recurrentgemma-9b / model.safetensors.index.json
alpindale's picture
Upload folder using huggingface_hub
686e396 verified
{
"metadata": {
"total_size": 19257106432
},
"weight_map": {
"lm_head.weight": "model-00004-of-00004.safetensors",
"model.embed_tokens.weight": "model-00001-of-00004.safetensors",
"model.final_norm.weight": "model-00004-of-00004.safetensors",
"model.layers.0.channel_pre_norm.weight": "model-00001-of-00004.safetensors",
"model.layers.0.mlp_block.down_proj.bias": "model-00001-of-00004.safetensors",
"model.layers.0.mlp_block.down_proj.weight": "model-00001-of-00004.safetensors",
"model.layers.0.mlp_block.gate_proj.bias": "model-00001-of-00004.safetensors",
"model.layers.0.mlp_block.gate_proj.weight": "model-00001-of-00004.safetensors",
"model.layers.0.mlp_block.up_proj.bias": "model-00001-of-00004.safetensors",
"model.layers.0.mlp_block.up_proj.weight": "model-00001-of-00004.safetensors",
"model.layers.0.temporal_block.conv_1d.bias": "model-00001-of-00004.safetensors",
"model.layers.0.temporal_block.conv_1d.weight": "model-00001-of-00004.safetensors",
"model.layers.0.temporal_block.linear_out.bias": "model-00001-of-00004.safetensors",
"model.layers.0.temporal_block.linear_out.weight": "model-00001-of-00004.safetensors",
"model.layers.0.temporal_block.linear_x.bias": "model-00001-of-00004.safetensors",
"model.layers.0.temporal_block.linear_x.weight": "model-00001-of-00004.safetensors",
"model.layers.0.temporal_block.linear_y.bias": "model-00001-of-00004.safetensors",
"model.layers.0.temporal_block.linear_y.weight": "model-00001-of-00004.safetensors",
"model.layers.0.temporal_block.rg_lru.input_gate_bias": "model-00001-of-00004.safetensors",
"model.layers.0.temporal_block.rg_lru.input_gate_weight": "model-00001-of-00004.safetensors",
"model.layers.0.temporal_block.rg_lru.recurrent_gate_bias": "model-00001-of-00004.safetensors",
"model.layers.0.temporal_block.rg_lru.recurrent_gate_weight": "model-00001-of-00004.safetensors",
"model.layers.0.temporal_block.rg_lru.recurrent_param": "model-00001-of-00004.safetensors",
"model.layers.0.temporal_pre_norm.weight": "model-00001-of-00004.safetensors",
"model.layers.1.channel_pre_norm.weight": "model-00001-of-00004.safetensors",
"model.layers.1.mlp_block.down_proj.bias": "model-00001-of-00004.safetensors",
"model.layers.1.mlp_block.down_proj.weight": "model-00001-of-00004.safetensors",
"model.layers.1.mlp_block.gate_proj.bias": "model-00001-of-00004.safetensors",
"model.layers.1.mlp_block.gate_proj.weight": "model-00001-of-00004.safetensors",
"model.layers.1.mlp_block.up_proj.bias": "model-00001-of-00004.safetensors",
"model.layers.1.mlp_block.up_proj.weight": "model-00001-of-00004.safetensors",
"model.layers.1.temporal_block.conv_1d.bias": "model-00001-of-00004.safetensors",
"model.layers.1.temporal_block.conv_1d.weight": "model-00001-of-00004.safetensors",
"model.layers.1.temporal_block.linear_out.bias": "model-00001-of-00004.safetensors",
"model.layers.1.temporal_block.linear_out.weight": "model-00001-of-00004.safetensors",
"model.layers.1.temporal_block.linear_x.bias": "model-00001-of-00004.safetensors",
"model.layers.1.temporal_block.linear_x.weight": "model-00001-of-00004.safetensors",
"model.layers.1.temporal_block.linear_y.bias": "model-00001-of-00004.safetensors",
"model.layers.1.temporal_block.linear_y.weight": "model-00001-of-00004.safetensors",
"model.layers.1.temporal_block.rg_lru.input_gate_bias": "model-00001-of-00004.safetensors",
"model.layers.1.temporal_block.rg_lru.input_gate_weight": "model-00001-of-00004.safetensors",
"model.layers.1.temporal_block.rg_lru.recurrent_gate_bias": "model-00001-of-00004.safetensors",
"model.layers.1.temporal_block.rg_lru.recurrent_gate_weight": "model-00001-of-00004.safetensors",
"model.layers.1.temporal_block.rg_lru.recurrent_param": "model-00001-of-00004.safetensors",
"model.layers.1.temporal_pre_norm.weight": "model-00001-of-00004.safetensors",
"model.layers.10.channel_pre_norm.weight": "model-00002-of-00004.safetensors",
"model.layers.10.mlp_block.down_proj.bias": "model-00002-of-00004.safetensors",
"model.layers.10.mlp_block.down_proj.weight": "model-00002-of-00004.safetensors",
"model.layers.10.mlp_block.gate_proj.bias": "model-00002-of-00004.safetensors",
"model.layers.10.mlp_block.gate_proj.weight": "model-00002-of-00004.safetensors",
"model.layers.10.mlp_block.up_proj.bias": "model-00002-of-00004.safetensors",
"model.layers.10.mlp_block.up_proj.weight": "model-00002-of-00004.safetensors",
"model.layers.10.temporal_block.conv_1d.bias": "model-00002-of-00004.safetensors",
"model.layers.10.temporal_block.conv_1d.weight": "model-00002-of-00004.safetensors",
"model.layers.10.temporal_block.linear_out.bias": "model-00002-of-00004.safetensors",
"model.layers.10.temporal_block.linear_out.weight": "model-00002-of-00004.safetensors",
"model.layers.10.temporal_block.linear_x.bias": "model-00002-of-00004.safetensors",
"model.layers.10.temporal_block.linear_x.weight": "model-00002-of-00004.safetensors",
"model.layers.10.temporal_block.linear_y.bias": "model-00002-of-00004.safetensors",
"model.layers.10.temporal_block.linear_y.weight": "model-00002-of-00004.safetensors",
"model.layers.10.temporal_block.rg_lru.input_gate_bias": "model-00002-of-00004.safetensors",
"model.layers.10.temporal_block.rg_lru.input_gate_weight": "model-00002-of-00004.safetensors",
"model.layers.10.temporal_block.rg_lru.recurrent_gate_bias": "model-00002-of-00004.safetensors",
"model.layers.10.temporal_block.rg_lru.recurrent_gate_weight": "model-00002-of-00004.safetensors",
"model.layers.10.temporal_block.rg_lru.recurrent_param": "model-00002-of-00004.safetensors",
"model.layers.10.temporal_pre_norm.weight": "model-00002-of-00004.safetensors",
"model.layers.11.channel_pre_norm.weight": "model-00002-of-00004.safetensors",
"model.layers.11.mlp_block.down_proj.bias": "model-00002-of-00004.safetensors",
"model.layers.11.mlp_block.down_proj.weight": "model-00002-of-00004.safetensors",
"model.layers.11.mlp_block.gate_proj.bias": "model-00002-of-00004.safetensors",
"model.layers.11.mlp_block.gate_proj.weight": "model-00002-of-00004.safetensors",
"model.layers.11.mlp_block.up_proj.bias": "model-00002-of-00004.safetensors",
"model.layers.11.mlp_block.up_proj.weight": "model-00002-of-00004.safetensors",
"model.layers.11.temporal_block.k_proj.weight": "model-00002-of-00004.safetensors",
"model.layers.11.temporal_block.o_proj.bias": "model-00002-of-00004.safetensors",
"model.layers.11.temporal_block.o_proj.weight": "model-00002-of-00004.safetensors",
"model.layers.11.temporal_block.q_proj.weight": "model-00002-of-00004.safetensors",
"model.layers.11.temporal_block.v_proj.weight": "model-00002-of-00004.safetensors",
"model.layers.11.temporal_pre_norm.weight": "model-00002-of-00004.safetensors",
"model.layers.12.channel_pre_norm.weight": "model-00002-of-00004.safetensors",
"model.layers.12.mlp_block.down_proj.bias": "model-00002-of-00004.safetensors",
"model.layers.12.mlp_block.down_proj.weight": "model-00002-of-00004.safetensors",
"model.layers.12.mlp_block.gate_proj.bias": "model-00002-of-00004.safetensors",
"model.layers.12.mlp_block.gate_proj.weight": "model-00002-of-00004.safetensors",
"model.layers.12.mlp_block.up_proj.bias": "model-00002-of-00004.safetensors",
"model.layers.12.mlp_block.up_proj.weight": "model-00002-of-00004.safetensors",
"model.layers.12.temporal_block.conv_1d.bias": "model-00002-of-00004.safetensors",
