
+    PHI2             = auto()
+    PHI3             = auto()
+    PHIMOE           = auto()
+    PLAMO            = auto()
+    CODESHELL        = auto()
+    ORION            = auto()
+    INTERNLM2        = auto()
+    MINICPM          = auto()
+    MINICPM3         = auto()
+    GEMMA            = auto()
+    GEMMA2           = auto()
+    STARCODER2       = auto()
+    RWKV6            = auto()
+    RWKV6QWEN2       = auto()
+    MAMBA            = auto()
+    XVERSE           = auto()
+    COMMAND_R        = auto()
+    COHERE2          = auto()
+    DBRX             = auto()
+    OLMO             = auto()
+    OLMO2            = auto()
+    OLMOE            = auto()
+    OPENELM          = auto()
+    ARCTIC           = auto()
+    DEEPSEEK         = auto()
+    DEEPSEEK2        = auto()
+    CHATGLM          = auto()
+    BITNET           = auto()
+    T5               = auto()
+    T5ENCODER        = auto()
+    JAIS             = auto()
+    NEMOTRON         = auto()
+    EXAONE           = auto()
+    GRANITE          = auto()
+    GRANITE_MOE      = auto()
+    CHAMELEON        = auto()
+    WAVTOKENIZER_DEC = auto()
 
 
 class MODEL_TENSOR(IntEnum):
@@ -298,6 +317,7 @@ class MODEL_TENSOR(IntEnum):
     FFN_GATE_SHEXP       = auto()
     FFN_DOWN_SHEXP       = auto()
     FFN_UP_SHEXP         = auto()
+    FFN_EXP_PROBS_B      = auto()
     ATTN_Q_NORM          = auto()
     ATTN_K_NORM          = auto()
     LAYER_OUT_NORM       = auto()
@@ -315,6 +335,7 @@ class MODEL_TENSOR(IntEnum):
     TIME_MIX_LERP_V      = auto()
     TIME_MIX_LERP_R      = auto()
     TIME_MIX_LERP_G      = auto()
+    TIME_MIX_LERP_FUSED  = auto()
     TIME_MIX_LERP_W      = auto()
     TIME_MIX_FIRST       = auto()
     TIME_MIX_DECAY       = auto()
@@ -369,60 +390,82 @@ class MODEL_TENSOR(IntEnum):
     ENC_OUTPUT_NORM      = auto()
     CLS                  = auto() # classifier
     CLS_OUT              = auto() # classifier output projection
+    CONV1D               = auto()
+    CONVNEXT_DW          = auto()
+    CONVNEXT_NORM        = auto()
+    CONVNEXT_PW1         = auto()
+    CONVNEXT_PW2         = auto()
+    CONVNEXT_GAMMA       = auto()
+    POSNET_CONV1         = auto()
+    POSNET_CONV2         = auto()
+    POSNET_NORM          = auto()
+    POSNET_NORM1         = auto()
+    POSNET_NORM2         = auto()
+    POSNET_ATTN_NORM     = auto()
+    POSNET_ATTN_Q        = auto()
+    POSNET_ATTN_K        = auto()
+    POSNET_ATTN_V        = auto()
+    POSNET_ATTN_OUT      = auto()
 
