{ "encoder.downsample.conv_end.bias": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 256 ], "data_offsets": [ 316504448, 316504960 ] }, "encoder.downsample.conv_end.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 256, 1, 3, 3 ], "data_offsets": [ 316504960, 316509568 ] }, "encoder.downsample.conv_in.bias": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 256 ], "data_offsets": [ 316509568, 316510080 ] }, "encoder.downsample.conv_in.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 256, 1, 3, 3 ], "data_offsets": [ 316510080, 316514688 ] }, "encoder.downsample.conv_mid.bias": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 256 ], "data_offsets": [ 316514688, 316515200 ] }, "encoder.downsample.conv_mid.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 256, 1, 3, 3 ], "data_offsets": [ 316515200, 316519808 ] }, "encoder.downsample.lin_end.bias": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 256 ], "data_offsets": [ 316519808, 316520320 ] }, "encoder.downsample.lin_end.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 256, 256, 1, 1 ], "data_offsets": [ 316520320, 316651392 ] }, "encoder.downsample.lin_mid.bias": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 256 ], "data_offsets": [ 316651392, 316651904 ] }, "encoder.downsample.lin_mid.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 256, 256, 1, 1 ], "data_offsets": [ 316651904, 316782976 ] }, "encoder.downsample.lin_out.bias": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 768 ], "data_offsets": [ 316782976, 316784512 ] }, "encoder.downsample.lin_out.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 768, 4096, 1 ], "data_offsets": [ 316784512, 323075968 ] }, "encoder.lfm.lfm.embedding_norm.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 768 ], "data_offsets": [ 323075968, 323077504 ] }, "encoder.lfm.lfm.layers.0.conv.conv.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 768, 1, 3 ], "data_offsets": [ 323077504, 323082112 ] }, "encoder.lfm.lfm.layers.0.conv.in_proj.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 2304, 768 ], "data_offsets": [ 323082112, 326621056 ] }, "encoder.lfm.lfm.layers.0.conv.out_proj.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 768, 768 ], "data_offsets": [ 326621056, 327800704 ] }, "encoder.lfm.lfm.layers.0.feed_forward.w1.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 1792, 768 ], "data_offsets": [ 327800704, 330553216 ] }, "encoder.lfm.lfm.layers.0.feed_forward.w2.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 768, 1792 ], "data_offsets": [ 330553216, 333305728 ] }, "encoder.lfm.lfm.layers.0.feed_forward.w3.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 1792, 768 ], "data_offsets": [ 333305728, 336058240 ] }, "encoder.lfm.lfm.layers.0.ffn_norm.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 768 ], "data_offsets": [ 336058240, 336059776 ] }, "encoder.lfm.lfm.layers.0.operator_norm.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 768 ], "data_offsets": [ 336059776, 336061312 ] }, "encoder.lfm.lfm.layers.1.conv.conv.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 768, 1, 3 ], "data_offsets": [ 336061312, 336065920 ] }, "encoder.lfm.lfm.layers.1.conv.in_proj.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 2304, 768 ], "data_offsets": [ 336065920, 339604864 ] }, "encoder.lfm.lfm.layers.1.conv.out_proj.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 768, 768 ], "data_offsets": [ 339604864, 340784512 ] }, "encoder.lfm.lfm.layers.1.feed_forward.w1.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 1792, 768 ], "data_offsets": [ 340784512, 343537024 ] }, "encoder.lfm.lfm.layers.1.feed_forward.w2.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 768, 1792 ], "data_offsets": [ 343537024, 346289536 ] }, "encoder.lfm.lfm.layers.1.feed_forward.w3.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 1792, 768 ], "data_offsets": [ 346289536, 349042048 ] }, "encoder.lfm.lfm.layers.1.ffn_norm.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 768 ], "data_offsets": [ 349042048, 349043584 ] }, "encoder.lfm.lfm.layers.1.operator_norm.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 768 ], "data_offsets": [ 349043584, 349045120 ] }, "encoder.lfm.lfm.layers.10.conv.conv.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 768, 1, 3 ], "data_offsets": [ 349045120, 349049728 ] }, "encoder.lfm.lfm.layers.10.conv.in_proj.