Update tiny models for Qwen3_5MoeForCausalLM

config.json

@@ -4,14 +4,13 @@
   ],
   "attention_bias": false,
   "attention_dropout": 0.0,
-  "bos_token_id": 0,
+  "bos_token_id": 1,
   "dtype": "float32",
   "eos_token_id": 248046,
-  "head_dim": 8,
+  "head_dim": 16,
   "hidden_act": "silu",
   "hidden_size": 32,
   "initializer_range": 0.02,
-  "intermediate_size": 37,
   "is_decoder": true,
   "layer_types": [
     "full_attention",
@@ -25,7 +24,7 @@
   "max_position_embeddings": 512,
   "model_type": "qwen3_5_moe_text",
   "moe_intermediate_size": 16,
-  "num_attention_heads": 4,
+  "num_attention_heads": 2,
   "num_experts": 8,
   "num_experts_per_tok": 2,
   "num_hidden_layers": 2,
@@ -35,19 +34,13 @@
   "partial_rotary_factor": 0.25,
   "rms_norm_eps": 1e-06,
   "rope_parameters": {
-    "mrope_interleaved": true,
-    "mrope_section": [
-      16,
-      8,
-      8
-    ],
     "partial_rotary_factor": 0.25,
-    "rope_theta": 10000,
+    "rope_theta": 10000.0,
     "rope_type": "default"
   },
   "router_aux_loss_coef": 0.001,
   "shared_expert_intermediate_size": 36,
-  "tie_word_embeddings": true,
+  "tie_word_embeddings": false,
   "transformers_version": "5.16.0.dev0",
   "use_cache": true,
   "vocab_size": 248077

generation_config.json

@@ -1,6 +1,6 @@
 {
   "_from_model_config": true,
-  "bos_token_id": 0,
+  "bos_token_id": 1,
   "eos_token_id": 248046,
   "output_attentions": false,
   "output_hidden_states": false,

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:f9f3d753fae929283f8638167316db74431f4b8c553617b362e9536b1ea2f6ad
-size 31977856
+oid sha256:abda792afc18b46645a8e1b6d60d017eefa419eab965dc3af1c15d21948183cb
+size 63735960

tokenizer_config.json

@@ -11,7 +11,7 @@
   "image_token": "<|image_pad|>",
   "is_local": true,
   "local_files_only": false,
-  "model_max_length": 262144,
+  "model_max_length": 512,
   "model_specific_special_tokens": {
     "audio_bos_token": "<|audio_start|>",
     "audio_eos_token": "<|audio_end|>",