TensorRT-LLM
TensorRT-LLM copied to clipboard
Medusa convert cpt doesn't work with safetensors version
System Info
N/A
Who can help?
No response
Information
- [x] The official example scripts
- [ ] My own modified scripts
Tasks
- [x] An officially supported task in the
examples
folder (such as GLUE/SQuAD, ...) - [ ] My own task or dataset (give details below)
Reproduction
try converting a medusa checkpoint using text-generation-inference/Mistral-7B-Instruct-v0.2-medusa
Expected behavior
Should load from safe tensors
actual behavior
script expects a "medusa_lm_head.pt" in folder
additional notes
n/a
Thanks for reporting this. @nivibilla You are correct. The script needs to be updated to be able to load safe tensors as well.
Opened a PR #1535 for this
@ncomly-nvidia kindly review this PR
This has been merged, we can close this issue