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 2304, 768 ], "data_offsets": [ 349049728, 352588672 ] }, "encoder.lfm.lfm.layers.10.conv.out_proj.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 768, 768 ], "data_offsets": [ 352588672, 353768320 ] }, "encoder.lfm.lfm.layers.10.feed_forward.w1.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 1792, 768 ], "data_offsets": [ 353768320, 356520832 ] }, "encoder.lfm.lfm.layers.10.feed_forward.w2.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 768, 1792 ], "data_offsets": [ 356520832, 359273344 ] }, "encoder.lfm.lfm.layers.10.feed_forward.w3.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 1792, 768 ], "data_offsets": [ 359273344, 362025856 ] }, "encoder.lfm.lfm.layers.10.ffn_norm.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 768 ], "data_offsets": [ 362025856, 362027392 ] }, "encoder.lfm.lfm.layers.10.operator_norm.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 768 ], "data_offsets": [ 362027392, 362028928 ] }, "encoder.lfm.lfm.layers.11.feed_forward.w1.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 1792, 768 ], "data_offsets": [ 362028928, 364781440 ] }, "encoder.lfm.lfm.layers.11.feed_forward.w2.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 768, 1792 ], "data_offsets": [ 364781440, 367533952 ] }, "encoder.lfm.lfm.layers.11.feed_forward.w3.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 1792, 768 ], "data_offsets": [ 367533952, 370286464 ] }, "encoder.lfm.lfm.layers.11.ffn_norm.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 768 ], "data_offsets": [ 370286464, 370288000 ] }, "encoder.lfm.lfm.layers.11.operator_norm.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 768 ], "data_offsets": [ 370288000, 370289536 ] }, "encoder.lfm.lfm.layers.11.self_attn.k_layernorm.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 96 ], "data_offsets": [ 370289536, 370289728 ] }, "encoder.lfm.lfm.layers.11.self_attn.k_proj.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 384, 768 ], "data_offsets": [ 370289728, 370879552 ] }, "encoder.lfm.lfm.layers.11.self_attn.out_proj.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 768, 768 ], "data_offsets": [ 370879552, 372059200 ] }, "encoder.lfm.lfm.layers.11.self_attn.q_layernorm.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 96 ], "data_offsets": [ 372059200, 372059392 ] }, "encoder.lfm.lfm.layers.11.self_attn.q_proj.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 768, 768 ], "data_offsets": [ 372059392, 373239040 ] }, "encoder.lfm.lfm.layers.11.self_attn.v_proj.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 384, 768 ], "data_offsets": [ 373239040, 373828864 ] }, "encoder.lfm.lfm.layers.12.conv.conv.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 768, 1, 3 ], "data_offsets": [ 373828864, 373833472 ] }, "encoder.lfm.lfm.layers.12.conv.in_proj.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 2304, 768 ], "data_offsets": [ 373833472, 377372416 ] }, "encoder.lfm.lfm.layers.12.conv.out_proj.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 768, 768 ], "data_offsets": [ 377372416, 378552064 ] }, "encoder.lfm.lfm.layers.12.feed_forward.w1.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 1792, 768 ], "data_offsets": [ 378552064, 381304576 ] }, "encoder.lfm.lfm.layers.12.feed_forward.w2.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 768, 1792 ], "data_offsets": [ 381304576, 384057088 ] }, "encoder.lfm.lfm.layers.12.feed_forward.w3.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 1792, 768 ], "data_offsets": [ 384057088, 386809600 ] }, "encoder.lfm.lfm.layers.12.ffn_norm.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 768 ], "data_offsets": [ 386809600, 386811136 ] }, "encoder.lfm.lfm.layers.12.operator_norm.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 768 ], "data_offsets": [ 386811136, 386812672 ] }, "encoder.lfm.lfm.layers.13.conv.conv.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 768, 1, 3 ], "data_offsets": [ 386812672, 386817280 ] }, "encoder.lfm.lfm.layers.13.conv.in_proj.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 2304, 768 ], "data_offsets": [ 386817280, 390356224 ] }, "encoder.lfm.lfm.layers.13.conv.out_proj.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 768, 768 ], "data_offsets": [ 390356224, 391535872 ] }, "encoder.lfm.lfm.layers.13.feed_forward.w1.