You can deploy this model and start earning money today!

zhiyuanhusf/
tmax-ornith-397b

Name: tmax-ornith-397b
Availability: InStock

transformers

deepreinforce-ai/Ornith-1.0-397B

finetune

other

endpoints_compatible

Views

No views yet

TMax Ornith-1.0-397B DPPO

Full-parameter DPPO checkpoints trained on allenai/tmax-15k-open-instruct.

global_stepN branches contain resumable DeepSpeed state. Only the newest five state branches are retained.
step_N branches contain merged Hugging Face model weights. All merged branches are retained.
main contains metadata only.

Training uses nine 8xH200 learner nodes with ZeRO-3, CPU optimizer offload, and sequence parallelism, plus one 8xH200 tensor-parallel vLLM rollout node. Sampling uses temperature 1.0 and top-p 1.0.

zhiyuanhusf/tmax-ornith-397b

TMax Ornith-1.0-397B DPPO

zhiyuanhusf/
tmax-ornith-397b