allow setting tokenChunkSize of WebGPU mode

backend-python/rwkv_pip/webgpu/model.py

@@ -26,12 +26,19 @@ class RWKV:
             if s.startswith("layer")
         )
 
+        chunk_size = (
+            int(s.lstrip("chunk"))
+            for s in strategy.split()
+            for s in s.split(",")
+            if s.startswith("chunk")
+        )
+
         args = {
             "file": model_path,
             "turbo": True,
             "quant": next(layer, 31) if "i8" in strategy else 0,
             "quant_nf4": next(layer, 26) if "i4" in strategy else 0,
-            "token_chunk_size": 128,
+            "token_chunk_size": next(chunk_size, 32),
             "lora": None,
         }
         self.model = self.wrp.Model(**args)