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 1792, 768 ], "data_offsets": [ 391535872, 394288384 ] }, "encoder.lfm.lfm.layers.13.feed_forward.w2.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 768, 1792 ], "data_offsets": [ 394288384, 397040896 ] }, "encoder.lfm.lfm.layers.13.feed_forward.w3.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 1792, 768 ], "data_offsets": [ 397040896, 399793408 ] }, "encoder.lfm.lfm.layers.13.ffn_norm.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 768 ], "data_offsets": [ 399793408, 399794944 ] }, "encoder.lfm.lfm.layers.13.operator_norm.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 768 ], "data_offsets": [ 399794944, 399796480 ] }, "encoder.lfm.lfm.layers.14.feed_forward.w1.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 1792, 768 ], "data_offsets": [ 399796480, 402548992 ] }, "encoder.lfm.lfm.layers.14.feed_forward.w2.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 768, 1792 ], "data_offsets": [ 402548992, 405301504 ] }, "encoder.lfm.lfm.layers.14.feed_forward.w3.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 1792, 768 ], "data_offsets": [ 405301504, 408054016 ] }, "encoder.lfm.lfm.layers.14.ffn_norm.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 768 ], "data_offsets": [ 408054016, 408055552 ] }, "encoder.lfm.lfm.layers.14.operator_norm.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 768 ], "data_offsets": [ 408055552, 408057088 ] }, "encoder.lfm.lfm.layers.14.self_attn.k_layernorm.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 96 ], "data_offsets": [ 408057088, 408057280 ] }, "encoder.lfm.lfm.layers.14.self_attn.k_proj.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 384, 768 ], "data_offsets": [ 408057280, 408647104 ] }, "encoder.lfm.lfm.layers.14.self_attn.out_proj.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 768, 768 ], "data_offsets": [ 408647104, 409826752 ] }, "encoder.lfm.lfm.layers.14.self_attn.q_layernorm.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 96 ], "data_offsets": [ 409826752, 409826944 ] }, "encoder.lfm.lfm.layers.14.self_attn.q_proj.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 768, 768 ], "data_offsets": [ 409826944, 411006592 ] }, "encoder.lfm.lfm.layers.14.self_attn.v_proj.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 384, 768 ], "data_offsets": [ 411006592, 411596416 ] }, "encoder.lfm.lfm.layers.15.feed_forward.w1.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 1792, 768 ], "data_offsets": [ 411596416, 414348928 ] }, "encoder.lfm.lfm.layers.15.feed_forward.w2.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 768, 1792 ], "data_offsets": [ 414348928, 417101440 ] }, "encoder.lfm.lfm.layers.15.feed_forward.w3.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 1792, 768 ], "data_offsets": [ 417101440, 419853952 ] }, "encoder.lfm.lfm.layers.15.ffn_norm.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 768 ], "data_offsets": [ 419853952, 419855488 ] }, "encoder.lfm.lfm.layers.15.operator_norm.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 768 ], "data_offsets": [ 419855488, 419857024 ] }, "encoder.lfm.lfm.layers.15.self_attn.k_layernorm.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 96 ], "data_offsets": [ 419857024, 419857216 ] }, "encoder.lfm.lfm.layers.15.self_attn.k_proj.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 384, 768 ], "data_offsets": [ 419857216, 420447040 ] }, "encoder.lfm.lfm.layers.15.self_attn.out_proj.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 768, 768 ], "data_offsets": [ 420447040, 421626688 ] }, "encoder.lfm.lfm.layers.15.self_attn.q_layernorm.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 96 ], "data_offsets": [ 421626688, 421626880 ] }, "encoder.lfm.lfm.layers.15.self_attn.q_proj.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 768, 768 ], "data_offsets": [ 421626880, 422806528 ] }, "encoder.lfm.lfm.layers.15.self_attn.v_proj.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 384, 768 ], "data_offsets": [ 422806528, 423396352 ] }, "encoder.lfm.lfm.layers.16.feed_forward.w1.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 1792, 768 ], "data_offsets": [ 423396352, 426148864 ] }, "encoder.lfm.lfm.layers.16.feed_forward.w2.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 768, 1792 ], "data_offsets": [ 426148864, 428901376 ] }, "encoder.lfm.lfm.layers.16.feed_forward.w3.