Llama 3.2 Vision is a collection of instruction-tuned image reasoning generative models in 11B and 90B sizes.
vision
11b
90b
138.3K Pulls Updated 2 weeks ago
11f274007f09 · 6.0GB
-
general.architecturemllama
-
general.file_type15
-
general.nameModel
-
general.quantization_version2
-
general.size_label10B
-
general.typemodel
-
mllama.attention.cross_attention_layers[3, 8, 13, 18, 23, ...]
-
mllama.attention.head_count32
-
mllama.attention.head_count_kv8
-
mllama.attention.layer_norm_rms_epsilon1e-05
-
mllama.block_count40
-
mllama.context_length131072
-
mllama.embedding_length4096
-
mllama.feed_forward_length14336
-
mllama.rope.dimension_count128
-
mllama.rope.freq_base500000
-
mllama.vocab_size128256
-
tokenizer.ggml.add_bos_tokentrue
-
tokenizer.ggml.bos_token_id128000
-
tokenizer.ggml.eos_token_id128009
-
tokenizer.ggml.merges[Ġ Ġ, Ġ ĠĠĠ, ĠĠ ĠĠ, ĠĠĠ Ġ, i n, ...]
-
tokenizer.ggml.modelgpt2
-
tokenizer.ggml.padding_token_id128004
-
tokenizer.ggml.prellama-bpe
-
tokenizer.ggml.token_type[1, 1, 1, 1, 1, ...]
-
tokenizer.ggml.tokens[!, ", #, $, %, ...]
-
NameTypeShape
-
token_embd.weightQ4_K[4096, 128264]
-
blk.0.attn_k.weightQ4_K[4096, 1024]
-
blk.0.attn_norm.weightF32[4096]
-
blk.0.attn_output.weightQ4_K[4096, 4096]
-
blk.0.attn_q.weightQ4_K[4096, 4096]
-
blk.0.attn_v.weightQ6_K[4096, 1024]
-
blk.0.ffn_down.weightQ6_K[14336, 4096]
-
blk.0.ffn_gate.weightQ4_K[4096, 14336]
-
blk.0.ffn_norm.weightF32[4096]
-
blk.0.ffn_up.weightQ4_K[4096, 14336]
-
blk.1.attn_k.weightQ4_K[4096, 1024]
-
blk.1.attn_norm.weightF32[4096]
-
blk.1.attn_output.weightQ4_K[4096, 4096]
-
blk.1.attn_q.weightQ4_K[4096, 4096]
-
blk.1.attn_v.weightQ6_K[4096, 1024]
-
blk.1.ffn_down.weightQ6_K[14336, 4096]
-
blk.1.ffn_gate.weightQ4_K[4096, 14336]
-
blk.1.ffn_norm.weightF32[4096]
-
blk.1.ffn_up.weightQ4_K[4096, 14336]
-
blk.2.attn_k.weightQ4_K[4096, 1024]
-
blk.2.attn_norm.weightF32[4096]
-
blk.2.attn_output.weightQ4_K[4096, 4096]
-
blk.2.attn_q.weightQ4_K[4096, 4096]
-
blk.2.attn_v.weightQ6_K[4096, 1024]
-
blk.2.ffn_down.weightQ6_K[14336, 4096]
-
blk.2.ffn_gate.weightQ4_K[4096, 14336]
-
blk.2.ffn_norm.weightF32[4096]
-
blk.2.ffn_up.weightQ4_K[4096, 14336]
-
blk.3.attn_norm.weightF32[4096]
-
blk.3.cross_attn_attn_gateF32[1]
-
blk.3.cross_attn_k_norm.weightF32[128]
-
blk.3.cross_attn_k_proj.weightQ4_K[4096, 1024]
-
blk.3.cross_attn_mlp_gateF32[1]
-
blk.3.cross_attn_o_proj.weightQ4_K[4096, 4096]
-
blk.3.cross_attn_q_norm.weightF32[128]
-
blk.3.cross_attn_q_proj.weightQ4_K[4096, 4096]
