Fix flash-attn + qlora not working with llama models

[tmm1]

fixes this error:

  File "/mnt/ml/axolotl/src/axolotl/flash_attn.py", line 98, in forward                                                    
    output_unpad = flash_attn_varlen_qkvpacked_func(                                                                       
                   ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^                                                                       
  File "/mnt/ml/flash-attention/flash_attn/flash_attn_interface.py", line 406, in flash_attn_varlen_qkvpacked_func         
    return FlashAttnVarlenQKVPackedFunc.apply(                                                                             
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^                                                                             
  File "/home/tmm1/micromamba/envs/test/lib/python3.11/site-packages/torch/autograd/function.py", line 539, in apply       
    return super().apply(*args, **kwargs)  # type: ignore[misc]                                                            
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^                                                                                  
  File "/mnt/ml/flash-attention/flash_attn/flash_attn_interface.py", line 123, in forward                                  
    out, q, k, v, out_padded, softmax_lse, S_dmask, rng_state = _flash_attn_varlen_forward(                                
                                                                ^^^^^^^^^^^^^^^^^^^^^^^^^^^                                
  File "/mnt/ml/flash-attention/flash_attn/flash_attn_interface.py", line 52, in _flash_attn_varlen_forward                
    out, q, k, v, out_padded, softmax_lse, S_dmask, rng_state = flash_attn_cuda.varlen_fwd(                                
                                                                ^^^^^^^^^^^^^^^^^^^^^^^^^^^                                
RuntimeError: FlashAttention only support fp16 and bf16 data type