Add chat template block (runtime-owned chat templating)

v81/phimoe.json

@@ -1,28 +1,76 @@
-{
-  "schema_version": 1,
-  "_comment": "phi-tiny-moe - Qualcomm Hexagon v81 (QHexRT) deployment manifest.",
-  "model": { "name": "phi-tiny-moe", "family": "llm", "dsp_arch": "v81", "tokenizer_pre": "default" },
-  "params": {
-    "hidden": 4096, "vocab": 32064, "n_layers": 32, "max_ctx": 2048,
-    "kv_dim": 512, "head_dim": 128, "rope_theta": 10000, "eos_token_id": 32000
-  },
-  "artifacts": {
-    "contexts": {
-      "a_lo":   { "bin": "a_lo.bin" },
-      "a_hi":   { "bin": "a_hi.bin" },
-      "ffn":    { "bin": "ffn2.bin" },
-      "lmhead": { "bin": "lmhead_ar1.bin" }
-    },
-    "embed": "embed_f16.bin",
-    "tokenizer": "tokenizer.json",
-    "fixture_dir": "."
-  },
-  "plan": {
-    "steps": [
-      { "host": "phimoe_generate",
-        "params": { "experts": "experts_i8.bin", "scales": "experts_scale.f32",
-                    "n_experts": 16, "ffn": 448, "jitter": 0.01, "neg": -50,
-                    "user_id": 32010, "end_id": 32007, "assistant_id": 32001, "max_new": 24 } }
-    ]
-  }
-}
+{
+  "schema_version": 1,
+  "_comment": "phi-tiny-moe - Qualcomm Hexagon v81 (QHexRT) deployment manifest.",
+  "model": {
+    "name": "phi-tiny-moe",
+    "family": "llm",
+    "dsp_arch": "v81",
+    "tokenizer_pre": "default"
+  },
+  "chat": {
+    "user": {
+      "prefix": [
+        32010
+      ],
+      "suffix": [
+        32007
+      ]
+    },
+    "assistant": {
+      "prefix": [
+        32001
+      ],
+      "suffix": [
+        32007
+      ]
+    }
+  },
+  "params": {
+    "hidden": 4096,
+    "vocab": 32064,
+    "n_layers": 32,
+    "max_ctx": 2048,
+    "kv_dim": 512,
+    "head_dim": 128,
+    "rope_theta": 10000,
+    "eos_token_id": 32000
+  },
+  "artifacts": {
+    "contexts": {
+      "a_lo": {
+        "bin": "a_lo.bin"
+      },
+      "a_hi": {
+        "bin": "a_hi.bin"
+      },
+      "ffn": {
+        "bin": "ffn2.bin"
+      },
+      "lmhead": {
+        "bin": "lmhead_ar1.bin"
+      }
+    },
+    "embed": "embed_f16.bin",
+    "tokenizer": "tokenizer.json",
+    "fixture_dir": "."
+  },
+  "plan": {
+    "steps": [
+      {
+        "host": "phimoe_generate",
+        "params": {
+          "experts": "experts_i8.bin",
+          "scales": "experts_scale.f32",
+          "n_experts": 16,
+          "ffn": 448,
+          "jitter": 0.01,
+          "neg": -50,
+          "user_id": 32010,
+          "end_id": 32007,
+          "assistant_id": 32001,
+          "max_new": 24
+        }
+      }
+    ]
+  }
+}