 
 MODEL_ARCH_NAMES: dict[MODEL_ARCH, str] = {
-    MODEL_ARCH.LLAMA:          "llama",
-    MODEL_ARCH.FALCON:         "falcon",
-    MODEL_ARCH.BAICHUAN:       "baichuan",
-    MODEL_ARCH.GROK:           "grok",
-    MODEL_ARCH.GPT2:           "gpt2",
-    MODEL_ARCH.GPTJ:           "gptj",
-    MODEL_ARCH.GPTNEOX:        "gptneox",
-    MODEL_ARCH.MPT:            "mpt",
-    MODEL_ARCH.STARCODER:      "starcoder",
-    MODEL_ARCH.REFACT:         "refact",
-    MODEL_ARCH.BERT:           "bert",
-    MODEL_ARCH.NOMIC_BERT:     "nomic-bert",
-    MODEL_ARCH.JINA_BERT_V2:   "jina-bert-v2",
-    MODEL_ARCH.BLOOM:          "bloom",
-    MODEL_ARCH.STABLELM:       "stablelm",
-    MODEL_ARCH.QWEN:           "qwen",
-    MODEL_ARCH.QWEN2:          "qwen2",
-    MODEL_ARCH.QWEN2MOE:       "qwen2moe",
-    MODEL_ARCH.QWEN2VL:        "qwen2vl",
-    MODEL_ARCH.PHI2:           "phi2",
-    MODEL_ARCH.PHI3:           "phi3",
-    MODEL_ARCH.PLAMO:          "plamo",
-    MODEL_ARCH.CODESHELL:      "codeshell",
-    MODEL_ARCH.ORION:          "orion",
-    MODEL_ARCH.INTERNLM2:      "internlm2",
-    MODEL_ARCH.MINICPM:        "minicpm",
-    MODEL_ARCH.MINICPM3:       "minicpm3",
-    MODEL_ARCH.GEMMA:          "gemma",
-    MODEL_ARCH.GEMMA2:         "gemma2",
-    MODEL_ARCH.STARCODER2:     "starcoder2",
-    MODEL_ARCH.RWKV6:          "rwkv6",
-    MODEL_ARCH.MAMBA:          "mamba",
-    MODEL_ARCH.XVERSE:         "xverse",
-    MODEL_ARCH.COMMAND_R:      "command-r",
-    MODEL_ARCH.DBRX:           "dbrx",
-    MODEL_ARCH.OLMO:           "olmo",
-    MODEL_ARCH.OLMO2:          "olmo2",
-    MODEL_ARCH.OLMOE:          "olmoe",
-    MODEL_ARCH.OPENELM:        "openelm",
-    MODEL_ARCH.ARCTIC:         "arctic",
-    MODEL_ARCH.DEEPSEEK2:      "deepseek2",
-    MODEL_ARCH.CHATGLM:        "chatglm",
-    MODEL_ARCH.BITNET:         "bitnet",
-    MODEL_ARCH.T5:             "t5",
-    MODEL_ARCH.T5ENCODER:      "t5encoder",
-    MODEL_ARCH.JAIS:           "jais",
-    MODEL_ARCH.NEMOTRON:       "nemotron",
-    MODEL_ARCH.EXAONE:         "exaone",
-    MODEL_ARCH.GRANITE:        "granite",
-    MODEL_ARCH.GRANITE_MOE:    "granitemoe",
-    MODEL_ARCH.CHAMELEON:      "chameleon",
+    MODEL_ARCH.LLAMA:            "llama",
+    MODEL_ARCH.DECI:             "deci",
+    MODEL_ARCH.FALCON:           "falcon",
+    MODEL_ARCH.BAICHUAN:         "baichuan",
+    MODEL_ARCH.GROK:             "grok",
+    MODEL_ARCH.GPT2:             "gpt2",
+    MODEL_ARCH.GPTJ:             "gptj",
+    MODEL_ARCH.GPTNEOX:          "gptneox",
+    MODEL_ARCH.MPT:              "mpt",
+    MODEL_ARCH.STARCODER:        "starcoder",
+    MODEL_ARCH.REFACT:           "refact",
+    MODEL_ARCH.BERT:             "bert",
+    MODEL_ARCH.NOMIC_BERT:       "nomic-bert",
+    MODEL_ARCH.JINA_BERT_V2:     "jina-bert-v2",
+    MODEL_ARCH.BLOOM:            "bloom",
+    MODEL_ARCH.STABLELM:         "stablelm",
+    MODEL_ARCH.QWEN:             "qwen",
+    MODEL_ARCH.QWEN2:            "qwen2",
+    MODEL_ARCH.QWEN2MOE:         "qwen2moe",
+    MODEL_ARCH.QWEN2VL:          "qwen2vl",
+    MODEL_ARCH.PHI2:             "phi2",
+    MODEL_ARCH.PHI3:             "phi3",
+    MODEL_ARCH.PHIMOE:           "phimoe",
+    MODEL_ARCH.PLAMO:            "plamo",
+    MODEL_ARCH.CODESHELL:        "codeshell",
+    MODEL_ARCH.ORION:            "orion",
+    MODEL_ARCH.INTERNLM2:        "internlm2",
+    MODEL_ARCH.MINICPM:          "minicpm",
+    MODEL_ARCH.MINICPM3:         "minicpm3",
+    MODEL_ARCH.GEMMA:            "gemma",
+    MODEL_ARCH.GEMMA2:           "gemma2",
+    MODEL_ARCH.STARCODER2:       "starcoder2",
+    MODEL_ARCH.RWKV6:            "rwkv6",
+    MODEL_ARCH.RWKV6QWEN2:       "rwkv6qwen2",
+    MODEL_ARCH.MAMBA:            "mamba",
+    MODEL_ARCH.XVERSE:           "xverse",
+    MODEL_ARCH.COMMAND_R:        "command-r",
+    MODEL_ARCH.COHERE2:          "cohere2",
+    MODEL_ARCH.DBRX:             "dbrx",
+    MODEL_ARCH.OLMO:             "olmo",
+    MODEL_ARCH.OLMO2:            "olmo2",
+    MODEL_ARCH.OLMOE:            "olmoe",
+    MODEL_ARCH.OPENELM:          "openelm",
+    MODEL_ARCH.ARCTIC:           "arctic",
+    MODEL_ARCH.DEEPSEEK:         "deepseek",
+    MODEL_ARCH.DEEPSEEK2:        "deepseek2",
+    MODEL_ARCH.CHATGLM:          "chatglm",
+    MODEL_ARCH.BITNET:           "bitnet",
+    MODEL_ARCH.T5:               "t5",
+    MODEL_ARCH.T5ENCODER:        "t5encoder",
+    MODEL_ARCH.JAIS:             "jais",
+    MODEL_ARCH.NEMOTRON:         "nemotron",
+    MODEL_ARCH.EXAONE:           "exaone",
+    MODEL_ARCH.GRANITE:          "granite",
+    MODEL_ARCH.GRANITE_MOE:      "granitemoe",
+    MODEL_ARCH.CHAMELEON:        "chameleon",
+    MODEL_ARCH.WAVTOKENIZER_DEC: "wavtokenizer-dec",
 }
 
