Model save

README.md

@@ -39,8 +39,8 @@ The following hyperparameters were used during training:
 - seed: 42
 - distributed_type: multi-GPU
 - num_devices: 2
-- gradient_accumulation_steps: 8
-- total_train_batch_size: 16
+- gradient_accumulation_steps: 16
+- total_train_batch_size: 32
 - total_eval_batch_size: 2
 - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
 - lr_scheduler_type: cosine

adapter_config.json

@@ -9,7 +9,16 @@
   "layers_to_transform": null,
   "lora_alpha": 32,
   "lora_dropout": 0.05,
-  "modules_to_save": null,
+  "modules_to_save": [
+    "connector",
+    "mm_projector",
+    "response_head",
+    "related_head",
+    "lm_head",
+    "informative_head",
+    "relevance_head",
+    "uncertainty_head"
+  ],
   "peft_type": "LORA",
   "r": 16,
   "revision": null,

adapter_model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:45586505cc6446579dfcc340c93a5a243f56e1d52f5eea2720702e3a443c28f9
-size 80792880
+oid sha256:1758145d5d7aed686432e0cb24dd5ac2c6d9559009e611ccf43e48d1a9f6ea36
+size 1204780872

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:717d8fddf9f7e6e836658c600c244485888a3feec38df308431fa6949df34b0f
-size 7608
+oid sha256:ebce320674f286e5ae592444650217ce3d75b35ec6e09f2c1a25cf8787d7f3d5
+size 7672