llama: add initial support for Falcon-H1 model family

convert_hf_to_gguf.py

@@ -686,6 +686,14 @@ def get_vocab_base_pre(self, tokenizer) -> str:
         if chkhsh == "9d032fcbd5501f4a38150912590928bfb36091efb5df11b8e2124b0390e3fb1e":
             # ref: https://huggingface.co/tiiuae/Falcon3-7B-Base
             res = "falcon3"
+        if (
+            chkhsh == "60476e1243776c4fb1b993dbd7a5f15ac22f83c80afdf425fa5ae01c8d44ef86" or
+            chkhsh == "3eda48b4c4dc7de733d1a8b3e3b4a85243dbbf704da2ee9d42c6beced8897896" or
+            chkhsh == "48f8e02c0359c0bbdd82f26909171fac1c18a457bb47573ed1fe3bbb2c1cfd4b" or
+            chkhsh == "a6b57017d60e6edb4d88ecc2845188e0eb333a70357e45dcc9b53964a73bbae6"
+        ):
+            # ref: https://huggingface.co/collections/tiiuae/falcon-h1-6819f2795bc406da60fab8df
+            res = "falcon_h1"
         if chkhsh == "8e62295832751ca1e8f92f2226f403dea30dc5165e448b5bfa05af5340c64ec7":
             # ref: https://huggingface.co/BAAI/bge-large-zh-v1.5
             res = "bert-bge-large"
@@ -4896,7 +4904,7 @@ def set_vocab(self):
     def set_gguf_parameters(self):
         d_model = self.find_hparam(["hidden_size", "d_model", "dim"])
         d_conv  = self.find_hparam(["conv_kernel",       "d_conv"],  optional=True) or 4
-        d_inner = self.find_hparam(["intermediate_size", "d_inner"], optional=True) or 2 * d_model
+        d_inner = self.find_hparam(["mamba_d_ssm", "intermediate_size", "d_inner"], optional=True) or 2 * d_model
         d_state = self.find_hparam(["state_size",        "d_state"], optional=True) or 128
         head_dim = self.find_hparam(["head_dim"],                    optional=True) or 64
         n_group = self.find_hparam(["n_groups"],                     optional=True) or 1
@@ -4905,8 +4913,10 @@ def set_gguf_parameters(self):
 
         # Fail early for models which don't have a block expansion factor of 2
         # TODO: does this really matter?
-        assert d_inner == 2 * d_model
-        assert d_inner % head_dim == 0
+        # skip the assertion for FalconH1 Model
+        if self.model_arch != gguf.MODEL_ARCH.FALCON_H1:
+            assert d_inner == 2 * d_model
+            assert d_inner % head_dim == 0
 
         self.gguf_writer.add_context_length(2**20)  # arbitrary value; for those who use the default
         self.gguf_writer.add_embedding_length(d_model)
@@ -4943,7 +4953,7 @@ def modify_tensors(self, data_torch: Tensor, name: str, bid: int | None) -> Iter
             data_torch = data_torch.reshape((*data_torch.shape, 1))
         elif self.match_model_tensor_name(new_name, gguf.MODEL_TENSOR.SSM_NORM, bid):
             d_model = self.find_hparam(["hidden_size", "d_model", "dim"])
-            d_inner = self.find_hparam(["intermediate_size", "d_inner"], optional=True) or 2 * d_model
+            d_inner = self.find_hparam(["mamba_d_ssm", "intermediate_size", "d_inner"], optional=True) or 2 * d_model
             n_group = self.hparams.get("n_groups", 1)
             data_torch = data_torch.reshape((n_group, d_inner // n_group))
 
@@ -6535,6 +6545,137 @@ def set_gguf_parameters(self):
         super().set_gguf_parameters()
         self.gguf_writer.add_audio_stack_factor(self.global_config["stack_factor"])
 