 TENSOR_NAMES: dict[MODEL_TENSOR, str] = {
@@ -463,6 +506,7 @@ TENSOR_NAMES: dict[MODEL_TENSOR, str] = {
     MODEL_TENSOR.FFN_GATE_EXP:              "blk.{bid}.ffn_gate_exps",
     MODEL_TENSOR.FFN_DOWN_EXP:              "blk.{bid}.ffn_down_exps",
     MODEL_TENSOR.FFN_UP_EXP:                "blk.{bid}.ffn_up_exps",
+    MODEL_TENSOR.FFN_EXP_PROBS_B:           "blk.{bid}.exp_probs_b",
     MODEL_TENSOR.LAYER_OUT_NORM:            "blk.{bid}.layer_output_norm",
     MODEL_TENSOR.SSM_IN:                    "blk.{bid}.ssm_in",
     MODEL_TENSOR.SSM_CONV1D:                "blk.{bid}.ssm_conv1d",
@@ -478,6 +522,7 @@ TENSOR_NAMES: dict[MODEL_TENSOR, str] = {
     MODEL_TENSOR.TIME_MIX_LERP_V:           "blk.{bid}.time_mix_lerp_v",
     MODEL_TENSOR.TIME_MIX_LERP_R:           "blk.{bid}.time_mix_lerp_r",
     MODEL_TENSOR.TIME_MIX_LERP_G:           "blk.{bid}.time_mix_lerp_g",
+    MODEL_TENSOR.TIME_MIX_LERP_FUSED:       "blk.{bid}.time_mix_lerp_fused",
     MODEL_TENSOR.TIME_MIX_LERP_W:           "blk.{bid}.time_mix_lerp_w",
     MODEL_TENSOR.TIME_MIX_FIRST:            "blk.{bid}.time_mix_first",
     MODEL_TENSOR.TIME_MIX_DECAY:            "blk.{bid}.time_mix_decay",
@@ -532,6 +577,22 @@ TENSOR_NAMES: dict[MODEL_TENSOR, str] = {
     MODEL_TENSOR.ENC_OUTPUT_NORM:           "enc.output_norm",
     MODEL_TENSOR.CLS:                       "cls",
     MODEL_TENSOR.CLS_OUT:                   "cls.output",
+    MODEL_TENSOR.CONV1D:                    "conv1d",
+    MODEL_TENSOR.CONVNEXT_DW:               "convnext.{bid}.dw",
+    MODEL_TENSOR.CONVNEXT_NORM:             "convnext.{bid}.norm",
+    MODEL_TENSOR.CONVNEXT_PW1:              "convnext.{bid}.pw1",
+    MODEL_TENSOR.CONVNEXT_PW2:              "convnext.{bid}.pw2",
+    MODEL_TENSOR.CONVNEXT_GAMMA:            "convnext.{bid}.gamma",
+    MODEL_TENSOR.POSNET_CONV1:              "posnet.{bid}.conv1",
+    MODEL_TENSOR.POSNET_CONV2:              "posnet.{bid}.conv2",
+    MODEL_TENSOR.POSNET_NORM:               "posnet.{bid}.norm",
+    MODEL_TENSOR.POSNET_NORM1:              "posnet.{bid}.norm1",
+    MODEL_TENSOR.POSNET_NORM2:              "posnet.{bid}.norm2",
+    MODEL_TENSOR.POSNET_ATTN_NORM:          "posnet.{bid}.attn_norm",
+    MODEL_TENSOR.POSNET_ATTN_Q:             "posnet.{bid}.attn_q",
+    MODEL_TENSOR.POSNET_ATTN_K:             "posnet.{bid}.attn_k",
+    MODEL_TENSOR.POSNET_ATTN_V:             "posnet.{bid}.attn_v",
+    MODEL_TENSOR.POSNET_ATTN_OUT:           "posnet.{bid}.attn_output",
 }
 
