Model save

README.md

@@ -38,10 +38,10 @@ The following hyperparameters were used during training:
 - eval_batch_size: 1
 - seed: 42
 - distributed_type: multi-GPU
-- num_devices: 2
-- gradient_accumulation_steps: 16
+- num_devices: 4
+- gradient_accumulation_steps: 8
 - total_train_batch_size: 32
-- total_eval_batch_size: 2
+- total_eval_batch_size: 4
 - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
 - lr_scheduler_type: cosine
 - lr_scheduler_warmup_ratio: 0.05

adapter_config.json

@@ -12,6 +12,8 @@
   "modules_to_save": [
     "connector",
     "mm_projector",
+    "response_head",
+    "related_head",
     "lm_head",
     "informative_head",
     "relevance_head",

adapter_model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:ea39f158db6fa103ede9e2a74c9c17f1e74f482be83742cbe1e5f6fb4031bdee
+oid sha256:5477f4dfacc81f6e31f056965a0c432039a9bbc8fac4b6c7dfb8c5422b215cbd
 size 1204780872

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:77f1651b3d95ece762f3396a93da978d57391d165cce4a217173de0366ea867f
-size 7608
+oid sha256:1852aba2ace7e3641ffaab3c86d119cebb4b6301c7f8879296ca7eb3878f113a
+size 7672