Best model so far - eval_loss: 1.7660 --------------------------------------------------------

adapter_config.json

@@ -1,7 +1,7 @@
 {
   "alpha_pattern": {},
   "auto_mapping": null,
-  "base_model_name_or_path": "unsloth/gemma-1.1-2b-it",
+  "base_model_name_or_path": "facebook/opt-125m",
   "bias": "none",
   "fan_in_fan_out": false,
   "inference_mode": true,
@@ -10,25 +10,23 @@
   "layers_pattern": null,
   "layers_to_transform": null,
   "loftq_config": {},
-  "lora_alpha": 128,
+  "lora_alpha": 96,
   "lora_dropout": 0.1,
   "megatron_config": null,
   "megatron_core": "megatron.core",
-  "modules_to_save": [
-    "lm_head"
-  ],
+  "modules_to_save": [],
   "peft_type": "LORA",
-  "r": 64,
+  "r": 48,
   "rank_pattern": {},
   "revision": null,
   "target_modules": [
-    "up_proj",
+    "o_proj",
     "v_proj",
+    "k_proj",
     "down_proj",
-    "o_proj",
-    "q_proj",
+    "up_proj",
     "gate_proj",
-    "k_proj"
+    "q_proj"
   ],
   "task_type": "CAUSAL_LM",
   "use_dora": false,

adapter_model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:4fcff88ca2c3ae6fcae25cac2e5b1dffa4cc67a0d587851e3f11fcb3b693ed9b
-size 1362396360
+oid sha256:c76b92b53cdd34b03eda45a0675079080ceea65b6a10e6ca81863dce77d7163f
+size 10626888