"model.layers.12.temporal_block.conv_1d.weight": "model-00002-of-00004.safetensors",
"model.layers.12.temporal_block.linear_out.bias": "model-00002-of-00004.safetensors",
"model.layers.12.temporal_block.linear_out.weight": "model-00002-of-00004.safetensors",
"model.layers.12.temporal_block.linear_x.bias": "model-00002-of-00004.safetensors",
"model.layers.12.temporal_block.linear_x.weight": "model-00002-of-00004.safetensors",
"model.layers.12.temporal_block.linear_y.bias": "model-00002-of-00004.safetensors",
"model.layers.12.temporal_block.linear_y.weight": "model-00002-of-00004.safetensors",
"model.layers.12.temporal_block.rg_lru.input_gate_bias": "model-00002-of-00004.safetensors",
"model.layers.12.temporal_block.rg_lru.input_gate_weight": "model-00002-of-00004.safetensors",
"model.layers.12.temporal_block.rg_lru.recurrent_gate_bias": "model-00002-of-00004.safetensors",
"model.layers.12.temporal_block.rg_lru.recurrent_gate_weight": "model-00002-of-00004.safetensors",
"model.layers.12.temporal_block.rg_lru.recurrent_param": "model-00002-of-00004.safetensors",
"model.layers.12.temporal_pre_norm.weight": "model-00002-of-00004.safetensors",
"model.layers.13.channel_pre_norm.weight": "model-00002-of-00004.safetensors",
"model.layers.13.mlp_block.down_proj.bias": "model-00002-of-00004.safetensors",
"model.layers.13.mlp_block.down_proj.weight": "model-00002-of-00004.safetensors",
"model.layers.13.mlp_block.gate_proj.bias": "model-00002-of-00004.safetensors",
"model.layers.13.mlp_block.gate_proj.weight": "model-00002-of-00004.safetensors",
"model.layers.13.mlp_block.up_proj.bias": "model-00002-of-00004.safetensors",
"model.layers.13.mlp_block.up_proj.weight": "model-00002-of-00004.safetensors",
"model.layers.13.temporal_block.conv_1d.bias": "model-00002-of-00004.safetensors",
"model.layers.13.temporal_block.conv_1d.weight": "model-00002-of-00004.safetensors",
"model.layers.13.temporal_block.linear_out.bias": "model-00002-of-00004.safetensors",
"model.layers.13.temporal_block.linear_out.weight": "model-00002-of-00004.safetensors",
"model.layers.13.temporal_block.linear_x.bias": "model-00002-of-00004.safetensors",
"model.layers.13.temporal_block.linear_x.weight": "model-00002-of-00004.safetensors",
"model.layers.13.temporal_block.linear_y.bias": "model-00002-of-00004.safetensors",
"model.layers.13.temporal_block.linear_y.weight": "model-00002-of-00004.safetensors",
"model.layers.13.temporal_block.rg_lru.input_gate_bias": "model-00002-of-00004.safetensors",
"model.layers.13.temporal_block.rg_lru.input_gate_weight": "model-00002-of-00004.safetensors",
"model.layers.13.temporal_block.rg_lru.recurrent_gate_bias": "model-00002-of-00004.safetensors",
"model.layers.13.temporal_block.rg_lru.recurrent_gate_weight": "model-00002-of-00004.safetensors",
"model.layers.13.temporal_block.rg_lru.recurrent_param": "model-00002-of-00004.safetensors",
"model.layers.13.temporal_pre_norm.weight": "model-00002-of-00004.safetensors",
"model.layers.14.channel_pre_norm.weight": "model-00002-of-00004.safetensors",
"model.layers.14.mlp_block.down_proj.bias": "model-00002-of-00004.safetensors",
"model.layers.14.mlp_block.down_proj.weight": "model-00002-of-00004.safetensors",
"model.layers.14.mlp_block.gate_proj.bias": "model-00002-of-00004.safetensors",
"model.layers.14.mlp_block.gate_proj.weight": "model-00002-of-00004.safetensors",
"model.layers.14.mlp_block.up_proj.bias": "model-00002-of-00004.safetensors",
"model.layers.14.mlp_block.up_proj.weight": "model-00002-of-00004.safetensors",
"model.layers.14.temporal_block.k_proj.weight": "model-00002-of-00004.safetensors",
"model.layers.14.temporal_block.o_proj.bias": "model-00002-of-00004.safetensors",
"model.layers.14.temporal_block.o_proj.weight": "model-00002-of-00004.safetensors",
"model.layers.14.temporal_block.q_proj.weight": "model-00002-of-00004.safetensors",
"model.layers.14.temporal_block.v_proj.weight": "model-00002-of-00004.safetensors",
"model.layers.14.temporal_pre_norm.weight": "model-00002-of-00004.safetensors",
"model.layers.15.channel_pre_norm.weight": "model-00002-of-00004.safetensors",
"model.layers.15.mlp_block.down_proj.bias": "model-00002-of-00004.safetensors",
"model.layers.15.mlp_block.down_proj.weight": "model-00002-of-00004.safetensors",
"model.layers.15.mlp_block.gate_proj.bias": "model-00002-of-00004.safetensors",
"model.layers.15.mlp_block.gate_proj.weight": "model-00002-of-00004.safetensors",
"model.layers.15.mlp_block.up_proj.bias": "model-00002-of-00004.safetensors",
"model.layers.15.mlp_block.up_proj.weight": "model-00002-of-00004.safetensors",
"model.layers.15.temporal_block.conv_1d.bias": "model-00002-of-00004.safetensors",
"model.layers.15.temporal_block.conv_1d.weight": "model-00002-of-00004.safetensors",
"model.layers.15.temporal_block.linear_out.bias": "model-00002-of-00004.safetensors",
"model.layers.15.temporal_block.linear_out.weight": "model-00002-of-00004.safetensors",
"model.layers.15.temporal_block.linear_x.bias": "model-00002-of-00004.safetensors",
"model.layers.15.temporal_block.linear_x.weight": "model-00002-of-00004.safetensors",
"model.layers.15.temporal_block.linear_y.bias": "model-00002-of-00004.safetensors",
"model.layers.15.temporal_block.linear_y.weight": "model-00002-of-00004.safetensors",
"model.layers.15.temporal_block.rg_lru.input_gate_bias": "model-00002-of-00004.safetensors",
"model.layers.15.temporal_block.rg_lru.input_gate_weight": "model-00002-of-00004.safetensors",
"model.layers.15.temporal_block.rg_lru.recurrent_gate_bias": "model-00002-of-00004.safetensors",
"model.layers.15.temporal_block.rg_lru.recurrent_gate_weight": "model-00002-of-00004.safetensors",
"model.layers.15.temporal_block.rg_lru.recurrent_param": "model-00002-of-00004.safetensors",
"model.layers.15.temporal_pre_norm.weight": "model-00002-of-00004.safetensors",
"model.layers.16.channel_pre_norm.weight": "model-00002-of-00004.safetensors",
"model.layers.16.mlp_block.down_proj.bias": "model-00002-of-00004.safetensors",
"model.layers.16.mlp_block.down_proj.weight": "model-00002-of-00004.safetensors",
"model.layers.16.mlp_block.gate_proj.bias": "model-00002-of-00004.safetensors",
"model.layers.16.mlp_block.gate_proj.weight": "model-00002-of-00004.safetensors",
"model.layers.16.mlp_block.up_proj.bias": "model-00002-of-00004.safetensors",
"model.layers.16.mlp_block.up_proj.weight": "model-00002-of-00004.safetensors",
"model.layers.16.temporal_block.conv_1d.bias": "model-00002-of-00004.safetensors",
"model.layers.16.temporal_block.conv_1d.weight": "model-00002-of-00004.safetensors",
"model.layers.16.temporal_block.linear_out.bias": "model-00002-of-00004.safetensors",
"model.layers.16.temporal_block.linear_out.weight": "model-00002-of-00004.safetensors",
"model.layers.16.temporal_block.linear_x.bias": "model-00002-of-00004.safetensors",
"model.layers.16.temporal_block.linear_x.weight": "model-00002-of-00004.safetensors",
"model.layers.16.temporal_block.linear_y.bias": "model-00002-of-00004.safetensors",
"model.layers.16.temporal_block.linear_y.weight": "model-00002-of-00004.safetensors",
"model.layers.16.temporal_block.rg_lru.input_gate_bias": "model-00002-of-00004.safetensors",
"model.layers.16.temporal_block.rg_lru.input_gate_weight": "model-00002-of-00004.safetensors",
"model.layers.16.temporal_block.rg_lru.recurrent_gate_bias": "model-00002-of-00004.safetensors",
"model.layers.16.temporal_block.rg_lru.recurrent_gate_weight": "model-00002-of-00004.safetensors",
"model.layers.16.temporal_block.rg_lru.recurrent_param": "model-00002-of-00004.safetensors",