 MODEL_TENSORS: dict[MODEL_ARCH, list[MODEL_TENSOR]] = {
@@ -555,6 +616,26 @@ MODEL_TENSORS: dict[MODEL_ARCH, list[MODEL_TENSOR]] = {
         MODEL_TENSOR.FFN_DOWN_EXP,
         MODEL_TENSOR.FFN_UP_EXP,
     ],
+    MODEL_ARCH.DECI: [
+        MODEL_TENSOR.TOKEN_EMBD,
+        MODEL_TENSOR.OUTPUT_NORM,
+        MODEL_TENSOR.OUTPUT,
+        MODEL_TENSOR.ROPE_FREQS,
+        MODEL_TENSOR.ATTN_NORM,
+        MODEL_TENSOR.ATTN_Q,
+        MODEL_TENSOR.ATTN_K,
+        MODEL_TENSOR.ATTN_V,
+        MODEL_TENSOR.ATTN_OUT,
+        MODEL_TENSOR.ATTN_ROT_EMBD,
+        MODEL_TENSOR.FFN_GATE_INP,
+        MODEL_TENSOR.FFN_NORM,
+        MODEL_TENSOR.FFN_GATE,
+        MODEL_TENSOR.FFN_DOWN,
+        MODEL_TENSOR.FFN_UP,
+        MODEL_TENSOR.FFN_GATE_EXP,
+        MODEL_TENSOR.FFN_DOWN_EXP,
+        MODEL_TENSOR.FFN_UP_EXP,
+    ],
     MODEL_ARCH.GROK: [
         MODEL_TENSOR.TOKEN_EMBD,
         MODEL_TENSOR.OUTPUT_NORM,
@@ -865,6 +946,24 @@ MODEL_TENSORS: dict[MODEL_ARCH, list[MODEL_TENSOR]] = {
         MODEL_TENSOR.FFN_DOWN,
         MODEL_TENSOR.FFN_UP,
     ],
+    MODEL_ARCH.PHIMOE: [
+        MODEL_TENSOR.TOKEN_EMBD,
+        MODEL_TENSOR.OUTPUT_NORM,
+        MODEL_TENSOR.OUTPUT,
+        MODEL_TENSOR.ROPE_FACTORS_LONG,
+        MODEL_TENSOR.ROPE_FACTORS_SHORT,
+        MODEL_TENSOR.ATTN_NORM,
+        MODEL_TENSOR.ATTN_QKV,
+        MODEL_TENSOR.ATTN_Q,
+        MODEL_TENSOR.ATTN_K,
+        MODEL_TENSOR.ATTN_V,
+        MODEL_TENSOR.ATTN_OUT,
+        MODEL_TENSOR.FFN_NORM,
+        MODEL_TENSOR.FFN_GATE_INP,
+        MODEL_TENSOR.FFN_GATE_EXP,
+        MODEL_TENSOR.FFN_DOWN_EXP,
+        MODEL_TENSOR.FFN_UP_EXP,
+    ],
     MODEL_ARCH.CODESHELL: [
         MODEL_TENSOR.TOKEN_EMBD,
         MODEL_TENSOR.POS_EMBD,
@@ -1008,6 +1107,7 @@ MODEL_TENSORS: dict[MODEL_ARCH, list[MODEL_TENSOR]] = {
         MODEL_TENSOR.TIME_MIX_LERP_R,
         MODEL_TENSOR.TIME_MIX_LERP_G,
         MODEL_TENSOR.TIME_MIX_LERP_W,
+        MODEL_TENSOR.TIME_MIX_LERP_FUSED,
         MODEL_TENSOR.TIME_MIX_FIRST,
         MODEL_TENSOR.TIME_MIX_DECAY,
         MODEL_TENSOR.TIME_MIX_DECAY_W1,
@@ -1024,6 +1124,35 @@ MODEL_TENSORS: dict[MODEL_ARCH, list[MODEL_TENSOR]] = {
         MODEL_TENSOR.CHANNEL_MIX_RECEPTANCE,
         MODEL_TENSOR.CHANNEL_MIX_VALUE,
     ],
+    MODEL_ARCH.RWKV6QWEN2: [
+        MODEL_TENSOR.TOKEN_EMBD,
+        MODEL_TENSOR.OUTPUT_NORM,
+        MODEL_TENSOR.OUTPUT,
+        MODEL_TENSOR.ATTN_NORM,
+        MODEL_TENSOR.TIME_MIX_W1,
+        MODEL_TENSOR.TIME_MIX_W2,
+        MODEL_TENSOR.TIME_MIX_LERP_X,
+        MODEL_TENSOR.TIME_MIX_LERP_K,
+        MODEL_TENSOR.TIME_MIX_LERP_V,
+        MODEL_TENSOR.TIME_MIX_LERP_R,
+        MODEL_TENSOR.TIME_MIX_LERP_G,
+        MODEL_TENSOR.TIME_MIX_LERP_W,
+        MODEL_TENSOR.TIME_MIX_LERP_FUSED,
+        MODEL_TENSOR.TIME_MIX_FIRST,
+        MODEL_TENSOR.TIME_MIX_DECAY,
+        MODEL_TENSOR.TIME_MIX_DECAY_W1,
+        MODEL_TENSOR.TIME_MIX_DECAY_W2,
+        MODEL_TENSOR.TIME_MIX_KEY,
+        MODEL_TENSOR.TIME_MIX_VALUE,
+        MODEL_TENSOR.TIME_MIX_RECEPTANCE,
+        MODEL_TENSOR.TIME_MIX_GATE,
+        MODEL_TENSOR.TIME_MIX_LN,
+        MODEL_TENSOR.TIME_MIX_OUTPUT,
+        MODEL_TENSOR.FFN_NORM,
+        MODEL_TENSOR.FFN_GATE,
+        MODEL_TENSOR.FFN_DOWN,
+        MODEL_TENSOR.FFN_UP,
