Can not load OPT-175B model.

🐛 Bug

I am trying to run OPT-175B model with the latest code in this repository. I have resharded the model into 16 shards after consolidating 992 shards from the download link. However, when i execute the metaseq.cli i get following error: RuntimeError: Error(s) in loading state_dict for ModelParallelTransformerLanguageModel: Missing key(s) in state_dict: "decoder.layers.0.self_attn.qkv_proj.weight", "decoder.layers.0.self_attn.qkv_proj.bias", "decoder.layers.1.self_attn.qkv_proj.weight", "decoder.layers.1.self_attn.qkv_proj.bias", "decoder.layers.2.self_attn.qkv_proj.weight", "decoder.layers.2.self_attn.qkv_proj.bias", "decoder.layers.3.self_attn.qkv_proj.weight", "decoder.layers.3.self_attn.qkv_proj.bias", "decoder.layers.4.self_attn.qkv_proj.weight", "decoder.layers.4.self_attn.qkv_proj.bias", "decoder.layers.5.self_attn.qkv_proj.weight", "decoder.layers.5.self_attn.qkv_proj.bias", "decoder.layers.6.self_attn.qkv_proj.weight", "decoder.layers.6.self_attn.qkv_proj.bias", "decoder.layers.7.self_attn.qkv_proj.weight", "decoder.layers.7.self_attn.qkv_proj.bias", "decoder.layers.8.self_attn.qkv_proj.weight", "decoder.layers.8.self_attn.qkv_proj.bias", "decoder.layers.9.self_attn.qkv_proj.weight", "decoder.layers.9.self_attn.qkv_proj.bias", "decoder.layers.10.self_attn.qkv_proj.weight", "decoder.layers.10.self_attn.qkv_proj.bias", "decoder.layers.11.self_attn.qkv_proj.weight", "decoder.layers.11.self_attn.qkv_proj.bias", "decoder.layers.12.self_attn.qkv_proj.weight", "decoder.layers.12.self_attn.qkv_proj.bias", "decoder.layers.13.self_attn.qkv_proj.weight", "decoder.layers.13.self_attn.qkv_proj.bias", "decoder.layers.14.self_attn.qkv_proj.weight", "decoder.layers.14.self_attn.qkv_proj.bias", "decoder.layers.15.self_attn.qkv_proj.weight", "decoder.layers.15.self_attn.qkv_proj.bias", "decoder.layers.16.self_attn.qkv_proj.weight", "decoder.layers.16.self_attn.qkv_proj.bias", "decoder.layers.17.self_attn.qkv_proj.weight", "decoder.layers.17.self_attn.qkv_proj.bias", "decoder.layers.18.self_attn.qkv_proj.weight", "decoder.layers.18.self_attn.qkv_proj.bias", "decoder.layers.19.self_attn.qkv_proj.weight", "decoder.layers.19.self_attn.qkv_proj.bias", "decoder.layers.20.self_attn.qkv_proj.weight", "decoder.layers.20.self_attn.qkv_proj.bias", "decoder.layers.21.self_attn.qkv_proj.weight", "decoder.layers.21.self_attn.qkv_proj.bias", "decoder.layers.22.self_attn.qkv_proj.weight", "decoder.layers.22.self_attn.qkv_proj.bias", "decoder.layers.23.self_attn.qkv_proj.weight", "decoder.layers.23.self_attn.qkv_proj.bias", "decoder.layers.24.self_attn.qkv_proj.weight", "decoder.layers.24.self_attn.qkv_proj.bias", "decoder.layers.25.self_attn.qkv_proj.weight", "decoder.layers.25.self_attn.qkv_proj.bias", "decoder.layers.26.self_attn.qkv_proj.weight", "decoder.layers.26.self_attn.qkv_proj.bias", "decoder.layers.27.self_attn.qkv_proj.weight", "decoder.layers.27.self_attn.qkv_proj.bias", "decoder.layers.28.self_attn.qkv_proj.weight", "decoder.layers.28.self_attn.qkv_proj.bias", "decoder.layers.29.self_attn.qkv_proj.weight", "decoder.layers.29.self_attn.qkv_proj.bias", "decoder.layers.30.self_attn.qkv_proj.weight", "decoder.layers.30.self_attn.qkv_proj.bias", "decoder.layers.31.self_attn.qkv_proj.weight", "decoder.layers.31.self_attn.qkv_proj.bias", "decoder.layers.32.self_attn.qkv_proj.weight", "decoder.layers.32.self_attn.qkv_proj.bias", "decoder.layers.33.self_attn.qkv_proj.weight", "decoder.layers.33.self_attn.qkv_proj.bias", "decoder.layers.34.self_attn.qkv_proj.weight", "decoder.layers.34.self_attn.qkv_proj.bias", "decoder.layers.35.self_attn.qkv_proj.weight", "decoder.layers.35.self_attn.qkv_proj.bias", "decoder.layers.36.self_attn.qkv_proj.weight", "decoder.layers.36.self_attn.qkv_proj.bias", "decoder.layers.37.self_attn.qkv_proj.weight", "decoder.layers.37.self_attn.qkv_proj.bias", "decoder.layers.38.self_attn.qkv_proj.weight", "decoder.layers.38.self_attn.qkv_proj.bias", "decoder.layers.39.self_attn.qkv_proj.weight", "decoder.layers.39.self_attn.qkv_proj.bias", "decoder.layers.40.self_attn.qkv_proj.weight", "decoder.layers.40.self_attn.qkv_proj.bias", "decoder.layers.41.self_attn.qkv_proj.weight", "decoder.layers.41.self_attn.qkv_proj.bias", "decoder.layers.42.self_attn.qkv_proj.weight", "decoder.layers.42.self_attn.qkv_proj.bias", "decoder.layers.43.self_attn.qkv_proj.weight", "decoder.layers.43.self_attn.qkv_proj.bias", "decoder.layers.44.self_attn.qkv_proj.weight", "decoder.layers.44.self_attn.qkv_proj.bias", "decoder.layers.45.self_attn.qkv_proj.weight", "decoder.layers.45.self_attn.qkv_proj.bias", "decoder.layers.46.self_attn.qkv_proj.weight", "decoder.layers.46.self_attn.qkv_proj.bias", "decoder.layers.47.self_attn.qkv_proj.weight", "decoder.layers.47.self_attn.qkv_proj.bias", "decoder.layers.48.self_attn.qkv_proj.weight", "decoder.layers.48.self_attn.qkv_proj.bias", "decoder.layers.49.self_attn.qkv_proj.weight", "decoder.layers.49.self_attn.qkv_proj.bias", "decoder.layers.50.self_attn.qkv_proj.weight", "decoder.layers.50.self_attn.qkv_proj.bias", "decoder.layers.51.self_attn.qkv_proj.weight", "decoder.layers.51.self_attn.qkv_proj.bias", "decoder.layers.52.self_attn.qkv_proj.weight", "decoder.layers.52.self_attn.qkv_proj.bias", "decoder.layers.53.self_attn.qkv_proj.weight", "decoder.layers.53.self_attn.qkv_proj.bias", "decoder.layers.54.self_attn.qkv_proj.weight", "decoder.layers.54.self_attn.qkv_proj.bias", "decoder.layers.55.self_attn.qkv_proj.weight", "decoder.layers.55.self_attn.qkv_proj.bias", "decoder.layers.56.self_attn.qkv_proj.weight", "decoder.layers.56.self_attn.qkv_proj.bias", "decoder.layers.57.self_attn.qkv_proj.weight", "decoder.layers.57.self_attn.qkv_proj.bias", "decoder.layers.58.self_attn.qkv_proj.weight", "decoder.layers.58.self_attn.qkv_proj.bias", "decoder.layers.59.self_attn.qkv_proj.weight", "decoder.layers.59.self_attn.qkv_proj.bias", "decoder.layers.60.self_attn.qkv_proj.weight", "decoder.layers.60.self_attn.qkv_proj.bias", "decoder.layers.61.self_attn.qkv_proj.weight", "decoder.layers.61.self_attn.qkv_proj.bias", "decoder.layers.62.self_attn.qkv_proj.weight", "decoder.layers.62.self_attn.qkv_proj.bias", "decoder.layers.63.self_attn.qkv_proj.weight", "decoder.layers.63.self_attn.qkv_proj.bias", "decoder.layers.64.self_attn.qkv_proj.weight", "decoder.layers.64.self_attn.qkv_proj.bias", "decoder.layers.65.self_attn.qkv_proj.weight", "decoder.layers.65.self_attn.qkv_proj.bias", "decoder.layers.66.self_attn.qkv_proj.weight", "decoder.layers.66.self_attn.qkv_proj.bias", "decoder.layers.67.self_attn.qkv_proj.weight", "decoder.layers.67.self_attn.qkv_proj.bias", "decoder.layers.68.self_attn.qkv_proj.weight", "decoder.layers.68.self_attn.qkv_proj.bias", "decoder.layers.69.self_attn.qkv_proj.weight", "decoder.layers.69.self_attn.qkv_proj.bias", "decoder.layers.70.self_attn.qkv_proj.weight", "decoder.layers.70.self_attn.qkv_proj.bias", "decoder.layers.71.self_attn.qkv_proj.weight", "decoder.layers.71.self_attn.qkv_proj.bias", "decoder.layers.72.self_attn.qkv_proj.weight", "decoder.layers.72.self_attn.qkv_proj.bias", "decoder.layers.73.self_attn.qkv_proj.weight", "decoder.layers.73.self_attn.qkv_proj.bias", "decoder.layers.74.self_attn.qkv_proj.weight", "decoder.layers.74.self_attn.qkv_proj.bias", "decoder.layers.75.self_attn.qkv_proj.weight", "decoder.layers.75.self_attn.qkv_proj.bias", "decoder.layers.76.self_attn.qkv_proj.weight", "decoder.layers.76.self_attn.qkv_proj.bias", "decoder.layers.77.self_attn.qkv_proj.weight", "decoder.layers.77.self_attn.qkv_proj.bias", "decoder.layers.78.self_attn.qkv_proj.weight", "decoder.layers.78.self_attn.qkv_proj.bias", "decoder.layers.79.self_attn.qkv_proj.weight", "decoder.layers.79.self_attn.qkv_proj.bias", "decoder.layers.80.self_attn.qkv_proj.weight", "decoder.layers.80.self_attn.qkv_proj.bias", "decoder.layers.81.self_attn.qkv_proj.weight", "decoder.layers.81.self_attn.qkv_proj.bias", "decoder.layers.82.self_attn.qkv_proj.weight", "decoder.layers.82.self_attn.qkv_proj.bias", "decoder.layers.83.self_attn.qkv_proj.weight", "decoder.layers.83.self_attn.qkv_proj.bias", "decoder.layers.84.self_attn.qkv_proj.weight", "decoder.layers.84.self_attn.qkv_proj.bias", "decoder.layers.85.self_attn.qkv_proj.weight", "decoder.layers.85.self_attn.qkv_proj.bias", "decoder.layers.86.self_attn.qkv_proj.weight", "decoder.layers.86.self_attn.qkv_proj.bias", "decoder.layers.87.self_attn.qkv_proj.weight", "decoder.layers.87.self_attn.qkv_proj.bias", "decoder.layers.88.self_attn.qkv_proj.weight", "decoder.layers.88.self_attn.qkv_proj.bias", "decoder.layers.89.self_attn.qkv_proj.weight", "decoder.layers.89.self_attn.qkv_proj.bias", "decoder.layers.90.self_attn.qkv_proj.weight", "decoder.layers.90.self_attn.qkv_proj.bias", "decoder.layers.91.self_attn.qkv_proj.weight", "decoder.layers.91.self_attn.qkv_proj.bias", "decoder.layers.92.self_attn.qkv_proj.weight", "decoder.layers.92.self_attn.qkv_proj.bias", "decoder.layers.93.self_attn.qkv_proj.weight", "decoder.layers.93.self_attn.qkv_proj.bias", "decoder.layers.94.self_attn.qkv_proj.weight", "decoder.layers.94.self_attn.qkv_proj.bias", "decoder.layers.95.self_attn.qkv_proj.weight", "decoder.layers.95.self_attn.qkv_proj.bias". Unexpected