Smoldocling support

gguf-py/gguf/tensor_mapping.py

@@ -252,8 +252,8 @@ class TensorNameMap:
         ),
 
         MODEL_TENSOR.ATTN_POST_NORM: (
-            "model.layers.{bid}.post_attention_layernorm",     # gemma2 olmo2    # ge
-            "model.layers.{bid}.post_self_attn_layernorm",     # glm-4-0414
+            "model.layers.{bid}.post_attention_layernorm",                   # gemma2 olmo2    # ge
+            "model.layers.{bid}.post_self_attn_layernorm",                   # glm-4-0414
         ),
 
         # Rotary embeddings
@@ -378,18 +378,18 @@ class TensorNameMap:
 
         # Feed-forward gate
         MODEL_TENSOR.FFN_GATE: (
-            "model.layers.{bid}.mlp.gate_proj",           # llama-hf refact olmo2
-            "layers.{bid}.feed_forward.w1",               # llama-pth
-            "transformer.h.{bid}.mlp.w2",                 # qwen
-            "transformer.h.{bid}.mlp.c_fc2",              # jais
-            "model.layers.layers.{bid}.mlp.gate_proj",    # plamo
-            "model.layers.{bid}.feed_forward.w1",         # internlm2
-            "encoder.layers.{bid}.mlp.fc12",              # nomic-bert
-            "encoder.layer.{bid}.mlp.gated_layers_w",     # jina-bert-v2 (split up/gate, no longer used)
-            "transformer.h.{bid}.mlp.linear_1",           # refact
-            "model.layers.{bid}.residual_mlp.w1",         # arctic
-            "transformer.h.{bid}.mlp.c_fc_0",             # exaone
-            "model.layers.{bid}.feed_forward.gate_proj",  # llama4 jamba
+            "model.layers.{bid}.mlp.gate_proj",                     # llama-hf refact olmo2
+            "layers.{bid}.feed_forward.w1",                         # llama-pth
+            "transformer.h.{bid}.mlp.w2",                           # qwen
+            "transformer.h.{bid}.mlp.c_fc2",                        # jais
+            "model.layers.layers.{bid}.mlp.gate_proj",              # plamo
+            "model.layers.{bid}.feed_forward.w1",                   # internlm2
+            "encoder.layers.{bid}.mlp.fc12",                        # nomic-bert
+            "encoder.layer.{bid}.mlp.gated_layers_w",               # jina-bert-v2 (split up/gate, no longer used)
+            "transformer.h.{bid}.mlp.linear_1",                     # refact
+            "model.layers.{bid}.residual_mlp.w1",                   # arctic
+            "transformer.h.{bid}.mlp.c_fc_0",                       # exaone
+            "model.layers.{bid}.feed_forward.gate_proj",            # llama4 jamba
         ),
 
         MODEL_TENSOR.FFN_GATE_EXP: (
@@ -1024,7 +1024,7 @@ class TensorNameMap:
         ),
 
         MODEL_TENSOR.V_MMPROJ_FC: (
-            "model.connector.modality_projection.proj", # SmolVLM
+            "model.connector.modality_projection.proj",        # SmolVLM
         ),
 
         MODEL_TENSOR.V_MMPROJ_MLP: (

src/llama-model.cpp

@@ -40,6 +40,7 @@ const char * llm_type_name(llm_type type) {
         case LLM_TYPE_190M:          return "190M";
         case LLM_TYPE_220M:          return "220M";
         case LLM_TYPE_250M:          return "250M";
+        case LLM_TYPE_256M:          return "256M";
         case LLM_TYPE_270M:          return "270M";
         case LLM_TYPE_335M:          return "335M";
         case LLM_TYPE_410M:          return "410M";
@@ -581,6 +582,7 @@ void llama_model::load_hparams(llama_model_loader & ml) {
                         case 22: type = LLM_TYPE_1B; break;
                         case 26: type = LLM_TYPE_3B; break;
                         case 28: type = LLM_TYPE_3B; break; // Llama 3.2 3B
+                        case 30: type = LLM_TYPE_256M; break; // smoldocling 256M
                         // granite uses a vocab with len 49152
                         case 32: type = n_vocab == 49152 ? LLM_TYPE_3B : (n_vocab < 40000 ? LLM_TYPE_7B : LLM_TYPE_8B); break;
                         case 36: type = LLM_TYPE_8B; break; // granite

src/llama-model.h

@@ -32,6 +32,7 @@ enum llm_type {
     LLM_TYPE_190M,
     LLM_TYPE_220M,
     LLM_TYPE_250M,
+    LLM_TYPE_256M,
     LLM_TYPE_270M,
     LLM_TYPE_335M,
     LLM_TYPE_410M,

src/llama-vocab.cpp

@@ -1846,6 +1846,7 @@ void llama_vocab::impl::load(llama_model_loader & ml, const LLM_KV & kv) {
                         || t.first == "<EOT>"
                         || t.first == "_<EOT>"
                         || t.first == "<｜end▁of▁sentence｜>" // DeepSeek
+                        || t.first == "<end_of_utterance>" // smoldocling
                    ) {
                     special_eot_id = t.second;
                     if ((id_to_token[t.second].attr & LLAMA_TOKEN_ATTR_CONTROL) == 0) {
@@ -2005,6 +2006,7 @@ void llama_vocab::impl::load(llama_model_loader & ml, const LLM_KV & kv) {
                     || t.first == "<EOT>"
                     || t.first == "_<EOT>"
                     || t.first == "<|end_of_text|>"
+                    || t.first == "<end_of_utterance>" // smoldocling
                ) {
                 special_eog_ids.insert(t.second);
                 if ((id_to_token[t.second].attr & LLAMA_TOKEN_ATTR_CONTROL) == 0) {