update configs

kaistAI · Apr 10, 2024 · 5df27f3 · 5df27f3
1 parent e03f182
commit 5df27f3
Show file tree

Hide file tree

Showing 3 changed files with 11 additions and 17 deletions.
diff --git a/configs/official/OLMo-1B-105.yaml b/configs/official/OLMo-1B-105.yaml
@@ -1,4 +1,4 @@
-run_name: OLMo-1B
+run_name: OLMo-1B-sanity-check
 seed: 6198
 dry_run: false
 
@@ -58,13 +58,13 @@ tokenizer:
 save_folder: ${path.choose:${oc.env:SCRATCH_DIR,no_exist}/checkpoints,/results}/${oc.env:SLURM_JOB_ID,${run_name}}
 save_overwrite: false
 # Sharded checkpoints (best for restarts)
-save_interval: 1000
-save_num_checkpoints_to_keep: 9
+save_interval: 100
+save_num_checkpoints_to_keep: 11
 # Unsharded checkpoints (for final storage)
-save_interval_unsharded: 10000
+save_interval_unsharded: 1000
 save_num_unsharded_checkpoints_to_keep: -1
 
-load_path: null
+load_path: /mnt/nas/hoyeon/OLMo/official_checkpoints/step50000-unsharded
 
 max_duration: 739_328  # 3.1T tokens
 global_train_batch_size: 2048
@@ -85,9 +85,7 @@ speed_monitor:
 eval_interval: 999999999999
 eval_subset_num_batches: -1
 device_eval_batch_size: ${device_train_microbatch_size}
-evaluators:
-  - label: sciq
-    type: downstream
+evaluators: []
 
 data:
   pad_direction: right

diff --git a/configs/official/OLMo-1B.yaml b/configs/official/OLMo-1B.yaml
@@ -1,4 +1,4 @@
-run_name: OLMo-1B
+run_name: OLMo-1B-sanity-check
 seed: 6198
 dry_run: false
 
@@ -58,13 +58,13 @@ tokenizer:
 save_folder: ${path.choose:${oc.env:SCRATCH_DIR,no_exist}/checkpoints,/results}/${oc.env:SLURM_JOB_ID,${run_name}}
 save_overwrite: false
 # Sharded checkpoints (best for restarts)
-save_interval: 1000
+save_interval: 100
 save_num_checkpoints_to_keep: 9
 # Unsharded checkpoints (for final storage)
 save_interval_unsharded: 10000
 save_num_unsharded_checkpoints_to_keep: -1
 
-load_path: null
+load_path: /mnt/nas/hoyeon/OLMo/official_checkpoints/step50000-unsharded
 
 max_duration: 739_328  # 3.1T tokens
 global_train_batch_size: 2048
@@ -85,9 +85,7 @@ speed_monitor:
 eval_interval: 999999999999
 eval_subset_num_batches: -1
 device_eval_batch_size: ${device_train_microbatch_size}
-evaluators:
-  - label: sciq
-    type: downstream
+evaluators: []
 
 data:
   pad_direction: right

diff --git a/configs/official/OLMo-7B.yaml b/configs/official/OLMo-7B.yaml
@@ -91,9 +91,7 @@ speed_monitor:
 eval_interval: 99999999
 eval_subset_num_batches: -1
 device_eval_batch_size: ${device_train_microbatch_size}
-evaluators:
-  - label: sciq
-    type: downstream
+evaluators: []
 
 data:
   pad_direction: right