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: 16
-- total_train_batch_size: 32
+- gradient_accumulation_steps: 8
+- total_train_batch_size: 16
 - total_eval_batch_size: 2
 - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
 - lr_scheduler_type: cosine
@@ -54,7 +54,7 @@ The following hyperparameters were used during training:
 ### Framework versions
 
 - PEFT 0.4.0
-- Transformers 4.40.0.dev0
+- Transformers 4.40.0
 - Pytorch 2.5.1+cu124
 - Datasets 2.16.1
-- Tokenizers 0.15.2
+- Tokenizers 0.19.1

adapter_config.json

@@ -9,14 +9,7 @@
   "layers_to_transform": null,
   "lora_alpha": 32,
   "lora_dropout": 0.05,
-  "modules_to_save": [
-    "connector",
-    "mm_projector",
-    "lm_head",
-    "informative_head",
-    "relevance_head",
-    "uncertainty_head"
-  ],
+  "modules_to_save": null,
   "peft_type": "LORA",
   "r": 16,
   "revision": null,

adapter_model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:05869f2a4c856eaaac3380f22816e635116d1048eb686df5380a47578c684782
-size 1204780872
+oid sha256:45586505cc6446579dfcc340c93a5a243f56e1d52f5eea2720702e3a443c28f9
+size 80792880

tokenizer.json

@@ -82,6 +82,7 @@
     "end_of_word_suffix": "",
     "fuse_unk": false,
     "byte_fallback": false,
+    "ignore_merges": false,
     "vocab": {
       "!": 0,
       "\"": 1,

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:d8bea4b6855a51dc80f1ba22d86c24e3c6fa84e8ce043dc5c76b00cc099a20bc
-size 7544
+oid sha256:717d8fddf9f7e6e836658c600c244485888a3feec38df308431fa6949df34b0f
+size 7608