Support for dots.llm1 models (#573)

* Add llama.cpp changes for dots1 support

* Add python changes for dots1 support

* Fix to make it convert

* Remove V reshaping, remove BOS by default for dots1 and fix warmup to handle models without BOS

* Minor fix

* Remove commented lines

1 files changed, 26 insertions, 0 deletions

diff --git a/gguf-py/gguf/constants.py b/gguf-py/gguf/constants.py
index 489714c4..b3b2bc50 100644
--- a/gguf-py/gguf/constants.py
+++ b/gguf-py/gguf/constants.py
@@ -226,6 +226,7 @@ class MODEL_ARCH(IntEnum):
     T5           = auto()
     T5ENCODER    = auto()
     JAIS         = auto()
+    DOTS1        = auto()
 
 
 class MODEL_TENSOR(IntEnum):
@@ -362,6 +363,7 @@ MODEL_ARCH_NAMES: dict[MODEL_ARCH, str] = {
     MODEL_ARCH.T5:             "t5",
     MODEL_ARCH.T5ENCODER:      "t5encoder",
     MODEL_ARCH.JAIS:           "jais",
+    MODEL_ARCH.DOTS1:          "dots1",
 }
 
 TENSOR_NAMES: dict[MODEL_TENSOR, str] = {
@@ -1164,6 +1166,30 @@ MODEL_TENSORS: dict[MODEL_ARCH, list[MODEL_TENSOR]] = {
         MODEL_TENSOR.FFN_GATE,
         MODEL_TENSOR.FFN_UP,
     ],
+    MODEL_ARCH.DOTS1: [
+        MODEL_TENSOR.TOKEN_EMBD,
+        MODEL_TENSOR.OUTPUT_NORM,
+        MODEL_TENSOR.OUTPUT,
+        MODEL_TENSOR.ATTN_NORM,
+        MODEL_TENSOR.ATTN_Q,
+        MODEL_TENSOR.ATTN_Q_NORM,
+        MODEL_TENSOR.ATTN_K,
+        MODEL_TENSOR.ATTN_K_NORM,
+        MODEL_TENSOR.ATTN_V,
+        MODEL_TENSOR.ATTN_OUT,
+        MODEL_TENSOR.FFN_EXP_PROBS_B,
+        MODEL_TENSOR.FFN_NORM,
+        MODEL_TENSOR.FFN_GATE,
+        MODEL_TENSOR.FFN_GATE_EXP,
+        MODEL_TENSOR.FFN_GATE_INP,
+        MODEL_TENSOR.FFN_GATE_SHEXP,
+        MODEL_TENSOR.FFN_DOWN,
+        MODEL_TENSOR.FFN_DOWN_EXP,
+        MODEL_TENSOR.FFN_DOWN_SHEXP,
+        MODEL_TENSOR.FFN_UP,
+        MODEL_TENSOR.FFN_UP_EXP,
+        MODEL_TENSOR.FFN_UP_SHEXP,
+    ],
     # TODO
 }