Fix llama_v2_7b_16h for torch.jit.trace

Original error: Attention using SDPA can not be traced with torch.jit.trace
when no attention_mask is provided. To solve this issue, please either load
your model with the argument attn_implementation="eager" or
pass an attention_mask input when tracing the model.

torchbenchmark/util/framework/huggingface/model_factory.py

@@ -89,6 +89,12 @@ def __init__(self, name, test, device, batch_size=None, extra_args=[]):
  if class_models[name][2] == "ReformerConfig()" and not config.num_buckets:
  # silence "config.num_buckets is not set. Setting config.num_buckets to 128"
  config.num_buckets = 128
+ if name == "llama_v2_7b_16h":
+ # Workaround for the following error:
+ # Attention using SDPA can not be traced with torch.jit.trace when no attention_mask is provided.
+ # To solve this issue, please either load your model with the argument attn_implementation="eager" or
+ # pass an attention_mask input when tracing the model.
+ config._attn_implementation = "eager"
  class_ctor = getattr(transformers, class_models[name][3])
  kwargs = {}
  hugging_face_models_requiring_trust_remote_code = ["hf_Falcon_7b", "hf_MPT_7b_instruct", "phi_1_5", "hf_Yi"]