key(s) in state_dict: "decoder.layers.0.self_attn.k_proj.weight", "decoder.layers.0.self_attn.v_proj.weight", "decoder.layers.0.self_attn.q_proj.weight", "decoder.layers.0.self_attn.k_proj.bias", "decoder.layers.0.self_attn.v_proj.bias", "decoder.layers.0.self_attn.q_proj.bias", "decoder.layers.1.self_attn.k_proj.weight", "decoder.layers.1.self_attn.v_proj.weight", "decoder.layers.1.self_attn.q_proj.weight", "decoder.layers.1.self_attn.k_proj.bias", "decoder.layers.1.self_attn.v_proj.bias", "decoder.layers.1.self_attn.q_proj.bias", "decoder.layers.2.self_attn.k_proj.weight", "decoder.layers.2.self_attn.v_proj.weight", "decoder.layers.2.self_attn.q_proj.weight", "decoder.layers.2.self_attn.k_proj.bias", "decoder.layers.2.self_attn.v_proj.bias", "decoder.layers.2.self_attn.q_proj.bias", "decoder.layers.3.self_attn.k_proj.weight", "decoder.layers.3.self_attn.v_proj.weight", "decoder.layers.3.self_attn.q_proj.weight", "decoder.layers.3.self_attn.k_proj.bias", "decoder.layers.3.self_attn.v_proj.bias", "decoder.layers.3.self_attn.q_proj.bias", "decoder.layers.4.self_attn.k_proj.weight", "decoder.layers.4.self_attn.v_proj.weight", "decoder.layers.4.self_attn.q_proj.weight", "decoder.layers.4.self_attn.k_proj.bias", "decoder.layers.4.self_attn.v_proj.bias", "decoder.layers.4.self_attn.q_proj.bias", "decoder.layers.5.self_attn.k_proj.weight", "decoder.layers.5.self_attn.v_proj.weight", "decoder.layers.5.self_attn.q_proj.weight", "decoder.layers.5.self_attn.k_proj.bias", "decoder.layers.5.self_attn.v_proj.bias", "decoder.layers.5.self_attn.q_proj.bias", "decoder.layers.6.self_attn.k_proj.weight", "decoder.layers.6.self_attn.v_proj.weight", "decoder.layers.6.self_attn.q_proj.weight", "decoder.layers.6.self_attn.k_proj.bias", "decoder.layers.6.self_attn.v_proj.bias", "decoder.layers.6.self_attn.q_proj.bias", "decoder.layers.7.self_attn.k_proj.weight", "decoder.layers.7.self_attn.v_proj.weight", "decoder.layers.7.self_attn.q_proj.weight", "decoder.layers.7.self_attn.k_proj.bias", "decoder.layers.7.self_attn.v_proj.bias", "decoder.layers.7.self_attn.q_proj.bias", "decoder.layers.8.self_attn.k_proj.weight", "decoder.layers.8.self_attn.v_proj.weight", "decoder.layers.8.self_attn.q_proj.weight", "decoder.layers.8.self_attn.k_proj.bias", "decoder.layers.8.self_attn.v_proj.bias", "decoder.layers.8.self_attn.q_proj.bias", "decoder.layers.9.self_attn.k_proj.weight", "decoder.layers.9.self_attn.v_proj.weight", "decoder.layers.9.self_attn.q_proj.weight", "decoder.layers.9.self_attn.k_proj.bias", "decoder.layers.9.self_attn.v_proj.bias", "decoder.layers.9.self_attn.q_proj.bias", "decoder.layers.10.self_attn.k_proj.weight", "decoder.layers.10.self_attn.v_proj.weight", "decoder.layers.10.self_attn.q_proj.weight", "decoder.layers.10.self_attn.k_proj.bias", "decoder.layers.10.self_attn.v_proj.bias", "decoder.layers.10.self_attn.q_proj.bias", "decoder.layers.11.self_attn.k_proj.weight", "decoder.layers.11.self_attn.v_proj.weight", "decoder.layers.11.self_attn.q_proj.weight", "decoder.layers.11.self_attn.k_proj.bias", "decoder.layers.11.self_attn.v_proj.bias", "decoder.layers.11.self_attn.q_proj.bias", "decoder.layers.12.self_attn.k_proj.weight", "decoder.layers.12.self_attn.v_proj.weight", "decoder.layers.12.self_attn.q_proj.weight", "decoder.layers.12.self_attn.k_proj.bias", "decoder.layers.12.self_attn.v_proj.bias", "decoder.layers.12.self_attn.q_proj.bias", "decoder.layers.13.self_attn.k_proj.weight", "decoder.layers.13.self_attn.v_proj.weight", "decoder.layers.13.self_attn.q_proj.weight", "decoder.layers.13.self_attn.k_proj.bias", "decoder.layers.13.self_attn.v_proj.bias", "decoder.layers.13.self_attn.q_proj.bias", "decoder.layers.14.self_attn.k_proj.weight", "decoder.layers.14.self_attn.v_proj.weight", "decoder.layers.14.self_attn.q_proj.weight", "decoder.layers.14.self_attn.k_proj.bias", "decoder.layers.14.self_attn.v_proj.bias", "decoder.layers.14.self_attn.q_proj.bias", "decoder.layers.15.self_attn.k_proj.weight", "decoder.layers.15.self_attn.v_proj.weight", "decoder.layers.15.self_attn.q_proj.weight", "decoder.layers.15.self