+    ],
     MODEL_ARCH.MAMBA: [
         MODEL_TENSOR.TOKEN_EMBD,
         MODEL_TENSOR.OUTPUT_NORM,
@@ -1067,6 +1196,18 @@ MODEL_TENSORS: dict[MODEL_ARCH, list[MODEL_TENSOR]] = {
         MODEL_TENSOR.ATTN_K_NORM,
         MODEL_TENSOR.ATTN_Q_NORM,
     ],
+    MODEL_ARCH.COHERE2: [
+        MODEL_TENSOR.TOKEN_EMBD,
+        MODEL_TENSOR.OUTPUT_NORM,
+        MODEL_TENSOR.ATTN_NORM,
+        MODEL_TENSOR.ATTN_Q,
+        MODEL_TENSOR.ATTN_K,
+        MODEL_TENSOR.ATTN_V,
+        MODEL_TENSOR.ATTN_OUT,
+        MODEL_TENSOR.FFN_GATE,
+        MODEL_TENSOR.FFN_DOWN,
+        MODEL_TENSOR.FFN_UP,
+    ],
     MODEL_ARCH.DBRX: [
         MODEL_TENSOR.TOKEN_EMBD,
         MODEL_TENSOR.OUTPUT_NORM,
@@ -1158,6 +1299,29 @@ MODEL_TENSORS: dict[MODEL_ARCH, list[MODEL_TENSOR]] = {
         MODEL_TENSOR.FFN_DOWN_EXP,
         MODEL_TENSOR.FFN_UP_EXP,
     ],
+    MODEL_ARCH.DEEPSEEK: [
+        MODEL_TENSOR.TOKEN_EMBD,
+        MODEL_TENSOR.OUTPUT_NORM,
+        MODEL_TENSOR.OUTPUT,
+        MODEL_TENSOR.ROPE_FREQS,
+        MODEL_TENSOR.ATTN_NORM,
+        MODEL_TENSOR.ATTN_Q,
+        MODEL_TENSOR.ATTN_K,
+        MODEL_TENSOR.ATTN_V,
+        MODEL_TENSOR.ATTN_OUT,
+        MODEL_TENSOR.ATTN_ROT_EMBD,
+        MODEL_TENSOR.FFN_GATE_INP,
+        MODEL_TENSOR.FFN_NORM,
+        MODEL_TENSOR.FFN_GATE,
+        MODEL_TENSOR.FFN_DOWN,
+        MODEL_TENSOR.FFN_UP,
+        MODEL_TENSOR.FFN_GATE_EXP,
+        MODEL_TENSOR.FFN_DOWN_EXP,
+        MODEL_TENSOR.FFN_UP_EXP,
+        MODEL_TENSOR.FFN_GATE_SHEXP,
+        MODEL_TENSOR.FFN_DOWN_SHEXP,
+        MODEL_TENSOR.FFN_UP_SHEXP,
+    ],
     MODEL_ARCH.DEEPSEEK2: [
         MODEL_TENSOR.TOKEN_EMBD,
         MODEL_TENSOR.OUTPUT_NORM,
@@ -1184,6 +1348,7 @@ MODEL_TENSORS: dict[MODEL_ARCH, list[MODEL_TENSOR]] = {
         MODEL_TENSOR.FFN_GATE_SHEXP,
         MODEL_TENSOR.FFN_DOWN_SHEXP,
         MODEL_TENSOR.FFN_UP_SHEXP,
+        MODEL_TENSOR.FFN_EXP_PROBS_B,
     ],
     MODEL_ARCH.CHATGLM : [
         MODEL_TENSOR.TOKEN_EMBD,
@@ -1192,6 +1357,9 @@ MODEL_TENSORS: dict[MODEL_ARCH, list[MODEL_TENSOR]] = {
         MODEL_TENSOR.OUTPUT,
         MODEL_TENSOR.ATTN_NORM,
         MODEL_TENSOR.ATTN_QKV,
+        MODEL_TENSOR.ATTN_Q,
+        MODEL_TENSOR.ATTN_K,
+        MODEL_TENSOR.ATTN_V,
         MODEL_TENSOR.ATTN_OUT,
         MODEL_TENSOR.FFN_NORM,
         MODEL_TENSOR.FFN_DOWN,
@@ -1347,6 +1515,28 @@ MODEL_TENSORS: dict[MODEL_ARCH, list[MODEL_TENSOR]] = {
         MODEL_TENSOR.FFN_DOWN,
         MODEL_TENSOR.FFN_UP,
     ],
+    MODEL_ARCH.WAVTOKENIZER_DEC: [
+        MODEL_TENSOR.TOKEN_EMBD,
+        MODEL_TENSOR.TOKEN_EMBD_NORM,
+        MODEL_TENSOR.CONV1D,
+        MODEL_TENSOR.CONVNEXT_DW,
+        MODEL_TENSOR.CONVNEXT_NORM,
+        MODEL_TENSOR.CONVNEXT_PW1,
+        MODEL_TENSOR.CONVNEXT_PW2,
+        MODEL_TENSOR.CONVNEXT_GAMMA,
+        MODEL_TENSOR.OUTPUT,
+        MODEL_TENSOR.OUTPUT_NORM,
+        MODEL_TENSOR.POSNET_CONV1,
+        MODEL_TENSOR.POSNET_CONV2,
+        MODEL_TENSOR.POSNET_NORM,
+        MODEL_TENSOR.POSNET_NORM1,
+        MODEL_TENSOR.POSNET_NORM2,
+        MODEL_TENSOR.POSNET_ATTN_NORM,
+        MODEL_TENSOR.POSNET_ATTN_Q,
+        MODEL_TENSOR.POSNET_ATTN_K,
+        MODEL_TENSOR.POSNET_ATTN_V,
+        MODEL_TENSOR.POSNET_ATTN_OUT,
+    ],
     # TODO
 }
 
