Training in progress, step 420

adapter_config.json

@@ -23,13 +23,13 @@
   "rank_pattern": {},
   "revision": null,
   "target_modules": [
-    "k_proj",
-    "q_proj",
-    "down_proj",
-    "gate_proj",
     "v_proj",
     "up_proj",
-    "o_proj"
+    "o_proj",
+    "down_proj",
+    "k_proj",
+    "gate_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 sha256:494699bc211a993c174bdef9bee2e290c1feccec832a028cf4fff513b3ba1840
+oid sha256:a8130e83528e67064027d9f72500e72b9c8f84725905e2ed7556b045263bd1cd
 size 80792456

trainer_log.jsonl

@@ -1,12 +1,6 @@
-{"current_steps": 250, "total_steps": 2550, "loss": 0.6036, "lr": 0.00029489696943632825, "epoch": 0.29394473838918284, "percentage": 9.8, "elapsed_time": "0:50:52", "remaining_time": "7:48:02"}
-{"current_steps": 260, "total_steps": 2550, "loss": 0.6024, "lr": 0.0002944000922130167, "epoch": 0.30570252792475017, "percentage": 10.2, "elapsed_time": "1:41:43", "remaining_time": "14:55:57"}
-{"current_steps": 270, "total_steps": 2550, "loss": 0.5936, "lr": 0.00029388059386971724, "epoch": 0.31746031746031744, "percentage": 10.59, "elapsed_time": "2:32:30", "remaining_time": "21:27:49"}
-{"current_steps": 280, "total_steps": 2550, "loss": 0.5987, "lr": 0.0002933385557888875, "epoch": 0.3292181069958848, "percentage": 10.98, "elapsed_time": "3:23:18", "remaining_time": "1 day, 3:28:11"}
-{"current_steps": 290, "total_steps": 2550, "loss": 0.5955, "lr": 0.00029277406288396663, "epoch": 0.3409758965314521, "percentage": 11.37, "elapsed_time": "4:14:05", "remaining_time": "1 day, 9:00:13"}
-{"current_steps": 300, "total_steps": 2550, "loss": 0.5973, "lr": 0.00029218720358607363, "epoch": 0.3527336860670194, "percentage": 11.76, "elapsed_time": "5:05:05", "remaining_time": "1 day, 14:08:10"}
-{"current_steps": 310, "total_steps": 2550, "loss": 0.5897, "lr": 0.00029157806983015394, "epoch": 0.3644914756025867, "percentage": 12.16, "elapsed_time": "5:55:57", "remaining_time": "1 day, 18:52:05"}
-{"current_steps": 320, "total_steps": 2550, "loss": 0.5936, "lr": 0.00029094675704057724, "epoch": 0.37624926513815404, "percentage": 12.55, "elapsed_time": "6:46:54", "remaining_time": "1 day, 23:15:39"}
-{"current_steps": 330, "total_steps": 2550, "loss": 0.5876, "lr": 0.00029029336411618865, "epoch": 0.3880070546737213, "percentage": 12.94, "elapsed_time": "7:37:50", "remaining_time": "2 days, 3:19:59"}
-{"current_steps": 340, "total_steps": 2550, "loss": 0.5884, "lr": 0.0002896179934148158, "epoch": 0.39976484420928865, "percentage": 13.33, "elapsed_time": "8:28:43", "remaining_time": "2 days, 7:06:42"}
-{"current_steps": 350, "total_steps": 2550, "loss": 0.5846, "lr": 0.0002889207507372337, "epoch": 0.411522633744856, "percentage": 13.73, "elapsed_time": "9:19:33", "remaining_time": "2 days, 10:37:15"}
-{"current_steps": 360, "total_steps": 2550, "loss": 0.5862, "lr": 0.0002882017453105906, "epoch": 0.42328042328042326, "percentage": 14.12, "elapsed_time": "10:10:28", "remaining_time": "2 days, 13:53:43"}
+{"current_steps": 370, "total_steps": 2550, "loss": 0.5724, "lr": 0.0002865363126582549, "epoch": 0.4350382128159906, "percentage": 14.51, "elapsed_time": "0:51:02", "remaining_time": "5:00:45"}
+{"current_steps": 380, "total_steps": 2550, "loss": 0.589, "lr": 0.00028575266221296395, "epoch": 0.4467960023515579, "percentage": 14.9, "elapsed_time": "1:42:04", "remaining_time": "9:42:51"}
+{"current_steps": 390, "total_steps": 2550, "loss": 0.5887, "lr": 0.00028494798058030713, "epoch": 0.4585537918871252, "percentage": 15.29, "elapsed_time": "2:33:10", "remaining_time": "14:08:22"}
+{"current_steps": 400, "total_steps": 2550, "loss": 0.5764, "lr": 0.0002841223924238447, "epoch": 0.4703115814226925, "percentage": 15.69, "elapsed_time": "3:24:19", "remaining_time": "18:18:12"}
+{"current_steps": 410, "total_steps": 2550, "loss": 0.5808, "lr": 0.0002832760256460349, "epoch": 0.48206937095825986, "percentage": 16.08, "elapsed_time": "4:15:30", "remaining_time": "22:13:39"}
+{"current_steps": 420, "total_steps": 2550, "loss": 0.5733, "lr": 0.00028240901136841886, "epoch": 0.49382716049382713, "percentage": 16.47, "elapsed_time": "5:07:05", "remaining_time": "1 day, 1:57:23"}

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:ecad592cf6ef86b9231e395018e571efe68a92520f4ec6e1705ee19d3c997e50
+oid sha256:12eac2dba3df35b6c9379924bc2a67b96e3d059dafdc119c9ec63ba9d7c614e2
 size 5752