jvstokes
commented
1 week ago +1. Did you find a solution @engageandy?
Open engageandy opened 2 months ago
jvstokes
commented
1 week ago +1. Did you find a solution @engageandy?
Product Version
Issue Description
JumpStart throws an error when attempting to fine-tune Meta-Llama-3.1-70B-Instruct using ml.p4d.24xlarge.
No S3 objects found under S3 URL "s3://jumpstart-cache-prod-us-west-2/meta-training/train-meta-textgeneration-llama-3-1-70b-instruct.tar.gz" given in input data source. Please ensure that the bucket exists in the selected region (us-west-2), that objects exist under that S3 prefix, and that the role "arn:aws:iam::831875310015:role/service-role/AmazonSageMaker-ExecutionRole-20240812T070946" has "s3:ListBucket" permissions on bucket "jumpstart-cache-prod-us-west-2".
The only options are ml.p4d.24xlarge and ml.g5.48xlarge. The ml.g5 is too small to load the model so Meta-Llama-3.1-70B-Instruct is impossible to tune via JumpStart.
Expected Behavior
No response
Observed Behavior
No S3 objects found under S3 URL "s3://jumpstart-cache-prod-us-west-2/meta-training/train-meta-textgeneration-llama-3-1-70b-instruct.tar.gz" given in input data source. Please ensure that the bucket exists in the selected region (us-west-2), that objects exist under that S3 prefix, and that the role "arn:aws:iam::831875310015:role/service-role/AmazonSageMaker-ExecutionRole-20240812T070946" has "s3:ListBucket" permissions on bucket "jumpstart-cache-prod-us-west-2".
Product Category
JumpStart, Models
Feedback Category
Reliability and Stability
Other Details
No response