step 30000

config.json

@@ -79,6 +79,7 @@
     "vocab_size": 151936
   },
   "base_model_name_or_path": "Qwen/Qwen3-1.7B",
+  "convergence_epsilon": 0.01,
   "decoder_layer_indices": [
     27
   ],
@@ -86,8 +87,10 @@
     0,
     1
   ],
+  "halting_strategy": "threshold",
   "hidden_size": 2048,
   "model_type": "lds",
+  "q_eval_interval": 1,
   "q_threshold": 0.9,
   "reasoning_layer_indices": [
     2,

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:fb85a20a4217338ae92b00f5c3ae944d851d01b4245c979783872e7cbf8a3468
-size 4064586530
+oid sha256:c9941e40dc7bc4b53b026ac31948d48b3ba5bcce90b32f895423a15d09cd010d
+size 4065655516

tokenizer_config.json

@@ -20,7 +20,7 @@
     "<|image_pad|>",
     "<|video_pad|>"
   ],
-  "is_local": true,
+  "is_local": false,
   "model_max_length": 131072,
   "pad_token": "<|endoftext|>",
   "split_special_tokens": false,