+
+@ModelBase.register("FalconH1ForCausalLM")
+class FalconH1Model(Mamba2Model):
+    model_arch = gguf.MODEL_ARCH.FALCON_H1
+
+    def __init__(self, *args, **kwargs):
+        # Set the hparam prefixes for Falcon Mamba2
+        self.hparam_prefixes = ["mamba"]
+
+        # Initialize the base Mamba2Model
+        super().__init__(*args, **kwargs)
+
+        # Use Llama conversion for attention
+        self._transformer_model_class = LlamaModel
+
+        # n_group and d_inner are used during reshape_tensors for mamaba2
+        self.d_model = self.find_hparam(["hidden_size", "d_model"])
+        self.n_group = self.find_hparam(["n_groups"])
+        self.d_inner = self.find_hparam(["expand"]) * self.d_model
+
+        # Initialize any Falcon Mamba2 specific attributes
+        self.has_attention = True  # Falcon Mamba2 has attention components
+
+        # Load Falcon-H1 multipliers from hyperparameters
+        self.attention_in_multiplier = self.find_hparam(["attention_in_multiplier"], optional=True)
+        self.attention_out_multiplier = self.find_hparam(["attention_out_multiplier"], optional=True)
+        self.ssm_in_multiplier = self.find_hparam(["ssm_in_multiplier"], optional=True)
+        self.ssm_out_multiplier = self.find_hparam(["ssm_out_multiplier"], optional=True)
+        self.mlp_multipliers = self.find_hparam(["mlp_multipliers"], optional=True)
+        self.ssm_multipliers = self.find_hparam(["ssm_multipliers"], optional=True)
+        self.intermediate_size = self.find_hparam(["intermediate_size"])
+        self.key_multiplier = self.find_hparam(["key_multiplier"], optional=True)
+
+    def find_hparam(self, keys: Iterable[str], *args, **kwargs) -> Any:
+        prefixed = []
+        for pfx in self.hparam_prefixes:
+            prefixed.extend(
+                "_".join([pfx, k])
+                for k in keys
+            )
+        keys = list(keys) + prefixed
+        return super().find_hparam(keys, *args, **kwargs)
+
+    def set_vocab(self):
+        self._set_vocab_gpt2()
+
+    def _generate_mup_vector(self, block_id: int) -> torch.Tensor:
+        zxbcdt_multipliers = self.hparams["ssm_multipliers"]
+        intermediate_size = self.hparams["mamba_d_ssm"]
+        groups_time_state_size = self.hparams["mamba_n_groups"] * self.hparams["mamba_d_state"]
+        vector_shape = (2 * intermediate_size + 2 * groups_time_state_size + self.hparams["mamba_n_heads"])
+
+        mup_vector = torch.ones(1, 1, vector_shape, dtype=torch.float64)
+        mup_vector[:, :, :intermediate_size] *= zxbcdt_multipliers[0]
+        mup_vector[:, :, intermediate_size:2 * intermediate_size] *= zxbcdt_multipliers[1]
+        mup_vector[:, :, 2 * intermediate_size:2 * intermediate_size + groups_time_state_size] *= zxbcdt_multipliers[2]
+        mup_vector[:, :, 2 * intermediate_size + groups_time_state_size:2 * intermediate_size + 2 * groups_time_state_size] *= zxbcdt_multipliers[3]
+        mup_vector[:, :, 2 * intermediate_size + 2 * groups_time_state_size:] *= zxbcdt_multipliers[4]
+
+        return mup_vector
+
+    def modify_tensors(self, data_torch: Tensor, name: str, bid: int | None) -> Iterable[tuple[str, Tensor]]:
+        tensors = list(super().modify_tensors(data_torch, name, bid))
+        tensor = tensors[0][1] 
+
+        if "down_proj" in name:
+            tensor = tensor  * self.mlp_multipliers[1]
+        elif "gate_proj" in name:
+            tensor = tensor * self.mlp_multipliers[0]
+        elif "k_proj" in name:
