Merge branch 'add-fh1-rebased' of https://github.com/tiiuae/llama.cpp-public into add-fh1-rebased

gguf-py/gguf/constants.py

@@ -289,7 +289,7 @@ class MODEL_ARCH(IntEnum):
     LLAMA4           = auto()
     DECI             = auto()
     FALCON           = auto()
-    FALCON_H1           = auto()
+    FALCON_H1        = auto()
     BAICHUAN         = auto()
     GROK             = auto()
     GPT2             = auto()
@@ -662,7 +662,7 @@ MODEL_ARCH_NAMES: dict[MODEL_ARCH, str] = {
     MODEL_ARCH.DOTS1:            "dots1",
     MODEL_ARCH.ARCEE:            "arcee",
     MODEL_ARCH.ERNIE4_5:         "ernie4_5",
-    MODEL_ARCH.FALCON_H1:         "falcon_h1",
+    MODEL_ARCH.FALCON_H1:        "falcon_h1",
 }
 
 VISION_PROJECTOR_TYPE_NAMES: dict[VISION_PROJECTOR_TYPE, str] = {

gguf-py/gguf/tensor_mapping.py

@@ -583,6 +583,7 @@ class TensorNameMap:
         ),
 
         MODEL_TENSOR.SSM_NORM: (
+            "model.layers.{bid}.mamba.norm", # falcon-h1
             "backbone.layers.{bid}.mixer.norm",  # mamba2
         ),
 
@@ -1177,10 +1178,6 @@ class TensorNameMap:
             "resampler.attn.out_proj",
         ),
         
-        MODEL_TENSOR.SSM_NORM: (
-            "model.layers.{bid}.mamba.norm",
-        ),
-
         MODEL_TENSOR.V_RESMPL_KV: (
             "resampler.kv_proj",
         ),

src/llama-arch.cpp

@@ -1951,8 +1951,7 @@ bool llm_arch_is_recurrent(const llm_arch & arch) {
 }
 
 bool llm_arch_is_hybrid(const llm_arch & arch) {
-    // TODO: There are currently no hybrid models! Once there are, this will be
-    //  the place to identify them
+    // List all mamba-attention hybrid models here
     switch (arch) {
         case LLM_ARCH_FALCON_H1:
             return true;