Training in progress, step 16500

adapter_config.json

@@ -25,12 +25,12 @@
   "revision": null,
   "target_modules": [
     "q_proj",
-    "o_proj",
-    "gate_proj",
-    "down_proj",
+    "up_proj",
     "v_proj",
+    "gate_proj",
+    "o_proj",
     "k_proj",
-    "up_proj"
+    "down_proj"
   ],
   "task_type": "CAUSAL_LM",
   "trainable_token_indices": null,

adapter_model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:f91e3c5f21cf129551bcf7c69bf4d8d16a4ef7f114f3eb09634873a47bacc359
+oid sha256:41c469e4040081e0ecfe377eeb4ac213060f49bafda35052223ed0946061838f
 size 161533160

trainer_log.jsonl

@@ -1,4 +1,2 @@
-{"current_steps": 16250, "total_steps": 18120, "loss": 0.2717, "lr": 3.2127666818151046e-06, "epoch": 2.6903973509933774, "percentage": 89.68, "elapsed_time": "0:05:28", "remaining_time": "0:00:37"}
-{"current_steps": 16300, "total_steps": 18120, "loss": 0.277, "lr": 3.045088789355488e-06, "epoch": 2.698675496688742, "percentage": 89.96, "elapsed_time": "0:11:01", "remaining_time": "0:01:13"}
-{"current_steps": 16350, "total_steps": 18120, "loss": 0.2652, "lr": 2.8817671791847634e-06, "epoch": 2.706953642384106, "percentage": 90.23, "elapsed_time": "0:16:10", "remaining_time": "0:01:45"}
-{"current_steps": 16400, "total_steps": 18120, "loss": 0.2617, "lr": 2.7228170036058153e-06, "epoch": 2.7152317880794703, "percentage": 90.51, "elapsed_time": "0:21:29", "remaining_time": "0:02:15"}
+{"current_steps": 16450, "total_steps": 18120, "loss": 0.2751, "lr": 2.5682530093579626e-06, "epoch": 2.7235099337748343, "percentage": 90.78, "elapsed_time": "0:05:13", "remaining_time": "0:00:31"}
+{"current_steps": 16500, "total_steps": 18120, "loss": 0.2741, "lr": 2.418089536248769e-06, "epoch": 2.7317880794701987, "percentage": 91.06, "elapsed_time": "0:10:04", "remaining_time": "0:00:59"}

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:4e7fd7045b4a15373c4d905f915edbb1436664635b44fe1df1bfa95de5188043
+oid sha256:e16cd32295896d2f57f1f19dbc49310a6ac6c29a2ec738cf4fb0c11d34db588b
 size 5752