"model.layers.16.temporal_pre_norm.weight": "model-00002-of-00004.safetensors",
"model.layers.17.channel_pre_norm.weight": "model-00002-of-00004.safetensors",
"model.layers.17.mlp_block.down_proj.bias": "model-00002-of-00004.safetensors",
"model.layers.17.mlp_block.down_proj.weight": "model-00002-of-00004.safetensors",
"model.layers.17.mlp_block.gate_proj.bias": "model-00002-of-00004.safetensors",
"model.layers.17.mlp_block.gate_proj.weight": "model-00002-of-00004.safetensors",
"model.layers.17.mlp_block.up_proj.bias": "model-00002-of-00004.safetensors",
"model.layers.17.mlp_block.up_proj.weight": "model-00002-of-00004.safetensors",
"model.layers.17.temporal_block.k_proj.weight": "model-00002-of-00004.safetensors",
"model.layers.17.temporal_block.o_proj.bias": "model-00002-of-00004.safetensors",
"model.layers.17.temporal_block.o_proj.weight": "model-00002-of-00004.safetensors",
"model.layers.17.temporal_block.q_proj.weight": "model-00002-of-00004.safetensors",
"model.layers.17.temporal_block.v_proj.weight": "model-00002-of-00004.safetensors",
"model.layers.17.temporal_pre_norm.weight": "model-00002-of-00004.safetensors",
"model.layers.18.channel_pre_norm.weight": "model-00002-of-00004.safetensors",
"model.layers.18.mlp_block.down_proj.bias": "model-00002-of-00004.safetensors",
"model.layers.18.mlp_block.down_proj.weight": "model-00002-of-00004.safetensors",
"model.layers.18.mlp_block.gate_proj.bias": "model-00002-of-00004.safetensors",
"model.layers.18.mlp_block.gate_proj.weight": "model-00002-of-00004.safetensors",
"model.layers.18.mlp_block.up_proj.bias": "model-00002-of-00004.safetensors",
"model.layers.18.mlp_block.up_proj.weight": "model-00002-of-00004.safetensors",
"model.layers.18.temporal_block.conv_1d.bias": "model-00002-of-00004.safetensors",
"model.layers.18.temporal_block.conv_1d.weight": "model-00002-of-00004.safetensors",
"model.layers.18.temporal_block.linear_out.bias": "model-00002-of-00004.safetensors",
"model.layers.18.temporal_block.linear_out.weight": "model-00002-of-00004.safetensors",
"model.layers.18.temporal_block.linear_x.bias": "model-00002-of-00004.safetensors",
"model.layers.18.temporal_block.linear_x.weight": "model-00002-of-00004.safetensors",
"model.layers.18.temporal_block.linear_y.bias": "model-00002-of-00004.safetensors",
"model.layers.18.temporal_block.linear_y.weight": "model-00002-of-00004.safetensors",
"model.layers.18.temporal_block.rg_lru.input_gate_bias": "model-00002-of-00004.safetensors",
"model.layers.18.temporal_block.rg_lru.input_gate_weight": "model-00002-of-00004.safetensors",
"model.layers.18.temporal_block.rg_lru.recurrent_gate_bias": "model-00002-of-00004.safetensors",
"model.layers.18.temporal_block.rg_lru.recurrent_gate_weight": "model-00002-of-00004.safetensors",
"model.layers.18.temporal_block.rg_lru.recurrent_param": "model-00002-of-00004.safetensors",
"model.layers.18.temporal_pre_norm.weight": "model-00002-of-00004.safetensors",
"model.layers.19.channel_pre_norm.weight": "model-00002-of-00004.safetensors",
"model.layers.19.mlp_block.down_proj.bias": "model-00003-of-00004.safetensors",
"model.layers.19.mlp_block.down_proj.weight": "model-00003-of-00004.safetensors",
"model.layers.19.mlp_block.gate_proj.bias": "model-00002-of-00004.safetensors",
"model.layers.19.mlp_block.gate_proj.weight": "model-00002-of-00004.safetensors",
"model.layers.19.mlp_block.up_proj.bias": "model-00002-of-00004.safetensors",
"model.layers.19.mlp_block.up_proj.weight": "model-00002-of-00004.safetensors",
"model.layers.19.temporal_block.conv_1d.bias": "model-00002-of-00004.safetensors",
"model.layers.19.temporal_block.conv_1d.weight": "model-00002-of-00004.safetensors",
"model.layers.19.temporal_block.linear_out.bias": "model-00002-of-00004.safetensors",
"model.layers.19.temporal_block.linear_out.weight": "model-00002-of-00004.safetensors",
"model.layers.19.temporal_block.linear_x.bias": "model-00002-of-00004.safetensors",
"model.layers.19.temporal_block.linear_x.weight": "model-00002-of-00004.safetensors",
"model.layers.19.temporal_block.linear_y.bias": "model-00002-of-00004.safetensors",
"model.layers.19.temporal_block.linear_y.weight": "model-00002-of-00004.safetensors",
"model.layers.19.temporal_block.rg_lru.input_gate_bias": "model-00002-of-00004.safetensors",
"model.layers.19.temporal_block.rg_lru.input_gate_weight": "model-00002-of-00004.safetensors",
"model.layers.19.temporal_block.rg_lru.recurrent_gate_bias": "model-00002-of-00004.safetensors",
"model.layers.19.temporal_block.rg_lru.recurrent_gate_weight": "model-00002-of-00004.safetensors",
"model.layers.19.temporal_block.rg_lru.recurrent_param": "model-00002-of-00004.safetensors",
"model.layers.19.temporal_pre_norm.weight": "model-00002-of-00004.safetensors",
"model.layers.2.channel_pre_norm.weight": "model-00001-of-00004.safetensors",
"model.layers.2.mlp_block.down_proj.bias": "model-00001-of-00004.safetensors",
"model.layers.2.mlp_block.down_proj.weight": "model-00001-of-00004.safetensors",
"model.layers.2.mlp_block.gate_proj.bias": "model-00001-of-00004.safetensors",
"model.layers.2.mlp_block.gate_proj.weight": "model-00001-of-00004.safetensors",
"model.layers.2.mlp_block.up_proj.bias": "model-00001-of-00004.safetensors",
"model.layers.2.mlp_block.up_proj.weight": "model-00001-of-00004.safetensors",
"model.layers.2.temporal_block.k_proj.weight": "model-00001-of-00004.safetensors",
"model.layers.2.temporal_block.o_proj.bias": "model-00001-of-00004.safetensors",
"model.layers.2.temporal_block.o_proj.weight": "model-00001-of-00004.safetensors",
"model.layers.2.temporal_block.q_proj.weight": "model-00001-of-00004.safetensors",
"model.layers.2.temporal_block.v_proj.weight": "model-00001-of-00004.safetensors",
"model.layers.2.temporal_pre_norm.weight": "model-00001-of-00004.safetensors",
"model.layers.20.channel_pre_norm.weight": "model-00003-of-00004.safetensors",
"model.layers.20.mlp_block.down_proj.bias": "model-00003-of-00004.safetensors",
"model.layers.20.mlp_block.down_proj.weight": "model-00003-of-00004.safetensors",
"model.layers.20.mlp_block.gate_proj.bias": "model-00003-of-00004.safetensors",
"model.layers.20.mlp_block.gate_proj.weight": "model-00003-of-00004.safetensors",
"model.layers.20.mlp_block.up_proj.bias": "model-00003-of-00004.safetensors",
"model.layers.20.mlp_block.up_proj.weight": "model-00003-of-00004.safetensors",
"model.layers.20.temporal_block.k_proj.weight": "model-00003-of-00004.safetensors",
"model.layers.20.temporal_block.o_proj.bias": "model-00003-of-00004.safetensors",
"model.layers.20.temporal_block.o_proj.weight": "model-00003-of-00004.safetensors",
"model.layers.20.temporal_block.q_proj.weight": "model-00003-of-00004.safetensors",
"model.layers.20.temporal_block.v_proj.weight": "model-00003-of-00004.safetensors",
"model.layers.20.temporal_pre_norm.weight": "model-00003-of-00004.safetensors",
"model.layers.21.channel_pre_norm.weight": "model-00003-of-00004.safetensors",
"model.layers.21.mlp_block.down_proj.bias": "model-00003-of-00004.safetensors",
"model.layers.21.mlp_block.down_proj.weight": "model-00003-of-00004.safetensors",
"model.layers.21.mlp_block.gate_proj.bias": "model-00003-of-00004.safetensors",
"model.layers.21.mlp_block.gate_proj.weight": "model-00003-of-00004.safetensors",
"model.layers.21.mlp_block.up_proj.bias": "model-00003-of-00004.safetensors",
"model.layers.21.mlp_block.up_proj.weight": "model-00003-of-00004.safetensors",