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 1792, 768 ], "data_offsets": [ 428901376, 431653888 ] }, "encoder.lfm.lfm.layers.16.ffn_norm.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 768 ], "data_offsets": [ 431653888, 431655424 ] }, "encoder.lfm.lfm.layers.16.operator_norm.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 768 ], "data_offsets": [ 431655424, 431656960 ] }, "encoder.lfm.lfm.layers.16.self_attn.k_layernorm.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 96 ], "data_offsets": [ 431656960, 431657152 ] }, "encoder.lfm.lfm.layers.16.self_attn.k_proj.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 384, 768 ], "data_offsets": [ 431657152, 432246976 ] }, "encoder.lfm.lfm.layers.16.self_attn.out_proj.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 768, 768 ], "data_offsets": [ 432246976, 433426624 ] }, "encoder.lfm.lfm.layers.16.self_attn.q_layernorm.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 96 ], "data_offsets": [ 433426624, 433426816 ] }, "encoder.lfm.lfm.layers.16.self_attn.q_proj.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 768, 768 ], "data_offsets": [ 433426816, 434606464 ] }, "encoder.lfm.lfm.layers.16.self_attn.v_proj.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 384, 768 ], "data_offsets": [ 434606464, 435196288 ] }, "encoder.lfm.lfm.layers.2.feed_forward.w1.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 1792, 768 ], "data_offsets": [ 435196288, 437948800 ] }, "encoder.lfm.lfm.layers.2.feed_forward.w2.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 768, 1792 ], "data_offsets": [ 437948800, 440701312 ] }, "encoder.lfm.lfm.layers.2.feed_forward.w3.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 1792, 768 ], "data_offsets": [ 440701312, 443453824 ] }, "encoder.lfm.lfm.layers.2.ffn_norm.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 768 ], "data_offsets": [ 443453824, 443455360 ] }, "encoder.lfm.lfm.layers.2.operator_norm.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 768 ], "data_offsets": [ 443455360, 443456896 ] }, "encoder.lfm.lfm.layers.2.self_attn.k_layernorm.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 96 ], "data_offsets": [ 443456896, 443457088 ] }, "encoder.lfm.lfm.layers.2.self_attn.k_proj.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 384, 768 ], "data_offsets": [ 443457088, 444046912 ] }, "encoder.lfm.lfm.layers.2.self_attn.out_proj.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 768, 768 ], "data_offsets": [ 444046912, 445226560 ] }, "encoder.lfm.lfm.layers.2.self_attn.q_layernorm.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 96 ], "data_offsets": [ 445226560, 445226752 ] }, "encoder.lfm.lfm.layers.2.self_attn.q_proj.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 768, 768 ], "data_offsets": [ 445226752, 446406400 ] }, "encoder.lfm.lfm.layers.2.self_attn.v_proj.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 384, 768 ], "data_offsets": [ 446406400, 446996224 ] }, "encoder.lfm.lfm.layers.3.conv.conv.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 768, 1, 3 ], "data_offsets": [ 446996224, 447000832 ] }, "encoder.lfm.lfm.layers.3.conv.in_proj.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 2304, 768 ], "data_offsets": [ 447000832, 450539776 ] }, "encoder.lfm.lfm.layers.3.conv.out_proj.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 768, 768 ], "data_offsets": [ 450539776, 451719424 ] }, "encoder.lfm.lfm.layers.3.feed_forward.w1.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 1792, 768 ], "data_offsets": [ 451719424, 454471936 ] }, "encoder.lfm.lfm.layers.3.feed_forward.w2.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 768, 1792 ], "data_offsets": [ 454471936, 457224448 ] }, "encoder.lfm.lfm.layers.3.feed_forward.w3.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 1792, 768 ], "data_offsets": [ 457224448, 459976960 ] }, "encoder.lfm.lfm.layers.3.ffn_norm.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 768 ], "data_offsets": [ 459976960, 459978496 ] }, "encoder.lfm.lfm.layers.3.operator_norm.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 768 ], "data_offsets": [ 459978496, 459980032 ] }, "encoder.lfm.lfm.layers.4.conv.conv.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 768, 1, 3 ], "data_offsets": [ 459980032, 459984640 ] }, "encoder.lfm.lfm.layers.4.conv.in_proj.