+            tensor = tensor * self.key_multiplier * self.attention_in_multiplier
+        elif "q_proj" in name:
+            tensor = tensor * self.attention_in_multiplier
+        elif "v_proj" in name:
+            tensor = tensor * self.attention_in_multiplier
+        elif "o_proj" in name:
+            tensor = tensor * self.attention_out_multiplier
+        elif "out_proj" in name:
+            tensor = tensor * self.ssm_out_multiplier
+        elif "in_proj" in name:
+            tensor = tensor * self.ssm_in_multiplier
+            zxbcdt_multipliers = self.hparams["ssm_multipliers"]
+            intermediate_size = self.hparams["mamba_d_ssm"]
+            groups_time_state_size = self.hparams["mamba_n_groups"] * self.hparams["mamba_d_state"]
+            tensor[:intermediate_size, :] *= zxbcdt_multipliers[0]
+            tensor[intermediate_size:2 * intermediate_size, :] *= zxbcdt_multipliers[1]
+            tensor[2 * intermediate_size:2 * intermediate_size + groups_time_state_size, :] *= zxbcdt_multipliers[2]
+            tensor[2 * intermediate_size + groups_time_state_size:2 * intermediate_size + 2 * groups_time_state_size, :] *= zxbcdt_multipliers[3]
+            tensor[2 * intermediate_size + 2 * groups_time_state_size:, :] *= zxbcdt_multipliers[4]
+        elif "lm_head" in name:
+            tensor = tensor * self.hparams["lm_head_multiplier"]
+        elif "embed_tokens" in name:
+            tensor = tensor * self.hparams["embedding_multiplier"]
+
+        tensors = [(tensors[0][0], tensor)]
+        return tensors
+
+    def set_gguf_parameters(self):
+        super().set_gguf_parameters()
+
+        ## General Params ##
+        self.gguf_writer.add_block_count(self.block_count)
+        self.gguf_writer.add_context_length(self.hparams.get("max_position_embeddings", 0))
+        self.gguf_writer.add_vocab_size(self.hparams["vocab_size"])
+        self.gguf_writer.add_add_bos_token(False)
+        self.gguf_writer.add_feed_forward_length(self.hparams["intermediate_size"])
+
+        ## Mamba mixer params ##
+        self.gguf_writer.add_ssm_conv_kernel(self.find_hparam(["conv_kernel", "d_conv"]))
+        self.gguf_writer.add_ssm_group_count(self.n_group)
+        self.gguf_writer.add_ssm_inner_size(self.find_hparam(["mamba_d_ssm"]))
+        self.gguf_writer.add_ssm_head_dim(d_head := self.find_hparam(["d_head"]))
+        self.gguf_writer.add_ssm_time_step_rank(self.find_hparam(["n_heads"]))
+
+        ## Attention params ##
+        self.gguf_writer.add_head_count(self.hparams["num_attention_heads"])
+        self.gguf_writer.add_head_count_kv(self.hparams["num_key_value_heads"] if "num_key_value_heads" in self.hparams else self.hparams["num_attention_heads"])
+        self.gguf_writer.add_layer_norm_rms_eps(self.hparams["rms_norm_eps"])
+        self.gguf_writer.add_key_length(self.hparams["head_dim"])
+        self.gguf_writer.add_value_length(self.hparams["head_dim"])
+
+        ## Validation ##
+        assert self.hparams.get("hidden_act") in [None, "silu"], "Only SILU activation supported"
+        assert self.d_inner % d_head == 0, f"SSM inner size {self.d_inner} not a multiple of head dim {d_head}"
+        self.gguf_writer.add_head_count_kv(self.find_hparam(["num_key_value_heads"], optional=True) or 
+            self.find_hparam(["num_attention_heads"]))
+
+
+        # Add any other Falcon Mamba2 specific configuration
+        self.gguf_writer.add_rope_freq_base(self.find_hparam(["rope_theta"]))
+
 ###### CONVERSION LOGIC ######
 