@@ -1356,6 +1546,10 @@ MODEL_TENSOR_SKIP: dict[MODEL_ARCH, list[MODEL_TENSOR]] = {
         MODEL_TENSOR.ROPE_FREQS,
         MODEL_TENSOR.ATTN_ROT_EMBD,
     ],
+    MODEL_ARCH.DECI: [
+        MODEL_TENSOR.ROPE_FREQS,
+        MODEL_TENSOR.ATTN_ROT_EMBD,
+    ],
     MODEL_ARCH.BAICHUAN: [
         MODEL_TENSOR.ROPE_FREQS,
         MODEL_TENSOR.ATTN_ROT_EMBD,
@@ -1380,6 +1574,10 @@ MODEL_TENSOR_SKIP: dict[MODEL_ARCH, list[MODEL_TENSOR]] = {
         MODEL_TENSOR.ROPE_FREQS,
         MODEL_TENSOR.ATTN_ROT_EMBD,
     ],
+    MODEL_ARCH.DEEPSEEK: [
+        MODEL_TENSOR.ROPE_FREQS,
+        MODEL_TENSOR.ATTN_ROT_EMBD,
+    ],
     MODEL_ARCH.DEEPSEEK2: [
         MODEL_TENSOR.ROPE_FREQS,
         MODEL_TENSOR.ATTN_ROT_EMBD,
@@ -1454,6 +1652,11 @@ class GGMLQuantizationType(IntEnum):
     TQ2_0   = 35
 