_attn.k_proj.bias", "decoder.layers.15.self_attn.v_proj.bias", "decoder.layers.15.self_attn.q_proj.bias", "decoder.layers.16.self_attn.k_proj.weight", "decoder.layers.16.self_attn.v_proj.weight", "decoder.layers.16.self_attn.q_proj.weight", "decoder.layers.16.self_attn.k_proj.bias", "decoder.layers.16.self_attn.v_proj.bias", "decoder.layers.16.self_attn.q_proj.bias", "decoder.layers.17.self_attn.k_proj.weight", "decoder.layers.17.self_attn.v_proj.weight", "decoder.layers.17.self_attn.q_proj.weight", "decoder.layers.17.self_attn.k_proj.bias", "decoder.layers.17.self_attn.v_proj.bias", "decoder.layers.17.self_attn.q_proj.bias", "decoder.layers.18.self_attn.k_proj.weight", "decoder.layers.18.self_attn.v_proj.weight", "decoder.layers.18.self_attn.q_proj.weight", "decoder.layers.18.self_attn.k_proj.bias", "decoder.layers.18.self_attn.v_proj.bias", "decoder.layers.18.self_attn.q_proj.bias", "decoder.layers.19.self_attn.k_proj.weight", "decoder.layers.19.self_attn.v_proj.weight", "decoder.layers.19.self_attn.q_proj.weight", "decoder.layers.19.self_attn.k_proj.bias", "decoder.layers.19.self_attn.v_proj.bias", "decoder.layers.19.self_attn.q_proj.bias", "decoder.layers.20.self_attn.k_proj.weight", "decoder.layers.20.self_attn.v_proj.weight", "decoder.layers.20.self_attn.q_proj.weight", "decoder.layers.20.self_attn.k_proj.bias", "decoder.layers.20.self_attn.v_proj.bias", "decoder.layers.20.self_attn.q_proj.bias", "decoder.layers.21.self_attn.k_proj.weight", "decoder.layers.21.self_attn.v_proj.weight", "decoder.layers.21.self_attn.q_proj.weight", "decoder.layers.21.self_attn.k_proj.bias", "decoder.layers.21.self_attn.v_proj.bias", "decoder.layers.21.self_attn.q_proj.bias", "decoder.layers.22.self_attn.k_proj.weight", "decoder.layers.22.self_attn.v_proj.weight", "decoder.layers.22.self_attn.q_proj.weight", "decoder.layers.22.self_attn.k_proj.bias", "decoder.layers.22.self_attn.v_proj.bias", "decoder.layers.22.self_attn.q_proj.bias", "decoder.layers.23.self_attn.k_proj.weight", "decoder.layers.23.self_attn.v_proj.weight", "decoder.layers.23.self_attn.q_proj.weight", "decoder.layers.23.self_attn.k_proj.bias", "decoder.layers.23.self_attn.v_proj.bias", "decoder.layers.23.self_attn.q_proj.bias", "decoder.layers.24.self_attn.k_proj.weight", "decoder.layers.24.self_attn.v_proj.weight", "decoder.layers.24.self_attn.q_proj.weight", "decoder.layers.24.self_attn.k_proj.bias", "decoder.layers.24.self_attn.v_proj.bias", "decoder.layers.24.self_attn.q_proj.bias", "decoder.layers.25.self_attn.k_proj.weight", "decoder.layers.25.self_attn.v_proj.weight", "decoder.layers.25.self_attn.q_proj.weight", "decoder.layers.25.self_attn.k_proj.bias", "decoder.layers.25.self_attn.v_proj.bias", "decoder.layers.25.self_attn.q_proj.bias", "decoder.layers.26.self_attn.k_proj.weight", "decoder.layers.26.self_attn.v_proj.weight", "decoder.layers.26.self_attn.q_proj.weight", "decoder.layers.26.self_attn.k_proj.bias", "decoder.layers.26.self_attn.v_proj.bias", "decoder.layers.26.self_attn.q_proj.bias", "decoder.layers.27.self_attn.k_proj.weight", "decoder.layers.27.self_attn.v_proj.weight", "decoder.layers.27.self_attn.q_proj.weight", "decoder.layers.27.self_attn.k_proj.bias", "decoder.layers.27.self_attn.v_proj.bias", "decoder.layers.27.self_attn.q_proj.bias", "decoder.layers.28.self_attn.k_proj.weight", "decoder.layers.28.self_attn.v_proj.weight", "decoder.layers.28.self_attn.q_proj.weight", "decoder.layers.28.self_attn.k_proj.bias", "decoder.layers.28.self_attn.v_proj.bias", "decoder.layers.28.self_attn.q_proj.bias", "decoder.layers.29.self_attn.k_proj.weight", "decoder.layers.29.self_attn.v_proj.weight", "decoder.layers.29.self_attn.q_proj.weight", "decoder.layers.29.self_attn.k_proj.bias", "decoder.layers.29.self_attn.v_proj.bias", "decoder.layers.29.self_attn.q_proj.bias", "decoder.layers.30.self_attn.k_proj.weight", "decoder.layers.30.self_attn.v_proj.weight", "decoder.layers.30.self_attn.q_proj.weight", "decoder.layers.30.self_attn.k_proj.bias", "decoder.layers.30.self_attn.v_proj.bias", "decoder.layers.30.self_attn.q_proj.bias", "decoder.layers.31.self_attn.k_proj.weight", "decoder.layers.31.self_attn.v_proj.weight", "decoder.layers.31.self_attn.q_proj.weight", "decoder.layers.31.self_attn.k_proj.bias", "decoder.layers.31.self_attn.v_proj.bias", "decoder.layers.31.self_attn.q_proj.bias", "decoder.layers.32.self_attn.k_proj.weight", "decoder.layers.32.self_attn.v_proj.weight", "decoder