"model.layers.21.temporal_block.conv_1d.bias": "model-00003-of-00004.safetensors",
"model.layers.21.temporal_block.conv_1d.weight": "model-00003-of-00004.safetensors",
"model.layers.21.temporal_block.linear_out.bias": "model-00003-of-00004.safetensors",
"model.layers.21.temporal_block.linear_out.weight": "model-00003-of-00004.safetensors",
"model.layers.21.temporal_block.linear_x.bias": "model-00003-of-00004.safetensors",
"model.layers.21.temporal_block.linear_x.weight": "model-00003-of-00004.safetensors",
"model.layers.21.temporal_block.linear_y.bias": "model-00003-of-00004.safetensors",
"model.layers.21.temporal_block.linear_y.weight": "model-00003-of-00004.safetensors",
"model.layers.21.temporal_block.rg_lru.input_gate_bias": "model-00003-of-00004.safetensors",
"model.layers.21.temporal_block.rg_lru.input_gate_weight": "model-00003-of-00004.safetensors",
"model.layers.21.temporal_block.rg_lru.recurrent_gate_bias": "model-00003-of-00004.safetensors",
"model.layers.21.temporal_block.rg_lru.recurrent_gate_weight": "model-00003-of-00004.safetensors",
"model.layers.21.temporal_block.rg_lru.recurrent_param": "model-00003-of-00004.safetensors",
"model.layers.21.temporal_pre_norm.weight": "model-00003-of-00004.safetensors",
"model.layers.22.channel_pre_norm.weight": "model-00003-of-00004.safetensors",
"model.layers.22.mlp_block.down_proj.bias": "model-00003-of-00004.safetensors",
"model.layers.22.mlp_block.down_proj.weight": "model-00003-of-00004.safetensors",
"model.layers.22.mlp_block.gate_proj.bias": "model-00003-of-00004.safetensors",
"model.layers.22.mlp_block.gate_proj.weight": "model-00003-of-00004.safetensors",
"model.layers.22.mlp_block.up_proj.bias": "model-00003-of-00004.safetensors",
"model.layers.22.mlp_block.up_proj.weight": "model-00003-of-00004.safetensors",
"model.layers.22.temporal_block.conv_1d.bias": "model-00003-of-00004.safetensors",
"model.layers.22.temporal_block.conv_1d.weight": "model-00003-of-00004.safetensors",
"model.layers.22.temporal_block.linear_out.bias": "model-00003-of-00004.safetensors",
"model.layers.22.temporal_block.linear_out.weight": "model-00003-of-00004.safetensors",
"model.layers.22.temporal_block.linear_x.bias": "model-00003-of-00004.safetensors",
"model.layers.22.temporal_block.linear_x.weight": "model-00003-of-00004.safetensors",
"model.layers.22.temporal_block.linear_y.bias": "model-00003-of-00004.safetensors",
"model.layers.22.temporal_block.linear_y.weight": "model-00003-of-00004.safetensors",
"model.layers.22.temporal_block.rg_lru.input_gate_bias": "model-00003-of-00004.safetensors",
"model.layers.22.temporal_block.rg_lru.input_gate_weight": "model-00003-of-00004.safetensors",
"model.layers.22.temporal_block.rg_lru.recurrent_gate_bias": "model-00003-of-00004.safetensors",
"model.layers.22.temporal_block.rg_lru.recurrent_gate_weight": "model-00003-of-00004.safetensors",
"model.layers.22.temporal_block.rg_lru.recurrent_param": "model-00003-of-00004.safetensors",
"model.layers.22.temporal_pre_norm.weight": "model-00003-of-00004.safetensors",
"model.layers.23.channel_pre_norm.weight": "model-00003-of-00004.safetensors",
"model.layers.23.mlp_block.down_proj.bias": "model-00003-of-00004.safetensors",
"model.layers.23.mlp_block.down_proj.weight": "model-00003-of-00004.safetensors",
"model.layers.23.mlp_block.gate_proj.bias": "model-00003-of-00004.safetensors",
"model.layers.23.mlp_block.gate_proj.weight": "model-00003-of-00004.safetensors",
"model.layers.23.mlp_block.up_proj.bias": "model-00003-of-00004.safetensors",
"model.layers.23.mlp_block.up_proj.weight": "model-00003-of-00004.safetensors",
"model.layers.23.temporal_block.k_proj.weight": "model-00003-of-00004.safetensors",
"model.layers.23.temporal_block.o_proj.bias": "model-00003-of-00004.safetensors",
"model.layers.23.temporal_block.o_proj.weight": "model-00003-of-00004.safetensors",
"model.layers.23.temporal_block.q_proj.weight": "model-00003-of-00004.safetensors",
"model.layers.23.temporal_block.v_proj.weight": "model-00003-of-00004.safetensors",
"model.layers.23.temporal_pre_norm.weight": "model-00003-of-00004.safetensors",
"model.layers.24.channel_pre_norm.weight": "model-00003-of-00004.safetensors",
"model.layers.24.mlp_block.down_proj.bias": "model-00003-of-00004.safetensors",
"model.layers.24.mlp_block.down_proj.weight": "model-00003-of-00004.safetensors",
"model.layers.24.mlp_block.gate_proj.bias": "model-00003-of-00004.safetensors",
"model.layers.24.mlp_block.gate_proj.weight": "model-00003-of-00004.safetensors",
"model.layers.24.mlp_block.up_proj.bias": "model-00003-of-00004.safetensors",
"model.layers.24.mlp_block.up_proj.weight": "model-00003-of-00004.safetensors",
"model.layers.24.temporal_block.conv_1d.bias": "model-00003-of-00004.safetensors",
"model.layers.24.temporal_block.conv_1d.weight": "model-00003-of-00004.safetensors",
"model.layers.24.temporal_block.linear_out.bias": "model-00003-of-00004.safetensors",
"model.layers.24.temporal_block.linear_out.weight": "model-00003-of-00004.safetensors",
"model.layers.24.temporal_block.linear_x.bias": "model-00003-of-00004.safetensors",
"model.layers.24.temporal_block.linear_x.weight": "model-00003-of-00004.safetensors",
"model.layers.24.temporal_block.linear_y.bias": "model-00003-of-00004.safetensors",
"model.layers.24.temporal_block.linear_y.weight": "model-00003-of-00004.safetensors",
"model.layers.24.temporal_block.rg_lru.input_gate_bias": "model-00003-of-00004.safetensors",
"model.layers.24.temporal_block.rg_lru.input_gate_weight": "model-00003-of-00004.safetensors",
"model.layers.24.temporal_block.rg_lru.recurrent_gate_bias": "model-00003-of-00004.safetensors",
"model.layers.24.temporal_block.rg_lru.recurrent_gate_weight": "model-00003-of-00004.safetensors",
"model.layers.24.temporal_block.rg_lru.recurrent_param": "model-00003-of-00004.safetensors",
"model.layers.24.temporal_pre_norm.weight": "model-00003-of-00004.safetensors",
"model.layers.25.channel_pre_norm.weight": "model-00003-of-00004.safetensors",
"model.layers.25.mlp_block.down_proj.bias": "model-00003-of-00004.safetensors",
"model.layers.25.mlp_block.down_proj.weight": "model-00003-of-00004.safetensors",
"model.layers.25.mlp_block.gate_proj.bias": "model-00003-of-00004.safetensors",
"model.layers.25.mlp_block.gate_proj.weight": "model-00003-of-00004.safetensors",
"model.layers.25.mlp_block.up_proj.bias": "model-00003-of-00004.safetensors",
"model.layers.25.mlp_block.up_proj.weight": "model-00003-of-00004.safetensors",
"model.layers.25.temporal_block.conv_1d.bias": "model-00003-of-00004.safetensors",
"model.layers.25.temporal_block.conv_1d.weight": "model-00003-of-00004.safetensors",
"model.layers.25.temporal_block.linear_out.bias": "model-00003-of-00004.safetensors",
"model.layers.25.temporal_block.linear_out.weight": "model-00003-of-00004.safetensors",
"model.layers.25.temporal_block.linear_x.bias": "model-00003-of-00004.safetensors",
"model.layers.25.temporal_block.linear_x.weight": "model-00003-of-00004.safetensors",
"model.layers.25.temporal_block.linear_y.bias": "model-00003-of-00004.safetensors",
"model.layers.25.temporal_block.linear_y.weight": "model-00003-of-00004.safetensors",
"model.layers.25.temporal_block.rg_lru.input_gate_bias": "model-00003-of-00004.safetensors",
"model.layers.25.temporal_block.rg_lru.input_gate_weight": "model-00003-of-00004.safetensors",
"model.layers.25.temporal_block.rg_lru.recurrent_gate_bias": "model-00003-of-00004.safetensors",
"model.layers.25.temporal_block.rg_lru.recurrent_gate_weight": "model-00003-of-00004.safetensors",