 
+class ExpertGatingFuncType(IntEnum):
+    SOFTMAX  = 1
+    SIGMOID  = 2
+
+
 # TODO: add GGMLFileType from ggml_ftype in ggml.h
 
 
@@ -1636,7 +1839,6 @@ KEY_TOKENIZER_EOM_ID     = Keys.Tokenizer.EOM_ID
 KEY_TOKENIZER_UNK_ID     = Keys.Tokenizer.UNK_ID
 KEY_TOKENIZER_SEP_ID     = Keys.Tokenizer.SEP_ID
 KEY_TOKENIZER_PAD_ID     = Keys.Tokenizer.PAD_ID
-KEY_TOKENIZER_CLS_ID     = Keys.Tokenizer.CLS_ID
 KEY_TOKENIZER_MASK_ID    = Keys.Tokenizer.MASK_ID
 KEY_TOKENIZER_HF_JSON    = Keys.Tokenizer.HF_JSON
 KEY_TOKENIZER_RWKV       = Keys.Tokenizer.RWKV

bigdl/cpp/gguf-py/gguf/gguf_writer.py

@@ -26,6 +26,7 @@ from .constants import (
     RopeScalingType,
     PoolingType,
     TokenType,
+    ExpertGatingFuncType,
 )
 
 from .quants import quant_shape_from_byte_shape
@@ -631,6 +632,21 @@ class GGUFWriter:
     def add_embedding_length(self, length: int) -> None:
         self.add_uint32(Keys.LLM.EMBEDDING_LENGTH.format(arch=self.arch), length)
 
