v13: (README written by author)

README.md

@@ -33,10 +33,10 @@ while intermediate reasoning (Chain-of-Thought) is masked.
 
 - Base model: Qwen/Qwen3-4B-Instruct-2507
 - Method: QLoRA (4-bit)
-- Max sequence length: 512
+- Max sequence length: 1024
 - Epochs: 2
-- Learning rate: 2e-05
-- LoRA: r=64, alpha=128
+- Learning rate: 2e-06
+- LoRA: r=96, alpha=192
 
 ## Usage
 

adapter_config.json

@@ -20,7 +20,7 @@
   "layers_pattern": null,
   "layers_to_transform": null,
   "loftq_config": {},
-  "lora_alpha": 128,
+  "lora_alpha": 192,
   "lora_bias": false,
   "lora_dropout": 0.0,
   "megatron_config": null,
@@ -29,17 +29,17 @@
   "peft_type": "LORA",
   "peft_version": "0.18.1",
   "qalora_group_size": 16,
-  "r": 64,
+  "r": 96,
   "rank_pattern": {},
   "revision": null,
   "target_modules": [
+    "v_proj",
+    "up_proj",
+    "q_proj",
     "gate_proj",
-    "o_proj",
     "down_proj",
-    "k_proj",
-    "q_proj",
-    "v_proj",
-    "up_proj"
+    "o_proj",
+    "k_proj"
   ],
   "target_parameters": null,
   "task_type": "CAUSAL_LM",

adapter_model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:8fd7cb38d1015489db9769d2307431d6f54e1b28db185b52f6749f4aa6a73e60
-size 528550256
+oid sha256:2631fc1ec69ab1d7bbf7a98a098791dac89a0f382fc2ae1d99f43f4598b90158
+size 792791480