 

convert_hf_to_gguf_update.py

@@ -86,6 +86,7 @@ class TOKENIZER_TYPE(IntEnum):
     {"name": "falcon",           "tokt": TOKENIZER_TYPE.BPE, "repo": "https://huggingface.co/tiiuae/falcon-7b", },
     {"name": "bert-bge",         "tokt": TOKENIZER_TYPE.WPM, "repo": "https://huggingface.co/BAAI/bge-small-en-v1.5", },
     {"name": "falcon3",          "tokt": TOKENIZER_TYPE.BPE, "repo": "https://huggingface.co/tiiuae/Falcon3-7B-Base", },
+    {"name": "falcon-h1",        "tokt": TOKENIZER_TYPE.BPE, "repo": "https://huggingface.co/tiiuae/Falcon-H1-7B-Base", },
     {"name": "bert-bge-large",   "tokt": TOKENIZER_TYPE.BPE, "repo": "https://huggingface.co/BAAI/bge-large-zh-v1.5", },
     {"name": "mpt",              "tokt": TOKENIZER_TYPE.BPE, "repo": "https://huggingface.co/mosaicml/mpt-7b", },
     {"name": "starcoder",        "tokt": TOKENIZER_TYPE.BPE, "repo": "https://huggingface.co/bigcode/starcoder2-3b", },

gguf-py/gguf/constants.py

@@ -172,6 +172,7 @@ class SSM:
         TIME_STEP_RANK = "{arch}.ssm.time_step_rank"
         GROUP_COUNT    = "{arch}.ssm.group_count"
         DT_B_C_RMS     = "{arch}.ssm.dt_b_c_rms"
+        HEAD_DIM       = "{arch}.ssm.head_dim"
 
     class WKV:
         HEAD_SIZE = "{arch}.wkv.head_size"
@@ -288,6 +289,7 @@ class MODEL_ARCH(IntEnum):
     LLAMA4           = auto()
     DECI             = auto()
     FALCON           = auto()
+    FALCON_H1        = auto()
     BAICHUAN         = auto()
     GROK             = auto()
     GPT2             = auto()
@@ -660,6 +662,7 @@ class MODEL_TENSOR(IntEnum):
     MODEL_ARCH.DOTS1:            "dots1",
     MODEL_ARCH.ARCEE:            "arcee",
     MODEL_ARCH.ERNIE4_5:         "ernie4_5",
+    MODEL_ARCH.FALCON_H1:        "falcon_h1",
 }
 
 VISION_PROJECTOR_TYPE_NAMES: dict[VISION_PROJECTOR_TYPE, str] = {
@@ -2211,6 +2214,40 @@ class MODEL_TENSOR(IntEnum):
         MODEL_TENSOR.FFN_DOWN,
         MODEL_TENSOR.FFN_UP,
     ],
+    MODEL_ARCH.FALCON_H1: [
+        # Token embedding
+        MODEL_TENSOR.TOKEN_EMBD,
+
+        # Input layernorm
+        MODEL_TENSOR.ATTN_NORM,
+
+        # Attention components
+        MODEL_TENSOR.ATTN_Q,         # Query projection
+        MODEL_TENSOR.ATTN_K,         # Key projection
+        MODEL_TENSOR.ATTN_V,         # Value projection
+        MODEL_TENSOR.ATTN_OUT,       # Output projection
+
+        # SSM components (Mamba2 specific)
+        MODEL_TENSOR.SSM_IN,         # Input projection for SSM
+        MODEL_TENSOR.SSM_CONV1D,     # Convolution layer
+        MODEL_TENSOR.SSM_DT,         # Delta time projection
+        MODEL_TENSOR.SSM_A,          # A parameter (log form)
+        MODEL_TENSOR.SSM_D,          # D parameter
+        MODEL_TENSOR.SSM_NORM,       # Normalization in SSM
+        MODEL_TENSOR.SSM_OUT,        # Output projection
+
+        # Pre-feedforward layernorm
+        MODEL_TENSOR.FFN_PRE_NORM,
+
+        # Feed-forward network components
+        MODEL_TENSOR.FFN_GATE,       # Gate projection (SwiGLU)
+        MODEL_TENSOR.FFN_DOWN,       # Down projection
+        MODEL_TENSOR.FFN_UP,         # Up projection
+
+        # Post-feedforward layernorm
+        MODEL_TENSOR.OUTPUT_NORM,    # Final layer norm
+        MODEL_TENSOR.OUTPUT,         # Output projection (lm_head)
+    ],
     # TODO
 }
 

gguf-py/gguf/gguf_writer.py

@@ -867,6 +867,9 @@ def add_ssm_group_count(self, value: int) -> None:
     def add_ssm_dt_b_c_rms(self, value: bool) -> None:
         self.add_bool(Keys.SSM.DT_B_C_RMS.format(arch=self.arch), value)
 