"model.layers.25.temporal_block.rg_lru.recurrent_param": "model-00003-of-00004.safetensors",
"model.layers.25.temporal_pre_norm.weight": "model-00003-of-00004.safetensors",
"model.layers.26.channel_pre_norm.weight": "model-00003-of-00004.safetensors",
"model.layers.26.mlp_block.down_proj.bias": "model-00003-of-00004.safetensors",
"model.layers.26.mlp_block.down_proj.weight": "model-00003-of-00004.safetensors",
"model.layers.26.mlp_block.gate_proj.bias": "model-00003-of-00004.safetensors",
"model.layers.26.mlp_block.gate_proj.weight": "model-00003-of-00004.safetensors",
"model.layers.26.mlp_block.up_proj.bias": "model-00003-of-00004.safetensors",
"model.layers.26.mlp_block.up_proj.weight": "model-00003-of-00004.safetensors",
"model.layers.26.temporal_block.k_proj.weight": "model-00003-of-00004.safetensors",
"model.layers.26.temporal_block.o_proj.bias": "model-00003-of-00004.safetensors",
"model.layers.26.temporal_block.o_proj.weight": "model-00003-of-00004.safetensors",
"model.layers.26.temporal_block.q_proj.weight": "model-00003-of-00004.safetensors",
"model.layers.26.temporal_block.v_proj.weight": "model-00003-of-00004.safetensors",
"model.layers.26.temporal_pre_norm.weight": "model-00003-of-00004.safetensors",
"model.layers.27.channel_pre_norm.weight": "model-00003-of-00004.safetensors",
"model.layers.27.mlp_block.down_proj.bias": "model-00003-of-00004.safetensors",
"model.layers.27.mlp_block.down_proj.weight": "model-00003-of-00004.safetensors",
"model.layers.27.mlp_block.gate_proj.bias": "model-00003-of-00004.safetensors",
"model.layers.27.mlp_block.gate_proj.weight": "model-00003-of-00004.safetensors",
"model.layers.27.mlp_block.up_proj.bias": "model-00003-of-00004.safetensors",
"model.layers.27.mlp_block.up_proj.weight": "model-00003-of-00004.safetensors",
"model.layers.27.temporal_block.conv_1d.bias": "model-00003-of-00004.safetensors",
"model.layers.27.temporal_block.conv_1d.weight": "model-00003-of-00004.safetensors",
"model.layers.27.temporal_block.linear_out.bias": "model-00003-of-00004.safetensors",
"model.layers.27.temporal_block.linear_out.weight": "model-00003-of-00004.safetensors",
"model.layers.27.temporal_block.linear_x.bias": "model-00003-of-00004.safetensors",
"model.layers.27.temporal_block.linear_x.weight": "model-00003-of-00004.safetensors",
"model.layers.27.temporal_block.linear_y.bias": "model-00003-of-00004.safetensors",
"model.layers.27.temporal_block.linear_y.weight": "model-00003-of-00004.safetensors",
"model.layers.27.temporal_block.rg_lru.input_gate_bias": "model-00003-of-00004.safetensors",
"model.layers.27.temporal_block.rg_lru.input_gate_weight": "model-00003-of-00004.safetensors",
"model.layers.27.temporal_block.rg_lru.recurrent_gate_bias": "model-00003-of-00004.safetensors",
"model.layers.27.temporal_block.rg_lru.recurrent_gate_weight": "model-00003-of-00004.safetensors",
"model.layers.27.temporal_block.rg_lru.recurrent_param": "model-00003-of-00004.safetensors",
"model.layers.27.temporal_pre_norm.weight": "model-00003-of-00004.safetensors",
"model.layers.28.channel_pre_norm.weight": "model-00003-of-00004.safetensors",
"model.layers.28.mlp_block.down_proj.bias": "model-00003-of-00004.safetensors",
"model.layers.28.mlp_block.down_proj.weight": "model-00003-of-00004.safetensors",
"model.layers.28.mlp_block.gate_proj.bias": "model-00003-of-00004.safetensors",
"model.layers.28.mlp_block.gate_proj.weight": "model-00003-of-00004.safetensors",
"model.layers.28.mlp_block.up_proj.bias": "model-00003-of-00004.safetensors",
"model.layers.28.mlp_block.up_proj.weight": "model-00003-of-00004.safetensors",
"model.layers.28.temporal_block.conv_1d.bias": "model-00003-of-00004.safetensors",
"model.layers.28.temporal_block.conv_1d.weight": "model-00003-of-00004.safetensors",
"model.layers.28.temporal_block.linear_out.bias": "model-00003-of-00004.safetensors",
"model.layers.28.temporal_block.linear_out.weight": "model-00003-of-00004.safetensors",
"model.layers.28.temporal_block.linear_x.bias": "model-00003-of-00004.safetensors",
"model.layers.28.temporal_block.linear_x.weight": "model-00003-of-00004.safetensors",
"model.layers.28.temporal_block.linear_y.bias": "model-00003-of-00004.safetensors",
"model.layers.28.temporal_block.linear_y.weight": "model-00003-of-00004.safetensors",
"model.layers.28.temporal_block.rg_lru.input_gate_bias": "model-00003-of-00004.safetensors",
"model.layers.28.temporal_block.rg_lru.input_gate_weight": "model-00003-of-00004.safetensors",
"model.layers.28.temporal_block.rg_lru.recurrent_gate_bias": "model-00003-of-00004.safetensors",
"model.layers.28.temporal_block.rg_lru.recurrent_gate_weight": "model-00003-of-00004.safetensors",
"model.layers.28.temporal_block.rg_lru.recurrent_param": "model-00003-of-00004.safetensors",
"model.layers.28.temporal_pre_norm.weight": "model-00003-of-00004.safetensors",
"model.layers.29.channel_pre_norm.weight": "model-00003-of-00004.safetensors",
"model.layers.29.mlp_block.down_proj.bias": "model-00003-of-00004.safetensors",
"model.layers.29.mlp_block.down_proj.weight": "model-00003-of-00004.safetensors",
"model.layers.29.mlp_block.gate_proj.bias": "model-00003-of-00004.safetensors",
"model.layers.29.mlp_block.gate_proj.weight": "model-00003-of-00004.safetensors",
"model.layers.29.mlp_block.up_proj.bias": "model-00003-of-00004.safetensors",
"model.layers.29.mlp_block.up_proj.weight": "model-00003-of-00004.safetensors",
"model.layers.29.temporal_block.k_proj.weight": "model-00003-of-00004.safetensors",
"model.layers.29.temporal_block.o_proj.bias": "model-00003-of-00004.safetensors",
"model.layers.29.temporal_block.o_proj.weight": "model-00003-of-00004.safetensors",
"model.layers.29.temporal_block.q_proj.weight": "model-00003-of-00004.safetensors",
"model.layers.29.temporal_block.v_proj.weight": "model-00003-of-00004.safetensors",
"model.layers.29.temporal_pre_norm.weight": "model-00003-of-00004.safetensors",
"model.layers.3.channel_pre_norm.weight": "model-00001-of-00004.safetensors",
"model.layers.3.mlp_block.down_proj.bias": "model-00001-of-00004.safetensors",
"model.layers.3.mlp_block.down_proj.weight": "model-00001-of-00004.safetensors",
"model.layers.3.mlp_block.gate_proj.bias": "model-00001-of-00004.safetensors",
"model.layers.3.mlp_block.gate_proj.weight": "model-00001-of-00004.safetensors",
"model.layers.3.mlp_block.up_proj.bias": "model-00001-of-00004.safetensors",
"model.layers.3.mlp_block.up_proj.weight": "model-00001-of-00004.safetensors",
"model.layers.3.temporal_block.conv_1d.bias": "model-00001-of-00004.safetensors",
"model.layers.3.temporal_block.conv_1d.weight": "model-00001-of-00004.safetensors",
"model.layers.3.temporal_block.linear_out.bias": "model-00001-of-00004.safetensors",
"model.layers.3.temporal_block.linear_out.weight": "model-00001-of-00004.safetensors",
"model.layers.3.temporal_block.linear_x.bias": "model-00001-of-00004.safetensors",
"model.layers.3.temporal_block.linear_x.weight": "model-00001-of-00004.safetensors",
"model.layers.3.temporal_block.linear_y.bias": "model-00001-of-00004.safetensors",
"model.layers.3.temporal_block.linear_y.weight": "model-00001-of-00004.safetensors",
"model.layers.3.temporal_block.rg_lru.input_gate_bias": "model-00001-of-00004.safetensors",
"model.layers.3.temporal_block.rg_lru.input_gate_weight": "model-00001-of-00004.safetensors",
"model.layers.3.temporal_block.rg_lru.recurrent_gate_bias": "model-00001-of-00004.safetensors",