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 2304, 768 ], "data_offsets": [ 459984640, 463523584 ] }, "encoder.lfm.lfm.layers.4.conv.out_proj.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 768, 768 ], "data_offsets": [ 463523584, 464703232 ] }, "encoder.lfm.lfm.layers.4.feed_forward.w1.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 1792, 768 ], "data_offsets": [ 464703232, 467455744 ] }, "encoder.lfm.lfm.layers.4.feed_forward.w2.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 768, 1792 ], "data_offsets": [ 467455744, 470208256 ] }, "encoder.lfm.lfm.layers.4.feed_forward.w3.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 1792, 768 ], "data_offsets": [ 470208256, 472960768 ] }, "encoder.lfm.lfm.layers.4.ffn_norm.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 768 ], "data_offsets": [ 472960768, 472962304 ] }, "encoder.lfm.lfm.layers.4.operator_norm.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 768 ], "data_offsets": [ 472962304, 472963840 ] }, "encoder.lfm.lfm.layers.5.feed_forward.w1.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 1792, 768 ], "data_offsets": [ 472963840, 475716352 ] }, "encoder.lfm.lfm.layers.5.feed_forward.w2.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 768, 1792 ], "data_offsets": [ 475716352, 478468864 ] }, "encoder.lfm.lfm.layers.5.feed_forward.w3.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 1792, 768 ], "data_offsets": [ 478468864, 481221376 ] }, "encoder.lfm.lfm.layers.5.ffn_norm.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 768 ], "data_offsets": [ 481221376, 481222912 ] }, "encoder.lfm.lfm.layers.5.operator_norm.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 768 ], "data_offsets": [ 481222912, 481224448 ] }, "encoder.lfm.lfm.layers.5.self_attn.k_layernorm.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 96 ], "data_offsets": [ 481224448, 481224640 ] }, "encoder.lfm.lfm.layers.5.self_attn.k_proj.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 384, 768 ], "data_offsets": [ 481224640, 481814464 ] }, "encoder.lfm.lfm.layers.5.self_attn.out_proj.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 768, 768 ], "data_offsets": [ 481814464, 482994112 ] }, "encoder.lfm.lfm.layers.5.self_attn.q_layernorm.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 96 ], "data_offsets": [ 482994112, 482994304 ] }, "encoder.lfm.lfm.layers.5.self_attn.q_proj.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 768, 768 ], "data_offsets": [ 482994304, 484173952 ] }, "encoder.lfm.lfm.layers.5.self_attn.v_proj.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 384, 768 ], "data_offsets": [ 484173952, 484763776 ] }, "encoder.lfm.lfm.layers.6.conv.conv.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 768, 1, 3 ], "data_offsets": [ 484763776, 484768384 ] }, "encoder.lfm.lfm.layers.6.conv.in_proj.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 2304, 768 ], "data_offsets": [ 484768384, 488307328 ] }, "encoder.lfm.lfm.layers.6.conv.out_proj.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 768, 768 ], "data_offsets": [ 488307328, 489486976 ] }, "encoder.lfm.lfm.layers.6.feed_forward.w1.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 1792, 768 ], "data_offsets": [ 489486976, 492239488 ] }, "encoder.lfm.lfm.layers.6.feed_forward.w2.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 768, 1792 ], "data_offsets": [ 492239488, 494992000 ] }, "encoder.lfm.lfm.layers.6.feed_forward.w3.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 1792, 768 ], "data_offsets": [ 494992000, 497744512 ] }, "encoder.lfm.lfm.layers.6.ffn_norm.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 768 ], "data_offsets": [ 497744512, 497746048 ] }, "encoder.lfm.lfm.layers.6.operator_norm.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 768 ], "data_offsets": [ 497746048, 497747584 ] }, "encoder.lfm.lfm.layers.7.conv.conv.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 768, 1, 3 ], "data_offsets": [ 497747584, 497752192 ] }, "encoder.lfm.lfm.layers.7.conv.in_proj.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 2304, 768 ], "data_offsets": [ 497752192, 501291136 ] }, "encoder.lfm.lfm.layers.7.conv.out_proj.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 768, 768 ], "data_offsets": [ 501291136, 502470784 ] }, "encoder.lfm.lfm.layers.7.feed_forward.w1.