+    def add_features_length(self, length: int) -> None:
+        self.add_uint32(Keys.LLM.FEATURES_LENGTH.format(arch=self.arch), length)
+
+    def add_posnet_embedding_length(self, length: int) -> None:
+        self.add_uint32(Keys.PosNet.EMBEDDING_LENGTH.format(arch=self.arch), length)
+
+    def add_posnet_block_count(self, length: int) -> None:
+        self.add_uint32(Keys.PosNet.BLOCK_COUNT.format(arch=self.arch), length)
+
+    def add_convnext_embedding_length(self, length: int) -> None:
+        self.add_uint32(Keys.ConvNext.EMBEDDING_LENGTH.format(arch=self.arch), length)
+
+    def add_convnext_block_count(self, length: int) -> None:
+        self.add_uint32(Keys.ConvNext.BLOCK_COUNT.format(arch=self.arch), length)
+
     def add_block_count(self, length: int) -> None:
         self.add_uint32(Keys.LLM.BLOCK_COUNT.format(arch=self.arch), length)
 
@@ -700,6 +716,12 @@ class GGUFWriter:
     def add_expert_weights_scale(self, value: float) -> None:
         self.add_float32(Keys.LLM.EXPERT_WEIGHTS_SCALE.format(arch=self.arch), value)
 
+    def add_expert_weights_norm(self, value: bool) -> None:
+        self.add_bool(Keys.LLM.EXPERT_WEIGHTS_NORM.format(arch=self.arch), value)
+
+    def add_expert_gating_func(self, value: ExpertGatingFuncType) -> None:
+        self.add_uint32(Keys.LLM.EXPERT_GATING_FUNC.format(arch=self.arch), value.value)
+
     def add_swin_norm(self, value: bool) -> None:
         self.add_bool(Keys.LLM.SWIN_NORM.format(arch=self.arch), value)
 
@@ -721,12 +743,21 @@ class GGUFWriter:
     def add_wkv_head_size(self, size: int) -> None:
         self.add_uint32(Keys.WKV.HEAD_SIZE.format(arch=self.arch), size)
 
+    def add_token_shift_count(self, count: int) -> None:
+        self.add_uint32(Keys.LLM.TOKEN_SHIFT_COUNT.format(arch=self.arch), count)
+
     def add_layer_norm_eps(self, value: float) -> None:
         self.add_float32(Keys.Attention.LAYERNORM_EPS.format(arch=self.arch), value)
 
     def add_layer_norm_rms_eps(self, value: float) -> None:
         self.add_float32(Keys.Attention.LAYERNORM_RMS_EPS.format(arch=self.arch), value)
 
+    def add_group_norm_eps(self, value: float) -> None:
+        self.add_float32(Keys.Attention.GROUPNORM_EPS.format(arch=self.arch), value)
+
+    def add_group_norm_groups(self, value: int) -> None:
+        self.add_uint32(Keys.Attention.GROUPNORM_GROUPS.format(arch=self.arch), value)
+
     def add_causal_attention(self, value: bool) -> None:
         self.add_bool(Keys.Attention.CAUSAL.format(arch=self.arch), value)
 
@@ -826,9 +857,6 @@ class GGUFWriter:
     def add_pad_token_id(self, id: int) -> None:
         self.add_uint32(Keys.Tokenizer.PAD_ID, id)
 
-    def add_cls_token_id(self, id: int) -> None:
-        self.add_uint32(Keys.Tokenizer.CLS_ID, id)
-
     def add_mask_token_id(self, id: int) -> None:
         self.add_uint32(Keys.Tokenizer.MASK_ID, id)