"model.layers.3.temporal_block.rg_lru.recurrent_gate_weight": "model-00001-of-00004.safetensors",
"model.layers.3.temporal_block.rg_lru.recurrent_param": "model-00001-of-00004.safetensors",
"model.layers.3.temporal_pre_norm.weight": "model-00001-of-00004.safetensors",
"model.layers.30.channel_pre_norm.weight": "model-00003-of-00004.safetensors",
"model.layers.30.mlp_block.down_proj.bias": "model-00003-of-00004.safetensors",
"model.layers.30.mlp_block.down_proj.weight": "model-00003-of-00004.safetensors",
"model.layers.30.mlp_block.gate_proj.bias": "model-00003-of-00004.safetensors",
"model.layers.30.mlp_block.gate_proj.weight": "model-00003-of-00004.safetensors",
"model.layers.30.mlp_block.up_proj.bias": "model-00003-of-00004.safetensors",
"model.layers.30.mlp_block.up_proj.weight": "model-00003-of-00004.safetensors",
"model.layers.30.temporal_block.conv_1d.bias": "model-00003-of-00004.safetensors",
"model.layers.30.temporal_block.conv_1d.weight": "model-00003-of-00004.safetensors",
"model.layers.30.temporal_block.linear_out.bias": "model-00003-of-00004.safetensors",
"model.layers.30.temporal_block.linear_out.weight": "model-00003-of-00004.safetensors",
"model.layers.30.temporal_block.linear_x.bias": "model-00003-of-00004.safetensors",
"model.layers.30.temporal_block.linear_x.weight": "model-00003-of-00004.safetensors",
"model.layers.30.temporal_block.linear_y.bias": "model-00003-of-00004.safetensors",
"model.layers.30.temporal_block.linear_y.weight": "model-00003-of-00004.safetensors",
"model.layers.30.temporal_block.rg_lru.input_gate_bias": "model-00003-of-00004.safetensors",
"model.layers.30.temporal_block.rg_lru.input_gate_weight": "model-00003-of-00004.safetensors",
"model.layers.30.temporal_block.rg_lru.recurrent_gate_bias": "model-00003-of-00004.safetensors",
"model.layers.30.temporal_block.rg_lru.recurrent_gate_weight": "model-00003-of-00004.safetensors",
"model.layers.30.temporal_block.rg_lru.recurrent_param": "model-00003-of-00004.safetensors",
"model.layers.30.temporal_pre_norm.weight": "model-00003-of-00004.safetensors",
"model.layers.31.channel_pre_norm.weight": "model-00003-of-00004.safetensors",
"model.layers.31.mlp_block.down_proj.bias": "model-00003-of-00004.safetensors",
"model.layers.31.mlp_block.down_proj.weight": "model-00003-of-00004.safetensors",
"model.layers.31.mlp_block.gate_proj.bias": "model-00003-of-00004.safetensors",
"model.layers.31.mlp_block.gate_proj.weight": "model-00003-of-00004.safetensors",
"model.layers.31.mlp_block.up_proj.bias": "model-00003-of-00004.safetensors",
"model.layers.31.mlp_block.up_proj.weight": "model-00003-of-00004.safetensors",
"model.layers.31.temporal_block.conv_1d.bias": "model-00003-of-00004.safetensors",
"model.layers.31.temporal_block.conv_1d.weight": "model-00003-of-00004.safetensors",
"model.layers.31.temporal_block.linear_out.bias": "model-00003-of-00004.safetensors",
"model.layers.31.temporal_block.linear_out.weight": "model-00003-of-00004.safetensors",
"model.layers.31.temporal_block.linear_x.bias": "model-00003-of-00004.safetensors",
"model.layers.31.temporal_block.linear_x.weight": "model-00003-of-00004.safetensors",
"model.layers.31.temporal_block.linear_y.bias": "model-00003-of-00004.safetensors",
"model.layers.31.temporal_block.linear_y.weight": "model-00003-of-00004.safetensors",
"model.layers.31.temporal_block.rg_lru.input_gate_bias": "model-00003-of-00004.safetensors",
"model.layers.31.temporal_block.rg_lru.input_gate_weight": "model-00003-of-00004.safetensors",
"model.layers.31.temporal_block.rg_lru.recurrent_gate_bias": "model-00003-of-00004.safetensors",
"model.layers.31.temporal_block.rg_lru.recurrent_gate_weight": "model-00003-of-00004.safetensors",
"model.layers.31.temporal_block.rg_lru.recurrent_param": "model-00003-of-00004.safetensors",
"model.layers.31.temporal_pre_norm.weight": "model-00003-of-00004.safetensors",
"model.layers.32.channel_pre_norm.weight": "model-00003-of-00004.safetensors",
"model.layers.32.mlp_block.down_proj.bias": "model-00004-of-00004.safetensors",
"model.layers.32.mlp_block.down_proj.weight": "model-00004-of-00004.safetensors",
"model.layers.32.mlp_block.gate_proj.bias": "model-00004-of-00004.safetensors",
"model.layers.32.mlp_block.gate_proj.weight": "model-00004-of-00004.safetensors",
"model.layers.32.mlp_block.up_proj.bias": "model-00004-of-00004.safetensors",
"model.layers.32.mlp_block.up_proj.weight": "model-00004-of-00004.safetensors",
"model.layers.32.temporal_block.k_proj.weight": "model-00003-of-00004.safetensors",
"model.layers.32.temporal_block.o_proj.bias": "model-00003-of-00004.safetensors",
"model.layers.32.temporal_block.o_proj.weight": "model-00003-of-00004.safetensors",
"model.layers.32.temporal_block.q_proj.weight": "model-00003-of-00004.safetensors",
"model.layers.32.temporal_block.v_proj.weight": "model-00003-of-00004.safetensors",
"model.layers.32.temporal_pre_norm.weight": "model-00003-of-00004.safetensors",
"model.layers.33.channel_pre_norm.weight": "model-00004-of-00004.safetensors",
"model.layers.33.mlp_block.down_proj.bias": "model-00004-of-00004.safetensors",
"model.layers.33.mlp_block.down_proj.weight": "model-00004-of-00004.safetensors",
"model.layers.33.mlp_block.gate_proj.bias": "model-00004-of-00004.safetensors",
"model.layers.33.mlp_block.gate_proj.weight": "model-00004-of-00004.safetensors",
"model.layers.33.mlp_block.up_proj.bias": "model-00004-of-00004.safetensors",
"model.layers.33.mlp_block.up_proj.weight": "model-00004-of-00004.safetensors",
"model.layers.33.temporal_block.conv_1d.bias": "model-00004-of-00004.safetensors",
"model.layers.33.temporal_block.conv_1d.weight": "model-00004-of-00004.safetensors",
"model.layers.33.temporal_block.linear_out.bias": "model-00004-of-00004.safetensors",
"model.layers.33.temporal_block.linear_out.weight": "model-00004-of-00004.safetensors",
"model.layers.33.temporal_block.linear_x.bias": "model-00004-of-00004.safetensors",
"model.layers.33.temporal_block.linear_x.weight": "model-00004-of-00004.safetensors",
"model.layers.33.temporal_block.linear_y.bias": "model-00004-of-00004.safetensors",
"model.layers.33.temporal_block.linear_y.weight": "model-00004-of-00004.safetensors",
"model.layers.33.temporal_block.rg_lru.input_gate_bias": "model-00004-of-00004.safetensors",
"model.layers.33.temporal_block.rg_lru.input_gate_weight": "model-00004-of-00004.safetensors",
"model.layers.33.temporal_block.rg_lru.recurrent_gate_bias": "model-00004-of-00004.safetensors",
"model.layers.33.temporal_block.rg_lru.recurrent_gate_weight": "model-00004-of-00004.safetensors",
"model.layers.33.temporal_block.rg_lru.recurrent_param": "model-00004-of-00004.safetensors",
"model.layers.33.temporal_pre_norm.weight": "model-00004-of-00004.safetensors",
"model.layers.34.channel_pre_norm.weight": "model-00004-of-00004.safetensors",
"model.layers.34.mlp_block.down_proj.bias": "model-00004-of-00004.safetensors",
"model.layers.34.mlp_block.down_proj.weight": "model-00004-of-00004.safetensors",
"model.layers.34.mlp_block.gate_proj.bias": "model-00004-of-00004.safetensors",
"model.layers.34.mlp_block.gate_proj.weight": "model-00004-of-00004.safetensors",
"model.layers.34.mlp_block.up_proj.bias": "model-00004-of-00004.safetensors",
"model.layers.34.mlp_block.up_proj.weight": "model-00004-of-00004.safetensors",
"model.layers.34.temporal_block.conv_1d.bias": "model-00004-of-00004.safetensors",
"model.layers.34.temporal_block.conv_1d.weight": "model-00004-of-00004.safetensors",
