DeepSpeed
beed962c - [Bug fix] Add rope_theta for llama config (#4480)

Commit
1 year ago
[Bug fix] Add rope_theta for llama config (#4480) * Add rope_theta for llama config * Add rope_theta to bias_add_transform_0213 * Fix CI problems * Add rope_theta to linear layer --------- Co-authored-by: Michael Wyatt <michaelwyatt@microsoft.com> Co-authored-by: Lev Kurilenko <113481193+lekurile@users.noreply.github.com>
Author
Parents
  • csrc/transformer/inference
    • csrc
      • File
        apply_rotary_pos_emb.cu
      • File
        pt_binding.cpp
      • File
        transform.cu
    • includes
      • File
        inference_cuda_layers.h
  • deepspeed
    • module_inject/containers
      • File
        llama.py
    • ops/transformer/inference
      • File
        config.py
      • op_binding
        • File
          linear.py
        • File
          softmax_context.py