Model save

README.md

@@ -34,12 +34,12 @@ More information needed
 
 The following hyperparameters were used during training:
 - learning_rate: 2e-05
-- train_batch_size: 8
+- train_batch_size: 1
 - eval_batch_size: 16
 - seed: 42
 - distributed_type: multi-GPU
 - gradient_accumulation_steps: 4
-- total_train_batch_size: 32
+- total_train_batch_size: 4
 - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
 - lr_scheduler_type: cosine
 - num_epochs: 1

adapter_config.json

@@ -16,10 +16,10 @@
   "rank_pattern": {},
   "revision": null,
   "target_modules": [
-    "k_proj",
+    "v_proj",
     "q_proj",
     "o_proj",
-    "v_proj"
+    "k_proj"
   ],
   "task_type": "CAUSAL_LM"
 }

runs/Feb03_23-57-37_r2d2-devbox-fqvnh/events.out.tfevents.1738627072.r2d2-devbox-fqvnh.15348.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a6a3e9bb096613460d8e76a6a99ef4a9d745fb8042c04ff53b19d01d3a6b323e
+size 4702

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:a91780eb02618f8e2fed0d61580aacdf469917027f977a79a01463d4d5977c22
+oid sha256:d5fa82360df7b0a39892c65caa6a581965e3261b781d79a03239ee6191468da2
 size 5880