diff --git a/recipes/configs/llama3/70B_full.yaml b/recipes/configs/llama3/70B_full.yaml
index a8b7ba619c..cd345822b6 100644
--- a/recipes/configs/llama3/70B_full.yaml
+++ b/recipes/configs/llama3/70B_full.yaml
@@ -15,11 +15,6 @@
 #
 # This config is only tested on an 8xA100 machine.
 #
-# !!!!!!!!!!!!!
-# !!!!!!!!!!!!!
-# ATTENTION: It will only work with pytorch>=2.5 (nightlies). For other pytorch versions, it will OOM, even on 8xA100.
-# !!!!!!!!!!!!!
-# !!!!!!!!!!!!!
 
 
 # Tokenizer
diff --git a/recipes/configs/llama3_1/405B_qlora.yaml b/recipes/configs/llama3_1/405B_qlora.yaml
index 6398a840ec..58f5eb9b1f 100644
--- a/recipes/configs/llama3_1/405B_qlora.yaml
+++ b/recipes/configs/llama3_1/405B_qlora.yaml
@@ -12,11 +12,7 @@
 # This config needs 8 GPUs to run
 #   # tune run --nproc_per_node 8 lora_finetune_distributed --config llama3_1/405B_qlora
 #
-# !!!!!!!!!!!!!
-# !!!!!!!!!!!!!
-# ATTENTION: It will only work with pytorch>=2.5 (nightlies). For other pytorch versions, it will OOM, even on 8xA100.
-# !!!!!!!!!!!!!
-# !!!!!!!!!!!!!
+
 
 # Model Arguments
 model:
diff --git a/recipes/configs/llama3_1/70B_full.yaml b/recipes/configs/llama3_1/70B_full.yaml
index fcae062999..2a7d19399c 100644
--- a/recipes/configs/llama3_1/70B_full.yaml
+++ b/recipes/configs/llama3_1/70B_full.yaml
@@ -15,11 +15,6 @@
 #
 # This config is only tested on an 8xA100 machine.
 #
-# !!!!!!!!!!!!!
-# !!!!!!!!!!!!!
-# ATTENTION: It will only work with pytorch>=2.5 (nightlies). For other pytorch versions, it will OOM, even on 8xA100.
-# !!!!!!!!!!!!!
-# !!!!!!!!!!!!!
 
 # Tokenizer
 tokenizer: