Training in progress, step 18000

pytorch_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:d9fa2de829d24e99eb20c8dd7280e503f96b588ca8051fab51af5f6ba304a842
+oid sha256:0083259db3b58a2a002a48614b01db92f9a0d63a6d02a7aeff5ba6e221b37e9a
 size 1527847357

run.sh

@@ -10,13 +10,13 @@ deepspeed run-ba.py \
 	--text_column_name="sentence" \
 	--streaming="False" \
 	--ignore_data_skip \
-	--resume_from_checkpoint="checkpoint-14000" \
+	--resume_from_checkpoint="checkpoint-18000" \
 	--torch_compile="True" \
 	--torch_compile_mode="reduce-overhead" \
 	--torch_compile_mode="max-autotune" \
 	--logging_steps="25" \
 	--learning_rate="4e-4" \
-	--max_steps="20000" \
+	--max_steps="18000" \
 	--output_dir="./" \
 	--per_device_train_batch_size="32" \
 	--gradient_accumulation_steps="1" \

runs/Dec22_09-38-38_129-146-176-120/events.out.tfevents.1671701963.129-146-176-120.830862.0

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:23c5606d59b8f250f928216197bf876a80b90c197e5ec94b725c864e8b7fe909
-size 24193
+oid sha256:1ffcecc9660bb3d4c00e6ad3958a85ea665cc118c17ec5d5e6d0a9311cf60ec2
+size 30917