epoch 22427 | block 8073904 | steps 175 | loss 1.0663 | sim_abs 0.568 sq_acc 0.10

config.json

@@ -1,15 +1,17 @@
 {
   "architectures": [
-    "Qwen2ForCausalLM"
+    "Qwen3ForCausalLM"
   ],
+  "attention_bias": false,
   "attention_dropout": 0.0,
   "bos_token_id": 151643,
   "dtype": "bfloat16",
   "eos_token_id": 151645,
+  "head_dim": 128,
   "hidden_act": "silu",
-  "hidden_size": 1536,
+  "hidden_size": 2048,
   "initializer_range": 0.02,
-  "intermediate_size": 8960,
+  "intermediate_size": 6144,
   "layer_types": [
     "full_attention",
     "full_attention",
@@ -40,16 +42,16 @@
     "full_attention",
     "full_attention"
   ],
-  "max_position_embeddings": 32768,
-  "max_window_layers": 21,
-  "model_type": "qwen2",
-  "num_attention_heads": 12,
+  "max_position_embeddings": 40960,
+  "max_window_layers": 28,
+  "model_type": "qwen3",
+  "num_attention_heads": 16,
   "num_hidden_layers": 28,
-  "num_key_value_heads": 2,
+  "num_key_value_heads": 8,
   "pad_token_id": null,
   "rms_norm_eps": 1e-06,
   "rope_parameters": {
-    "rope_theta": 1000000.0,
+    "rope_theta": 1000000,
     "rope_type": "default"
   },
   "sliding_window": null,

generation_config.json

@@ -6,9 +6,8 @@
     151643
   ],
   "pad_token_id": 151643,
-  "repetition_penalty": 1.1,
-  "temperature": 0.7,
+  "temperature": 0.6,
   "top_k": 20,
-  "top_p": 0.8,
+  "top_p": 0.95,
   "transformers_version": "5.6.2"
 }

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:47d41fed457727a87ad02720acad950f3039d142f5cf0f7fb90f2ec21f8df658
-size 3087467144
+oid sha256:0f27d007440d494ae1fe78920ce736de8f0e1738d554e66d61a92cf2e835df5c
+size 3441185608