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 1792, 768 ], "data_offsets": [ 502470784, 505223296 ] }, "encoder.lfm.lfm.layers.7.feed_forward.w2.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 768, 1792 ], "data_offsets": [ 505223296, 507975808 ] }, "encoder.lfm.lfm.layers.7.feed_forward.w3.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 1792, 768 ], "data_offsets": [ 507975808, 510728320 ] }, "encoder.lfm.lfm.layers.7.ffn_norm.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 768 ], "data_offsets": [ 510728320, 510729856 ] }, "encoder.lfm.lfm.layers.7.operator_norm.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 768 ], "data_offsets": [ 510729856, 510731392 ] }, "encoder.lfm.lfm.layers.8.feed_forward.w1.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 1792, 768 ], "data_offsets": [ 510731392, 513483904 ] }, "encoder.lfm.lfm.layers.8.feed_forward.w2.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 768, 1792 ], "data_offsets": [ 513483904, 516236416 ] }, "encoder.lfm.lfm.layers.8.feed_forward.w3.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 1792, 768 ], "data_offsets": [ 516236416, 518988928 ] }, "encoder.lfm.lfm.layers.8.ffn_norm.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 768 ], "data_offsets": [ 518988928, 518990464 ] }, "encoder.lfm.lfm.layers.8.operator_norm.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 768 ], "data_offsets": [ 518990464, 518992000 ] }, "encoder.lfm.lfm.layers.8.self_attn.k_layernorm.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 96 ], "data_offsets": [ 518992000, 518992192 ] }, "encoder.lfm.lfm.layers.8.self_attn.k_proj.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 384, 768 ], "data_offsets": [ 518992192, 519582016 ] }, "encoder.lfm.lfm.layers.8.self_attn.out_proj.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 768, 768 ], "data_offsets": [ 519582016, 520761664 ] }, "encoder.lfm.lfm.layers.8.self_attn.q_layernorm.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 96 ], "data_offsets": [ 520761664, 520761856 ] }, "encoder.lfm.lfm.layers.8.self_attn.q_proj.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 768, 768 ], "data_offsets": [ 520761856, 521941504 ] }, "encoder.lfm.lfm.layers.8.self_attn.v_proj.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 384, 768 ], "data_offsets": [ 521941504, 522531328 ] }, "encoder.lfm.lfm.layers.9.conv.conv.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 768, 1, 3 ], "data_offsets": [ 522531328, 522535936 ] }, "encoder.lfm.lfm.layers.9.conv.in_proj.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 2304, 768 ], "data_offsets": [ 522535936, 526074880 ] }, "encoder.lfm.lfm.layers.9.conv.out_proj.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 768, 768 ], "data_offsets": [ 526074880, 527254528 ] }, "encoder.lfm.lfm.layers.9.feed_forward.w1.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 1792, 768 ], "data_offsets": [ 527254528, 530007040 ] }, "encoder.lfm.lfm.layers.9.feed_forward.w2.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 768, 1792 ], "data_offsets": [ 530007040, 532759552 ] }, "encoder.lfm.lfm.layers.9.feed_forward.w3.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 1792, 768 ], "data_offsets": [ 532759552, 535512064 ] }, "encoder.lfm.lfm.layers.9.ffn_norm.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 768 ], "data_offsets": [ 535512064, 535513600 ] }, "encoder.lfm.lfm.layers.9.operator_norm.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 768 ], "data_offsets": [ 535513600, 535515136 ] }, "mlp.model.0.bias": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 768 ], "data_offsets": [ 3993120256, 3993121792 ] }, "mlp.model.0.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 768 ], "data_offsets": [ 3993121792, 3993123328 ] }, "mlp.model.1.bias": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 2048 ], "data_offsets": [ 3993123328, 3993127424 ] }, "mlp.model.1.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 2048, 768 ], "data_offsets": [ 3993127424, 3996273152 ] }, "mlp.model.3.bias": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 2048 ], "data_offsets": [ 3996273152, 3996277248 ] }, "mlp.model.3.weight": { "source_shard": "model.safetensors", "dtype": "BF16", "shape": [ 2048, 2048 ], "data_offsets": [ 3996277248, 4004665856 ] } }