"model.layers.34.temporal_block.linear_out.bias": "model-00004-of-00004.safetensors",
"model.layers.34.temporal_block.linear_out.weight": "model-00004-of-00004.safetensors",
"model.layers.34.temporal_block.linear_x.bias": "model-00004-of-00004.safetensors",
"model.layers.34.temporal_block.linear_x.weight": "model-00004-of-00004.safetensors",
"model.layers.34.temporal_block.linear_y.bias": "model-00004-of-00004.safetensors",
"model.layers.34.temporal_block.linear_y.weight": "model-00004-of-00004.safetensors",
"model.layers.34.temporal_block.rg_lru.input_gate_bias": "model-00004-of-00004.safetensors",
"model.layers.34.temporal_block.rg_lru.input_gate_weight": "model-00004-of-00004.safetensors",
"model.layers.34.temporal_block.rg_lru.recurrent_gate_bias": "model-00004-of-00004.safetensors",
"model.layers.34.temporal_block.rg_lru.recurrent_gate_weight": "model-00004-of-00004.safetensors",
"model.layers.34.temporal_block.rg_lru.recurrent_param": "model-00004-of-00004.safetensors",
"model.layers.34.temporal_pre_norm.weight": "model-00004-of-00004.safetensors",
"model.layers.35.channel_pre_norm.weight": "model-00004-of-00004.safetensors",
"model.layers.35.mlp_block.down_proj.bias": "model-00004-of-00004.safetensors",
"model.layers.35.mlp_block.down_proj.weight": "model-00004-of-00004.safetensors",
"model.layers.35.mlp_block.gate_proj.bias": "model-00004-of-00004.safetensors",
"model.layers.35.mlp_block.gate_proj.weight": "model-00004-of-00004.safetensors",
"model.layers.35.mlp_block.up_proj.bias": "model-00004-of-00004.safetensors",
"model.layers.35.mlp_block.up_proj.weight": "model-00004-of-00004.safetensors",
"model.layers.35.temporal_block.k_proj.weight": "model-00004-of-00004.safetensors",
"model.layers.35.temporal_block.o_proj.bias": "model-00004-of-00004.safetensors",
"model.layers.35.temporal_block.o_proj.weight": "model-00004-of-00004.safetensors",
"model.layers.35.temporal_block.q_proj.weight": "model-00004-of-00004.safetensors",
"model.layers.35.temporal_block.v_proj.weight": "model-00004-of-00004.safetensors",
"model.layers.35.temporal_pre_norm.weight": "model-00004-of-00004.safetensors",
"model.layers.36.channel_pre_norm.weight": "model-00004-of-00004.safetensors",
"model.layers.36.mlp_block.down_proj.bias": "model-00004-of-00004.safetensors",
"model.layers.36.mlp_block.down_proj.weight": "model-00004-of-00004.safetensors",
"model.layers.36.mlp_block.gate_proj.bias": "model-00004-of-00004.safetensors",
"model.layers.36.mlp_block.gate_proj.weight": "model-00004-of-00004.safetensors",
"model.layers.36.mlp_block.up_proj.bias": "model-00004-of-00004.safetensors",
"model.layers.36.mlp_block.up_proj.weight": "model-00004-of-00004.safetensors",
"model.layers.36.temporal_block.conv_1d.bias": "model-00004-of-00004.safetensors",
"model.layers.36.temporal_block.conv_1d.weight": "model-00004-of-00004.safetensors",
"model.layers.36.temporal_block.linear_out.bias": "model-00004-of-00004.safetensors",
"model.layers.36.temporal_block.linear_out.weight": "model-00004-of-00004.safetensors",
"model.layers.36.temporal_block.linear_x.bias": "model-00004-of-00004.safetensors",
"model.layers.36.temporal_block.linear_x.weight": "model-00004-of-00004.safetensors",
"model.layers.36.temporal_block.linear_y.bias": "model-00004-of-00004.safetensors",
"model.layers.36.temporal_block.linear_y.weight": "model-00004-of-00004.safetensors",
"model.layers.36.temporal_block.rg_lru.input_gate_bias": "model-00004-of-00004.safetensors",
"model.layers.36.temporal_block.rg_lru.input_gate_weight": "model-00004-of-00004.safetensors",
"model.layers.36.temporal_block.rg_lru.recurrent_gate_bias": "model-00004-of-00004.safetensors",
"model.layers.36.temporal_block.rg_lru.recurrent_gate_weight": "model-00004-of-00004.safetensors",
"model.layers.36.temporal_block.rg_lru.recurrent_param": "model-00004-of-00004.safetensors",
"model.layers.36.temporal_pre_norm.weight": "model-00004-of-00004.safetensors",
"model.layers.37.channel_pre_norm.weight": "model-00004-of-00004.safetensors",
"model.layers.37.mlp_block.down_proj.bias": "model-00004-of-00004.safetensors",
"model.layers.37.mlp_block.down_proj.weight": "model-00004-of-00004.safetensors",
"model.layers.37.mlp_block.gate_proj.bias": "model-00004-of-00004.safetensors",
"model.layers.37.mlp_block.gate_proj.weight": "model-00004-of-00004.safetensors",
"model.layers.37.mlp_block.up_proj.bias": "model-00004-of-00004.safetensors",
"model.layers.37.mlp_block.up_proj.weight": "model-00004-of-00004.safetensors",
"model.layers.37.temporal_block.conv_1d.bias": "model-00004-of-00004.safetensors",
"model.layers.37.temporal_block.conv_1d.weight": "model-00004-of-00004.safetensors",
"model.layers.37.temporal_block.linear_out.bias": "model-00004-of-00004.safetensors",
"model.layers.37.temporal_block.linear_out.weight": "model-00004-of-00004.safetensors",
"model.layers.37.temporal_block.linear_x.bias": "model-00004-of-00004.safetensors",
"model.layers.37.temporal_block.linear_x.weight": "model-00004-of-00004.safetensors",
"model.layers.37.temporal_block.linear_y.bias": "model-00004-of-00004.safetensors",
"model.layers.37.temporal_block.linear_y.weight": "model-00004-of-00004.safetensors",
"model.layers.37.temporal_block.rg_lru.input_gate_bias": "model-00004-of-00004.safetensors",
"model.layers.37.temporal_block.rg_lru.input_gate_weight": "model-00004-of-00004.safetensors",
"model.layers.37.temporal_block.rg_lru.recurrent_gate_bias": "model-00004-of-00004.safetensors",
"model.layers.37.temporal_block.rg_lru.recurrent_gate_weight": "model-00004-of-00004.safetensors",
"model.layers.37.temporal_block.rg_lru.recurrent_param": "model-00004-of-00004.safetensors",
"model.layers.37.temporal_pre_norm.weight": "model-00004-of-00004.safetensors",
"model.layers.4.channel_pre_norm.weight": "model-00001-of-00004.safetensors",
"model.layers.4.mlp_block.down_proj.bias": "model-00001-of-00004.safetensors",
"model.layers.4.mlp_block.down_proj.weight": "model-00001-of-00004.safetensors",
"model.layers.4.mlp_block.gate_proj.bias": "model-00001-of-00004.safetensors",
"model.layers.4.mlp_block.gate_proj.weight": "model-00001-of-00004.safetensors",
"model.layers.4.mlp_block.up_proj.bias": "model-00001-of-00004.safetensors",
"model.layers.4.mlp_block.up_proj.weight": "model-00001-of-00004.safetensors",
"model.layers.4.temporal_block.conv_1d.bias": "model-00001-of-00004.safetensors",
"model.layers.4.temporal_block.conv_1d.weight": "model-00001-of-00004.safetensors",
"model.layers.4.temporal_block.linear_out.bias": "model-00001-of-00004.safetensors",
"model.layers.4.temporal_block.linear_out.weight": "model-00001-of-00004.safetensors",
"model.layers.4.temporal_block.linear_x.bias": "model-00001-of-00004.safetensors",
"model.layers.4.temporal_block.linear_x.weight": "model-00001-of-00004.safetensors",
"model.layers.4.temporal_block.linear_y.bias": "model-00001-of-00004.safetensors",
"model.layers.4.temporal_block.linear_y.weight": "model-00001-of-00004.safetensors",
"model.layers.4.temporal_block.rg_lru.input_gate_bias": "model-00001-of-00004.safetensors",
"model.layers.4.temporal_block.rg_lru.input_gate_weight": "model-00001-of-00004.safetensors",
"model.layers.4.temporal_block.rg_lru.recurrent_gate_bias": "model-00001-of-00004.safetensors",
"model.layers.4.temporal_block.rg_lru.recurrent_gate_weight": "model-00001-of-00004.safetensors",
"model.layers.4.temporal_block.rg_lru.recurrent_param": "model-00001-of-00004.safetensors",
"model.layers.4.temporal_pre_norm.weight": "model-00001-of-00004.safetensors",
