Merge branch 'upstream' into concedo_experimental

# Conflicts:
#	.github/workflows/update-ops-docs.yml
#	ggml/src/CMakeLists.txt
#	ggml/src/ggml-cpu/CMakeLists.txt

Author: LostRuins

embd_res/klite.embd

@@ -5481,6 +5481,34 @@ Current version indicated by LITEVER below.
 			&& String.fromCharCode(x.data[0]) == 'c')
 			&& String.fromCharCode(x.data[4]) == 'a');
 
+		if(!found || found.length==0)
+		{
+			let found2 = chunks.filter(x => (
+			x.name == "tEXt"
+			&& x.data.length > 6
+			&& String.fromCharCode(x.data[0]) == 'c')
+			&& String.fromCharCode(x.data[1]) == 'c'
+			&& String.fromCharCode(x.data[2]) == 'v'
+			&& String.fromCharCode(x.data[3]) == '3'); //its a stupid new v3 card
+
+			if(found2 && found2.length>0)
+			{
+				try {
+					let b64buf = "";
+					let bytes = found2[0].data; //skip the chara
+					for (var i = 5; i < bytes.length; i++) {
+						b64buf += String.fromCharCode(bytes[i]);
+					}
+					var decoded = JSON.parse(b64_decode_unicode(b64buf));
+					console.log(decoded);
+					return decoded;
+				} catch (e) {
+					console.log("Error decoding b64 in image: " + e);
+					return null;
+				}
+			}
+		}
+
 		//remove ext asset
 		found = found.filter(x => (
 			x.data.length > 12

src/llama-model.cpp

@@ -119,6 +119,7 @@ const char * llm_type_name(llm_type type) {
         case LLM_TYPE_17B_16E:       return "17Bx16E (Scout)";
         case LLM_TYPE_17B_128E:      return "17Bx128E (Maverick)";
         case LLM_TYPE_A13B:          return "A13B";
+        case LLM_TYPE_7B_A1B:        return "7B.A1B";
         case LLM_TYPE_8B_A1B:        return "8B.A1B";
         case LLM_TYPE_21B_A3B:       return "21B.A3B";
         case LLM_TYPE_30B_A3B:       return "30B.A3B";
@@ -1848,8 +1849,10 @@ void llama_model::load_hparams(llama_model_loader & ml) {
 
                 ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps);
 
-                switch (hparams.n_layer) {
-                    // TODO: Add llm type label (not sure this is useful)
+                switch (hparams.n_embd) {
+                    case 1536: type = LLM_TYPE_7B_A1B; break;
+                    case 2048: case 2560: type = LLM_TYPE_3B; break;
+                    case 4096: type = LLM_TYPE_32B; break;
                     default: type = LLM_TYPE_UNKNOWN;
                 }
 

src/llama-model.h

@@ -107,6 +107,7 @@ enum llm_type {
     LLM_TYPE_17B_16E, // llama4 Scout
     LLM_TYPE_17B_128E, // llama4 Maverick
     LLM_TYPE_A13B,
+    LLM_TYPE_7B_A1B,
     LLM_TYPE_8B_A1B, // lfm2moe
     LLM_TYPE_21B_A3B, // Ernie MoE small
     LLM_TYPE_30B_A3B,