DeepSpeed
ebbcfd52 - qkv_out can be a single tensor or a list. Handling these cases separetely. (#1850)

Commit

3 years ago

qkv_out can be a single tensor or a list. Handling these cases separetely. (#1850) Co-authored-by: Jeff Rasley <jerasley@microsoft.com>

References

Author

samyam

Parents