fix: Allow "output" layer in granite moe architecture (convert and cpp)

Branch: GraniteMoE Co-Authored-By: [email protected] Signed-off-by: Gabe Goodhart <[email protected]>
ggerganov · Sep 23, 2024 · e071bc8 · e071bc8
1 parent 3907795
commit e071bc8
Show file tree

Hide file tree

Showing 2 changed files with 2 additions and 0 deletions.
diff --git a/gguf-py/gguf/constants.py b/gguf-py/gguf/constants.py
@@ -1247,6 +1247,7 @@ class MODEL_TENSOR(IntEnum):
     MODEL_ARCH.GRANITE_MOE: [
         MODEL_TENSOR.TOKEN_EMBD,
         MODEL_TENSOR.OUTPUT_NORM,
+        MODEL_TENSOR.OUTPUT,
         MODEL_TENSOR.ATTN_NORM,
         MODEL_TENSOR.ATTN_Q,
         MODEL_TENSOR.ATTN_K,

diff --git a/src/llama.cpp b/src/llama.cpp
@@ -1485,6 +1485,7 @@ static const std::map<llm_arch, std::map<llm_tensor, std::string>> LLM_TENSOR_NA
         {
             { 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_Q,          "blk.%d.attn_q" },
             { LLM_TENSOR_ATTN_K,          "blk.%d.attn_k" },