Training in progress, step 5

config.json

@@ -1,12 +1,12 @@
 {
-  "_name_or_path": "LGAI-EXAONE/EXAONE-3.5-2.4B-Instruct",
+  "_name_or_path": "/home/edgeai/DeepSpeedExamples/training/autotuning/hf/gpt2-large/saved_based_exaone_block1024_Batch1_Rank28_last_layer_save/",
   "activation_function": "silu",
   "architectures": [
-    "ExaoneForCausalLM"
+    "ExaoneTDForCausalLM"
   ],
   "attention_dropout": 0.0,
   "auto_map": {
-    "AutoConfig": "LGAI-EXAONE/EXAONE-3.5-2.4B-Instruct--configuration_exaone.ExaoneConfig",
+    "AutoConfig": "configuration_exaone.ExaoneConfig",
     "AutoModelForCausalLM": "LGAI-EXAONE/EXAONE-3.5-2.4B-Instruct--modeling_exaone.ExaoneForCausalLM",
     "AutoModelForSequenceClassification": "LGAI-EXAONE/EXAONE-3.5-2.4B-Instruct--modeling_exaone.ExaoneForSequenceClassification"
   },

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:47b56cb24533b87f77d79f0b39f5adaae1f815d938355323f21756128d857c52
-size 4810685896
+oid sha256:f53877b20fc544d4ec0d5a09ec23349d6f195b9358d77ef06509e66544e87340
+size 634692008

runs/Mar04_15-52-58_bizon3/events.out.tfevents.1741132410.bizon3.1838119.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1987fda8d5cbaaa38c42b7c6b6c880904adab46117b53c4494e349fcd5235d4e
+size 11835

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:386a13d6bfc8acfa6af24fd1ce3aba7a3093f9483e085bd348973913bae5784e
-size 7352
+oid sha256:a2527c1ee0e66a904e5ddce5d19da8d9d4794a90739f95e50ea8f7412e23c33f
+size 7480