.layers.32.self_attn.q_proj.weight", "decoder.layers.32.self_attn.k_proj.bias", "decoder.layers.32.self_attn.v_proj.bias", "decoder.layers.32.self_attn.q_proj.bias", "decoder.layers.33.self_attn.k_proj.weight", "decoder.layers.33.self_attn.v_proj.weight", "decoder.layers.33.self_attn.q_proj.weight", "decoder.layers.33.self_attn.k_proj.bias", "decoder.layers.33.self_attn.v_proj.bias", "decoder.layers.33.self_attn.q_proj.bias", "decoder.layers.34.self_attn.k_proj.weight", "decoder.layers.34.self_attn.v_proj.weight", "decoder.layers.34.self_attn.q_proj.weight", "decoder.layers.34.self_attn.k_proj.bias", "decoder.layers.34.self_attn.v_proj.bias", "decoder.layers.34.self_attn.q_proj.bias", "decoder.layers.35.self_attn.k_proj.weight", "decoder.layers.35.self_attn.v_proj.weight", "decoder.layers.35.self_attn.q_proj.weight", "decoder.layers.35.self_attn.k_proj.bias", "decoder.layers.35.self_attn.v_proj.bias", "decoder.layers.35.self_attn.q_proj.bias", "decoder.layers.36.self_attn.k_proj.weight", "decoder.layers.36.self_attn.v_proj.weight", "decoder.layers.36.self_attn.q_proj.weight", "decoder.layers.36.self_attn.k_proj.bias", "decoder.layers.36.self_attn.v_proj.bias", "decoder.layers.36.self_attn.q_proj.bias", "decoder.layers.37.self_attn.k_proj.weight", "decoder.layers.37.self_attn.v_proj.weight", "decoder.layers.37.self_attn.q_proj.weight", "decoder.layers.37.self_attn.k_proj.bias", "decoder.layers.37.self_attn.v_proj.bias", "decoder.layers.37.self_attn.q_proj.bias", "decoder.layers.38.self_attn.k_proj.weight", "decoder.layers.38.self_attn.v_proj.weight", "decoder.layers.38.self_attn.q_proj.weight", "decoder.layers.38.self_attn.k_proj.bias", "decoder.layers.38.self_attn.v_proj.bias", "decoder.layers.38.self_attn.q_proj.bias", "decoder.layers.39.self_attn.k_proj.weight", "decoder.layers.39.self_attn.v_proj.weight", "decoder.layers.39.self_attn.q_proj.weight", "decoder.layers.39.self_attn.k_proj.bias", "decoder.layers.39.self_attn.v_proj.bias", "decoder.layers.39.self_attn.q_proj.bias", "decoder.layers.40.self_attn.k_proj.weight", "decoder.layers.40.self_attn.v_proj.weight", "decoder.layers.40.self_attn.q_proj.weight", "decoder.layers.40.self_attn.k_proj.bias", "decoder.layers.40.self_attn.v_proj.bias", "decoder.layers.40.self_attn.q_proj.bias", "decoder.layers.41.self_attn.k_proj.weight", "decoder.layers.41.self_attn.v_proj.weight", "decoder.layers.41.self_attn.q_proj.weight", "decoder.layers.41.self_attn.k_proj.bias", "decoder.layers.41.self_attn.v_proj.bias", "decoder.layers.41.self_attn.q_proj.bias", "decoder.layers.42.self_attn.k_proj.weight", "decoder.layers.42.self_attn.v_proj.weight", "decoder.layers.42.self_attn.q_proj.weight", "decoder.layers.42.self_attn.k_proj.bias", "decoder.layers.42.self_attn.v_proj.bias", "decoder.layers.42.self_attn.q_proj.bias", "decoder.layers.43.self_attn.k_proj.weight", "decoder.layers.43.self_attn.v_proj.weight", "decoder.layers.43.self_attn.q_proj.weight", "decoder.layers.43.self_attn.k_proj.bias", "decoder.layers.43.self_attn.v_proj.bias", "decoder.layers.43.self_attn.q_proj.bias", "decoder.layers.44.self_attn.k_proj.weight", "decoder.layers.44.self_attn.v_proj.weight", "decoder.layers.44.self_attn.q_proj.weight", "decoder.layers.44.self_attn.k_proj.bias", "decoder.layers.44.self_attn.v_proj.bias", "decoder.layers.44.self_attn.q_proj.bias", "decoder.layers.45.self_attn.k_proj.weight", "decoder.layers.45.self_attn.v_proj.weight", "decoder.layers.45.self_attn.q_proj.weight", "decoder.layers.45.self_attn.k_proj.bias", "decoder.layers.45.self_attn.v_proj.bias", "decoder.layers.45.self_attn.q_proj.bias", "decoder.layers.46.self_attn.k_proj.weight", "decoder.layers.46.self_attn.v_proj.weight", "decoder.layers.46.self_attn.q_proj.weight", "decoder.layers.46.self_attn.k_proj.bias", "decoder.layers.46.self_attn.v_proj.bias", "decoder.layers.46.self_attn.q_proj.bias", "decoder.layers.47.self_attn.k_proj.weight", "decoder.layers.47.self_attn.v_proj.weight", "decoder.layers.47.self_attn.q_proj.weight", "decoder.layers.47.self_attn.k_proj.bias", "decoder.layers.47.self_attn.v_proj.bias", "decoder.layers.47.self_attn.q_proj.bias", "decoder.layers.48.self_attn.k_proj.weight", "decoder.layers.48.self_attn.v_proj.weight", "decoder.layers.48.self_attn.q_proj.weight", "decoder.layers.48.self_attn.k_proj.bias", "decoder.layers.48.self_attn.v_proj.bias", "decoder.layers.48.self_attn.q_proj.bias", "decoder.layers.49.self_attn.k_proj.weight