"model.layers.5.channel_pre_norm.weight": "model-00001-of-00004.safetensors",
"model.layers.5.mlp_block.down_proj.bias": "model-00001-of-00004.safetensors",
"model.layers.5.mlp_block.down_proj.weight": "model-00001-of-00004.safetensors",
"model.layers.5.mlp_block.gate_proj.bias": "model-00001-of-00004.safetensors",
"model.layers.5.mlp_block.gate_proj.weight": "model-00001-of-00004.safetensors",
"model.layers.5.mlp_block.up_proj.bias": "model-00001-of-00004.safetensors",
"model.layers.5.mlp_block.up_proj.weight": "model-00001-of-00004.safetensors",
"model.layers.5.temporal_block.k_proj.weight": "model-00001-of-00004.safetensors",
"model.layers.5.temporal_block.o_proj.bias": "model-00001-of-00004.safetensors",
"model.layers.5.temporal_block.o_proj.weight": "model-00001-of-00004.safetensors",
"model.layers.5.temporal_block.q_proj.weight": "model-00001-of-00004.safetensors",
"model.layers.5.temporal_block.v_proj.weight": "model-00001-of-00004.safetensors",
"model.layers.5.temporal_pre_norm.weight": "model-00001-of-00004.safetensors",
"model.layers.6.channel_pre_norm.weight": "model-00001-of-00004.safetensors",
"model.layers.6.mlp_block.down_proj.bias": "model-00001-of-00004.safetensors",
"model.layers.6.mlp_block.down_proj.weight": "model-00001-of-00004.safetensors",
"model.layers.6.mlp_block.gate_proj.bias": "model-00001-of-00004.safetensors",
"model.layers.6.mlp_block.gate_proj.weight": "model-00001-of-00004.safetensors",
"model.layers.6.mlp_block.up_proj.bias": "model-00001-of-00004.safetensors",
"model.layers.6.mlp_block.up_proj.weight": "model-00001-of-00004.safetensors",
"model.layers.6.temporal_block.conv_1d.bias": "model-00001-of-00004.safetensors",
"model.layers.6.temporal_block.conv_1d.weight": "model-00001-of-00004.safetensors",
"model.layers.6.temporal_block.linear_out.bias": "model-00001-of-00004.safetensors",
"model.layers.6.temporal_block.linear_out.weight": "model-00001-of-00004.safetensors",
"model.layers.6.temporal_block.linear_x.bias": "model-00001-of-00004.safetensors",
"model.layers.6.temporal_block.linear_x.weight": "model-00001-of-00004.safetensors",
"model.layers.6.temporal_block.linear_y.bias": "model-00001-of-00004.safetensors",
"model.layers.6.temporal_block.linear_y.weight": "model-00001-of-00004.safetensors",
"model.layers.6.temporal_block.rg_lru.input_gate_bias": "model-00001-of-00004.safetensors",
"model.layers.6.temporal_block.rg_lru.input_gate_weight": "model-00001-of-00004.safetensors",
"model.layers.6.temporal_block.rg_lru.recurrent_gate_bias": "model-00001-of-00004.safetensors",
"model.layers.6.temporal_block.rg_lru.recurrent_gate_weight": "model-00001-of-00004.safetensors",
"model.layers.6.temporal_block.rg_lru.recurrent_param": "model-00001-of-00004.safetensors",
"model.layers.6.temporal_pre_norm.weight": "model-00001-of-00004.safetensors",
"model.layers.7.channel_pre_norm.weight": "model-00001-of-00004.safetensors",
"model.layers.7.mlp_block.down_proj.bias": "model-00002-of-00004.safetensors",
"model.layers.7.mlp_block.down_proj.weight": "model-00002-of-00004.safetensors",
"model.layers.7.mlp_block.gate_proj.bias": "model-00002-of-00004.safetensors",
"model.layers.7.mlp_block.gate_proj.weight": "model-00002-of-00004.safetensors",
"model.layers.7.mlp_block.up_proj.bias": "model-00002-of-00004.safetensors",
"model.layers.7.mlp_block.up_proj.weight": "model-00002-of-00004.safetensors",
"model.layers.7.temporal_block.conv_1d.bias": "model-00001-of-00004.safetensors",
"model.layers.7.temporal_block.conv_1d.weight": "model-00001-of-00004.safetensors",
"model.layers.7.temporal_block.linear_out.bias": "model-00001-of-00004.safetensors",
"model.layers.7.temporal_block.linear_out.weight": "model-00001-of-00004.safetensors",
"model.layers.7.temporal_block.linear_x.bias": "model-00001-of-00004.safetensors",
"model.layers.7.temporal_block.linear_x.weight": "model-00001-of-00004.safetensors",
"model.layers.7.temporal_block.linear_y.bias": "model-00001-of-00004.safetensors",
"model.layers.7.temporal_block.linear_y.weight": "model-00001-of-00004.safetensors",
"model.layers.7.temporal_block.rg_lru.input_gate_bias": "model-00001-of-00004.safetensors",
"model.layers.7.temporal_block.rg_lru.input_gate_weight": "model-00001-of-00004.safetensors",
"model.layers.7.temporal_block.rg_lru.recurrent_gate_bias": "model-00001-of-00004.safetensors",
"model.layers.7.temporal_block.rg_lru.recurrent_gate_weight": "model-00001-of-00004.safetensors",
"model.layers.7.temporal_block.rg_lru.recurrent_param": "model-00001-of-00004.safetensors",
"model.layers.7.temporal_pre_norm.weight": "model-00001-of-00004.safetensors",
"model.layers.8.channel_pre_norm.weight": "model-00002-of-00004.safetensors",
"model.layers.8.mlp_block.down_proj.bias": "model-00002-of-00004.safetensors",
"model.layers.8.mlp_block.down_proj.weight": "model-00002-of-00004.safetensors",
"model.layers.8.mlp_block.gate_proj.bias": "model-00002-of-00004.safetensors",
"model.layers.8.mlp_block.gate_proj.weight": "model-00002-of-00004.safetensors",
"model.layers.8.mlp_block.up_proj.bias": "model-00002-of-00004.safetensors",
"model.layers.8.mlp_block.up_proj.weight": "model-00002-of-00004.safetensors",
"model.layers.8.temporal_block.k_proj.weight": "model-00002-of-00004.safetensors",
"model.layers.8.temporal_block.o_proj.bias": "model-00002-of-00004.safetensors",
"model.layers.8.temporal_block.o_proj.weight": "model-00002-of-00004.safetensors",
"model.layers.8.temporal_block.q_proj.weight": "model-00002-of-00004.safetensors",
"model.layers.8.temporal_block.v_proj.weight": "model-00002-of-00004.safetensors",
"model.layers.8.temporal_pre_norm.weight": "model-00002-of-00004.safetensors",
"model.layers.9.channel_pre_norm.weight": "model-00002-of-00004.safetensors",
"model.layers.9.mlp_block.down_proj.bias": "model-00002-of-00004.safetensors",
"model.layers.9.mlp_block.down_proj.weight": "model-00002-of-00004.safetensors",
"model.layers.9.mlp_block.gate_proj.bias": "model-00002-of-00004.safetensors",
"model.layers.9.mlp_block.gate_proj.weight": "model-00002-of-00004.safetensors",
"model.layers.9.mlp_block.up_proj.bias": "model-00002-of-00004.safetensors",
"model.layers.9.mlp_block.up_proj.weight": "model-00002-of-00004.safetensors",
"model.layers.9.temporal_block.conv_1d.bias": "model-00002-of-00004.safetensors",
"model.layers.9.temporal_block.conv_1d.weight": "model-00002-of-00004.safetensors",
"model.layers.9.temporal_block.linear_out.bias": "model-00002-of-00004.safetensors",
"model.layers.9.temporal_block.linear_out.weight": "model-00002-of-00004.safetensors",
"model.layers.9.temporal_block.linear_x.bias": "model-00002-of-00004.safetensors",
"model.layers.9.temporal_block.linear_x.weight": "model-00002-of-00004.safetensors",
"model.layers.9.temporal_block.linear_y.bias": "model-00002-of-00004.safetensors",
"model.layers.9.temporal_block.linear_y.weight": "model-00002-of-00004.safetensors",
"model.layers.9.temporal_block.rg_lru.input_gate_bias": "model-00002-of-00004.safetensors",
"model.layers.9.temporal_block.rg_lru.input_gate_weight": "model-00002-of-00004.safetensors",
"model.layers.9.temporal_block.rg_lru.recurrent_gate_bias": "model-00002-of-00004.safetensors",
"model.layers.9.temporal_block.rg_lru.recurrent_gate_weight": "model-00002-of-00004.safetensors",
"model.layers.9.temporal_block.rg_lru.recurrent_param": "model-00002-of-00004.safetensors",
"model.layers.9.temporal_pre_norm.weight": "model-00002-of-00004.safetensors"
}
}