+    def add_ssm_head_dim(self, value: int) -> None:
+        self.add_uint32(Keys.SSM.HEAD_DIM.format(arch=self.arch), value)
+
     def add_tokenizer_model(self, model: str) -> None:
         self.add_string(Keys.Tokenizer.MODEL, model)
 

gguf-py/gguf/tensor_mapping.py

@@ -286,12 +286,14 @@ class TensorNameMap:
         # Post feed-forward norm
         MODEL_TENSOR.FFN_PRE_NORM: (
             "model.layers.{bid}.pre_feedforward_layernorm", # gemma2
+            "model.layers.{bid}.pre_ff_layernorm.weight",
         ),
 
         # Post feed-forward norm
         MODEL_TENSOR.FFN_POST_NORM: (
             "model.layers.{bid}.post_feedforward_layernorm", # gemma2 olmo2
             "model.layers.{bid}.post_mlp_layernorm", # glm-4-0414
+            "model.layers.{bid}.feed_forward.up_proj",  
         ),
 
         MODEL_TENSOR.FFN_GATE_INP: (
@@ -362,6 +364,7 @@ class TensorNameMap:
             "model.layers.{bid}.mlp.shared_expert.up_proj",          # qwen2moe
             "model.layers.{bid}.mlp.shared_experts.up_proj",         # deepseek deepseek2
             "model.layers.{bid}.feed_forward.shared_expert.up_proj", # llama4
+            "model.layers.{bid}.feed_forward.down_proj",    
         ),
 
         # AWQ-activation gate
@@ -547,11 +550,13 @@ class TensorNameMap:
         MODEL_TENSOR.SSM_IN: (
             "model.layers.{bid}.in_proj",
             "backbone.layers.{bid}.mixer.in_proj",
+            "model.layers.{bid}.mamba.in_proj",    
         ),
 
         MODEL_TENSOR.SSM_CONV1D: (
             "model.layers.{bid}.conv1d",
             "backbone.layers.{bid}.mixer.conv1d",
+            "model.layers.{bid}.mamba.conv1d",  
         ),
 
         MODEL_TENSOR.SSM_X: (
@@ -562,25 +567,30 @@ class TensorNameMap:
         MODEL_TENSOR.SSM_DT: (
             "model.layers.{bid}.dt_proj",
             "backbone.layers.{bid}.mixer.dt_proj",
+            "model.layers.{bid}.mamba.dt_proj",  
         ),
 
         MODEL_TENSOR.SSM_A: (
             "model.layers.{bid}.A_log",
             "backbone.layers.{bid}.mixer.A_log",
+            "model.layers.{bid}.mamba.A_log",   
         ),
 
         MODEL_TENSOR.SSM_D: (
             "model.layers.{bid}.D",
             "backbone.layers.{bid}.mixer.D",
+            "model.layers.{bid}.mamba.D",
         ),
 
         MODEL_TENSOR.SSM_NORM: (
+            "model.layers.{bid}.mamba.norm", # falcon-h1
             "backbone.layers.{bid}.mixer.norm",  # mamba2
         ),
 
         MODEL_TENSOR.SSM_OUT: (
             "model.layers.{bid}.out_proj",
             "backbone.layers.{bid}.mixer.out_proj",
+            "model.layers.{bid}.mamba.out_proj",          # falcon-h1
         ),
 
         MODEL_TENSOR.TIME_MIX_W0: (
@@ -1167,7 +1177,7 @@ class TensorNameMap:
         MODEL_TENSOR.V_RESMPL_ATTN_OUT: (
             "resampler.attn.out_proj",
         ),
-
+        
         MODEL_TENSOR.V_RESMPL_KV: (
             "resampler.kv_proj",
         ),