", "decoder.layers.49.self_attn.v_proj.weight", "decoder.layers.49.self_attn.q_proj.weight", "decoder.layers.49.self_attn.k_proj.bias", "decoder.layers.49.self_attn.v_proj.bias", "decoder.layers.49.self_attn.q_proj.bias", "decoder.layers.50.self_attn.k_proj.weight", "decoder.layers.50.self_attn.v_proj.weight", "decoder.layers.50.self_attn.q_proj.weight", "decoder.layers.50.self_attn.k_proj.bias", "decoder.layers.50.self_attn.v_proj.bias", "decoder.layers.50.self_attn.q_proj.bias", "decoder.layers.51.self_attn.k_proj.weight", "decoder.layers.51.self_attn.v_proj.weight", "decoder.layers.51.self_attn.q_proj.weight", "decoder.layers.51.self_attn.k_proj.bias", "decoder.layers.51.self_attn.v_proj.bias", "decoder.layers.51.self_attn.q_proj.bias", "decoder.layers.52.self_attn.k_proj.weight", "decoder.layers.52.self_attn.v_proj.weight", "decoder.layers.52.self_attn.q_proj.weight", "decoder.layers.52.self_attn.k_proj.bias", "decoder.layers.52.self_attn.v_proj.bias", "decoder.layers.52.self_attn.q_proj.bias", "decoder.layers.53.self_attn.k_proj.weight", "decoder.layers.53.self_attn.v_proj.weight", "decoder.layers.53.self_attn.q_proj.weight", "decoder.layers.53.self_attn.k_proj.bias", "decoder.layers.53.self_attn.v_proj.bias", "decoder.layers.53.self_attn.q_proj.bias", "decoder.layers.54.self_attn.k_proj.weight", "decoder.layers.54.self_attn.v_proj.weight", "decoder.layers.54.self_attn.q_proj.weight", "decoder.layers.54.self_attn.k_proj.bias", "decoder.layers.54.self_attn.v_proj.bias", "decoder.layers.54.self_attn.q_proj.bias", "decoder.layers.55.self_attn.k_proj.weight", "decoder.layers.55.self_attn.v_proj.weight", "decoder.layers.55.self_attn.q_proj.weight", "decoder.layers.55.self_attn.k_proj.bias", "decoder.layers.55.self_attn.v_proj.bias", "decoder.layers.55.self_attn.q_proj.bias", "decoder.layers.56.self_attn.k_proj.weight", "decoder.layers.56.self_attn.v_proj.weight", "decoder.layers.56.self_attn.q_proj.weight", "decoder.layers.56.self_attn.k_proj.bias", "decoder.layers.56.self_attn.v_proj.bias", "decoder.layers.56.self_attn.q_proj.bias", "decoder.layers.57.self_attn.k_proj.weight", "decoder.layers.57.self_attn.v_proj.weight", "decoder.layers.57.self_attn.q_proj.weight", "decoder.layers.57.self_attn.k_proj.bias", "decoder.layers.57.self_attn.v_proj.bias", "decoder.layers.57.self_attn.q_proj.bias", "decoder.layers.58.self_attn.k_proj.weight", "decoder.layers.58.self_attn.v_proj.weight", "decoder.layers.58.self_attn.q_proj.weight", "decoder.layers.58.self_attn.k_proj.bias", "decoder.layers.58.self_attn.v_proj.bias", "decoder.layers.58.self_attn.q_proj.bias", "decoder.layers.59.self_attn.k_proj.weight", "decoder.layers.59.self_attn.v_proj.weight", "decoder.layers.59.self_attn.q_proj.weight", "decoder.layers.59.self_attn.k_proj.bias", "decoder.layers.59.self_attn.v_proj.bias", "decoder.layers.59.self_attn.q_proj.bias", "decoder.layers.60.self_attn.k_proj.weight", "decoder.layers.60.self_attn.v_proj.weight", "decoder.layers.60.self_attn.q_proj.weight", "decoder.layers.60.self_attn.k_proj.bias", "decoder.layers.60.self_attn.v_proj.bias", "decoder.layers.60.self_attn.q_proj.bias", "decoder.layers.61.self_attn.k_proj.weight", "decoder.layers.61.self_attn.v_proj.weight", "decoder.layers.61.self_attn.q_proj.weight", "decoder.layers.61.self_attn.k_proj.bias", "decoder.layers.61.self_attn.v_proj.bias", "decoder.layers.61.self_attn.q_proj.bias", "decoder.layers.62.self_attn.k_proj.weight", "decoder.layers.62.self_attn.v_proj.weight", "decoder.layers.62.self_attn.q_proj.weight", "decoder.layers.62.self_attn.k_proj.bias", "decoder.layers.62.self_attn.v_proj.bias", "decoder.layers.62.self_attn.q_proj.bias", "decoder.layers.63.self_attn.k_proj.weight", "decoder.layers.63.self_attn.v_proj.weight", "decoder.layers.63.self_attn.q_proj.weight", "decoder.layers.63.self_attn.k_proj.bias", "decoder.layers.63.self_attn.v_proj.bias", "decoder.layers.63.self_attn.q_proj.bias", "decoder.layers.64.self_attn.k_proj.weight", "decoder.layers.64.self_attn.v_proj.weight", "decoder.layers.64.self_attn.q_proj.weight", "decoder.layers.64.self_attn.k_proj.bias", "decoder.layers.64.self_attn.v_proj.bias", "decoder.layers.64.self_attn.q_proj.bias", "decoder.layers.65.self_attn.k_proj.weight", "decoder.layers.65.self_attn.v_proj.weight", "decoder.layers.65.self_attn.q_proj.weight", "decoder.layers.65.self_attn.k_proj.bias", "decoder.layers.65.self_attn.v_proj.bias", "decoder.layers.65.self_attn.q_proj.bias", "decoder.layers.66.self_attn.k_proj.weight", "decoder.layers.66.self_attn.v_proj.weight", "decoder.layers.66.self_attn.q_proj.weight", "decoder.layers.66.self_attn.k_proj.bias", "decoder.layers.66.self_attn.v_proj.bias", "decoder.layers.66.self_attn.q_proj.bias", "decoder.layers.67.self_attn.k_proj.weight", "decoder.layers.67.self_attn.v_proj.weight", "decoder.layers.67.self_attn.q_proj.weight", "decoder.layers.67.self_attn.k_proj.bias", "decoder.layers.67.self_attn.v_proj.bias", "decoder.layers.67.self_attn.q_proj.bias", "decoder.layers.68.self_attn.k_proj.weight", "decoder.layers.68.self_attn.v_proj.weight", "decoder.layers.68.self_attn.q_proj.weight", "decoder.layers.68.self_attn.k_proj.bias", "decoder.layers.68.self_attn.v_proj.bias", "decoder.layers.68.self_attn.q_proj.bias", "decoder.layers.69.self_attn.k_proj.weight", "decoder.layers.69.self_attn.v_proj.weight", "decoder.layers.69.self_attn.q_proj.weight", "decoder.layers.69.self_attn.k_proj.bias", "decoder.layers.69.self_attn.v_proj.bias", "decoder.layers.69.self_attn.q_proj.bias", "decoder.layers.70.self_attn.k_proj.weight", "decoder.layers.70.self_attn.v_proj.weight", "decoder.layers.70.self_attn.q_proj.weight", "decoder.layers.70.self_attn.k_proj.bias", "decoder.layers.70.self_attn.v_proj.bias", "decoder.layers.70.self_attn.q_proj.bias", "decoder.layers.71.self_attn.k_proj.weight", "decoder.layers.71.self_attn.v_proj.weight", "decoder.layers.71.self_attn.q_proj.weight", "decoder.layers.71.self_attn.k_proj.bias", "decoder.layers.71.self_attn.v_proj.bias", "decoder.layers.71.self_attn.q_proj.bias", "decoder.layers.72.self_attn.k_proj.weight", "decoder.layers.72.self_attn.v_proj.weight", "decoder.layers.72.self_attn.q_proj.weight", "decoder.layers.72.self_attn.k_proj.bias", "decoder.layers.72.self_attn.v_proj.bias", "decoder.layers.72.self_attn.q_proj.bias", "decoder.layers.73.self_attn.k_proj.weight", "decoder.layers.73.self_attn.v_proj.weight", "decoder.layers.73.self_attn.q_proj.weight", "decoder.layers.73.self_attn.k_proj.bias", "decoder.layers.73.self_attn.v_proj.bias", "decoder.layers.73.self_attn.q_proj.bias", "decoder.layers.74.self_attn.k_proj.weight", "decoder.layers.74.self_attn.v_proj.weight", "decoder.layers.74.self_attn.q_proj.weight", "decoder.layers.74.self_attn.k_proj.bias", "decoder.layers.74.self_attn.v_proj.bias", "decoder.layers.74.self_attn.q_proj.bias", "decoder.layers.75.self_attn.k_proj.weight", "decoder.layers.75.self_attn.v_proj.weight", "decoder.layers.75.self_attn.q_proj.weight", "decoder.layers.75.self_attn.k_proj.bias", "decoder.layers.75.self_attn.v_proj.bias", "decoder.layers.75.self_attn.q_proj.bias", "decoder.layers.76.self_attn.k_proj.weight", "decoder.layers.76.self_attn.v_proj.weight", "decoder.layers.76.self_attn.q_proj.weight", "decoder.layers.76.self_attn.k_proj.bias", "decoder.layers.76.self_attn.v_proj.bias", "decoder.layers.76.self_attn.q_proj.bias", "decoder.layers.77.self_attn.k_proj.weight", "decoder.layers.77.self_attn.v_proj.weight", "decoder.layers.77.self_attn.q_proj.weight", "decoder.layers.77.self_attn.k_proj.bias", "decoder.layers.77.self_attn.v_proj.bias", "decoder.layers.77.self_attn.q_proj.bias", "decoder.layers.78.self_attn.k_proj.weight", "decoder.layers.78.self_attn.v_proj.weight", "decoder.layers.78.self_attn.q_proj.weight", "decoder.layers.78.self_attn.k_proj.bias", "decoder.layers.78.self_attn.v_proj.bias", "decoder.layers.78.self_attn.q_proj.bias", "decoder.layers.79.self_attn.k_proj.weight", "decoder.layers.79.self_attn.v_proj.weight", "decoder.layers.79.self_attn.q_proj.weight", "decoder.layers.79.self_attn.k_proj.bias", "decoder.layers.79.self_attn.v_proj.bias", "decoder.layers.79.self_attn.q_proj.bias", "decoder.layers.80.self_attn.k_proj.weight", "decoder.layers.80.self_attn.v_proj.weight", "decoder.layers.80.self_attn.q_proj.weight", "decoder.layers.80.self_attn.k_proj.bias", "decoder.layers.80.self_attn.v_proj.bias", "decoder.layers.80.self_attn.q_proj.bias", "decoder.layers.81.self_attn.k_proj.weight", "decoder.layers.81.self_attn.v_proj.weight", "decoder.layers.81.self_attn.q_proj.weight", "decoder.layers.81.self_attn.k_proj.bias", "decoder.layers.81.self_attn.v_proj.bias", "decoder.layers.81.self_attn.q_proj.bias", "decoder.layers.82.self_attn.k_proj.weight", "decoder.layers.82.self_attn.v_proj.weight", "decoder.layers.82.self_attn.q_proj.weight", "decoder.layers.82.self_attn.k_proj.bias", "decoder.layers.82.self_attn.v_proj.bias", "decoder.layers.82.self_attn.q_proj.bias", "decoder.layers.83.self_attn.k_proj.weight", "decoder.layers.83.self_attn.v_proj.weight", "decoder.layers.83.self_attn.q_proj.weight", "decoder.layers.83.self_attn.k_proj.bias", "decoder.layers.83.self_attn.v_proj.bias", "decoder.layers.83.self_attn.q_proj.bias", "decoder.layers.84.self_attn.k_proj.weight", "decoder.layers.84.self_attn.v_proj.weight", "decoder.layers.84.self_attn.q_proj.weight", "decoder.layers.84.self_attn.k_proj.bias", "decoder.layers.84.self_attn.v_proj.bias", "decoder.layers.84.self_attn.q_proj.bias", "decoder.layers.85.self_attn.k_proj.weight", "decoder.layers.85.self_attn.v_proj.weight", "decoder.layers.85.self_attn.q_proj.weight", "decoder.layers.85.self_attn.k_proj.bias", "decoder.layers.85.self_attn.v_proj.bias", "decoder.layers.85.self_attn.q_proj.bias", "decoder.layers.86.self_attn.k_proj.weight", "decoder.layers.86.self_attn.v_proj.weight", "decoder.layers.86.self_attn.q_proj.weight", "decoder.layers.86.self_attn.k_proj.bias", "decoder.layers.86.self_attn.v_proj.bias", "decoder.layers.86.self_attn.q_proj.bias", "decoder.layers.87.self_attn.k_proj.weight", "decoder.layers.87.self_attn.v_proj.weight", "decoder.layers.87.self_attn.q_proj.weight", "decoder.layers.87.self_attn.k_proj.bias", "decoder.layers.87.self_attn.v_proj.bias", "decoder.layers.87.self_attn.q_proj.bias", "decoder.layers.88.self_attn.k_proj.weight", "decoder.layers.88.self_attn.v_proj.weight", "decoder.layers.88.self_attn.q_proj.weight", "decoder.layers.88.self_attn.k_proj.bias", "decoder.layers.88.self_attn.v_proj.bias", "decoder.layers.88.self_attn.q_proj.bias", "decoder.layers.89.self_attn.k_proj.weight", "decoder.layers.89.self_attn.v_proj.weight", "decoder.layers.89.self_attn.q_proj.weight", "decoder.layers.89.self_attn.k_proj.bias", "decoder.layers.89.self_attn.v_proj.bias", "decoder.layers.89.self_attn.q_proj.bias", "decoder.layers.90.self_attn.k_proj.weight", "decoder.layers.90.self_attn.v_proj.weight", "decoder.layers.90.self_attn.q_proj.weight", "decoder.layers.90.self_attn.k_proj.bias", "decoder.layers.90.self_attn.v_proj.bias", "decoder.layers.90.self_attn.q_proj.bias", "decoder.layers.91.self_attn.k_proj.weight", "decoder.layers.91.self_attn.v_proj.weight", "decoder.layers.91.self_attn.q_proj.weight", "decoder.layers.91.self_attn.k_proj.bias", "decoder.layers.91.self_attn.v_proj.bias", "decoder.layers.91.self_attn.q_proj.bias", "decoder.layers.92.self_attn.k_proj.weight", "decoder.layers.92.self_attn.v_proj.weight", "decoder.layers.92.self_attn.q_proj.weight", "decoder.layers.92.self_attn.k_proj.bias", "decoder.layers.92.self_attn.v_proj.bias", "decoder.layers.92.self_attn.q_proj.bias", "decoder.layers.93.self_attn.k_proj.weight", "decoder.layers.93.self_attn.v_proj.weight", "decoder.layers.93.self_attn.q_proj.weight", "decoder.layers.93.self_attn.k_proj.bias", "decoder.layers.93.self_attn.v_proj.bias", "decoder.layers.93.self_attn.q_proj.bias", "decoder.layers.94.self_attn.k_proj.weight", "decoder.layers.94.self_attn.v_proj.weight", "decoder.layers.94.self_attn.q_proj.weight", "decoder.layers.94.self_attn.k_proj.bias", "decoder.layers.94.self_attn.v_proj.bias", "decoder.layers.94.self_attn.q_proj.bias", "decoder.layers.95.self_attn.k_proj.weight", "decoder.layers.95.self_attn.v_proj.weight", "decoder.layers.95.self_attn.q_proj.weight", "decoder.layers.95.self_attn.k_proj.bias", "decoder.layers.95.self_attn.v_proj.bias", "decoder.layers.95.self_attn.q_proj.bias". Done

It seems that the actual model files have changed but the pretrained model provided is different.

To Reproduce

Steps to reproduce the behavior (always include the command you ran):

1. Run cmd python3 -m metaseq.cli.interactive_hosted

• metaseq Version (e.g., 1.0 or master): latest cloned from this repository
• PyTorch Version (e.g., 1.0): 1.10
• OS (e.g., Linux, Windows, MacOS): Linux
• How you installed metaseq (pip, source): from original github code repository
• Build command you used (if compiling from source): pip install -e .;
• Python version: 3.8
• CUDA/cuDNN version:
• GPU models and configuration: A100-40GB

[gulzainali98]

https://github.com/facebookresearch/metaseq/issues/404

Updated the branch for fairscale … rerunning it now.

[EIFY]

@gulzainali98 “–use-sharded-state” needs to be in the LAUNCH_ARGS actually, see https://github.com/facebookresearch/metaseq/compare/punitkoura/debug-407