model : add LFM2-ColBert-350M (#18607)

* model : add LFM2-ColBert-350M

* llama_model_n_embd_out() - returns `hparams.n_embd_out` if set and fallbacks to `hparams.n_embd`

gguf-py/gguf/constants.py

@@ -104,6 +104,7 @@ class Keys:
         VOCAB_SIZE                        = "{arch}.vocab_size"
         CONTEXT_LENGTH                    = "{arch}.context_length"
         EMBEDDING_LENGTH                  = "{arch}.embedding_length"
+        EMBEDDING_LENGTH_OUT              = "{arch}.embedding_length_out"
         FEATURES_LENGTH                   = "{arch}.features_length"
         BLOCK_COUNT                       = "{arch}.block_count"
         LEADING_DENSE_BLOCK_COUNT         = "{arch}.leading_dense_block_count"
@@ -3038,6 +3039,7 @@ MODEL_TENSORS: dict[MODEL_ARCH, list[MODEL_TENSOR]] = {
         MODEL_TENSOR.ATTN_V,
         MODEL_TENSOR.ATTN_OUT,
         MODEL_TENSOR.OUTPUT,
+        MODEL_TENSOR.DENSE_2_OUT, # LFM2-ColBert-350M
     ],
     MODEL_ARCH.LFM2MOE: [
         MODEL_TENSOR.TOKEN_EMBD,

gguf-py/gguf/gguf_writer.py

@@ -681,6 +681,9 @@ class GGUFWriter:
     def add_embedding_length(self, length: int) -> None:
         self.add_uint32(Keys.LLM.EMBEDDING_LENGTH.format(arch=self.arch), length)
 
+    def add_embedding_length_out(self, length: int) -> None:
+        self.add_uint32(Keys.LLM.EMBEDDING_LENGTH_OUT.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)