-
blk.3.cross_attn_v_proj.weightQ4_K[4096, 1024]
-
blk.3.ffn_down.weightQ6_K[14336, 4096]
-
blk.3.ffn_gate.weightQ4_K[4096, 14336]
-
blk.3.ffn_norm.weightF32[4096]
-
blk.3.ffn_up.weightQ4_K[4096, 14336]
-
blk.4.attn_k.weightQ4_K[4096, 1024]
-
blk.4.attn_norm.weightF32[4096]
-
blk.4.attn_output.weightQ4_K[4096, 4096]
-
blk.4.attn_q.weightQ4_K[4096, 4096]
-
blk.4.attn_v.weightQ6_K[4096, 1024]
-
blk.4.ffn_down.weightQ6_K[14336, 4096]
-
blk.4.ffn_gate.weightQ4_K[4096, 14336]
-
blk.4.ffn_norm.weightF32[4096]
-
blk.4.ffn_up.weightQ4_K[4096, 14336]
-
blk.5.attn_k.weightQ4_K[4096, 1024]
-
blk.5.attn_norm.weightF32[4096]
-
blk.5.attn_output.weightQ4_K[4096, 4096]
-
blk.5.attn_q.weightQ4_K[4096, 4096]
-
blk.5.attn_v.weightQ4_K[4096, 1024]
-
blk.5.ffn_down.weightQ4_K[14336, 4096]
-
blk.5.ffn_gate.weightQ4_K[4096, 14336]
-
blk.5.ffn_norm.weightF32[4096]
-
blk.5.ffn_up.weightQ4_K[4096, 14336]
-
blk.6.attn_k.weightQ4_K[4096, 1024]
-
blk.6.attn_norm.weightF32[4096]
-
blk.6.attn_output.weightQ4_K[4096, 4096]
-
blk.6.attn_q.weightQ4_K[4096, 4096]
-
blk.6.attn_v.weightQ4_K[4096, 1024]
-
blk.6.ffn_down.weightQ4_K[14336, 4096]
-
blk.6.ffn_gate.weightQ4_K[4096, 14336]
-
blk.6.ffn_norm.weightF32[4096]
-
blk.6.ffn_up.weightQ4_K[4096, 14336]
-
blk.7.attn_k.weightQ4_K[4096, 1024]
-
blk.7.attn_norm.weightF32[4096]
-
blk.7.attn_output.weightQ4_K[4096, 4096]
-
blk.7.attn_q.weightQ4_K[4096, 4096]
-
blk.7.attn_v.weightQ6_K[4096, 1024]
-
blk.7.ffn_down.weightQ6_K[14336, 4096]
-
blk.7.ffn_gate.weightQ4_K[4096, 14336]
-
blk.7.ffn_norm.weightF32[4096]
-
blk.7.ffn_up.weightQ4_K[4096, 14336]
-
blk.8.attn_norm.weightF32[4096]
-
blk.8.cross_attn_attn_gateF32[1]
-
blk.8.cross_attn_k_norm.weightF32[128]
-
blk.8.cross_attn_k_proj.weightQ4_K[4096, 1024]
-
blk.8.cross_attn_mlp_gateF32[1]
-
blk.8.cross_attn_o_proj.weightQ4_K[4096, 4096]
-
blk.8.cross_attn_q_norm.weightF32[128]
-
blk.8.cross_attn_q_proj.weightQ4_K[4096, 4096]
-
blk.8.cross_attn_v_proj.weightQ4_K[4096, 1024]
-
blk.8.ffn_down.weightQ4_K[14336, 4096]
-
blk.8.ffn_gate.weightQ4_K[4096, 14336]
-
blk.8.ffn_norm.weightF32[4096]
-
blk.8.ffn_up.weightQ4_K[4096, 14336]
-
blk.9.attn_k.weightQ4_K[4096, 1024]
-
blk.9.attn_norm.weightF32[4096]
-
blk.9.attn_output.weightQ4_K[4096, 4096]
-
blk.9.attn_q.weightQ4_K[4096, 4096]
-
blk.9.attn_v.weightQ4_K[4096, 1024]
-
blk.9.ffn_down.weightQ4_K[14336, 4096]
-
blk.9.ffn_gate.weightQ4_K[4096, 14336]
-
blk.9.ffn_norm.weightF32[4096]
-
blk.9.ffn_up.weightQ4_K[4096, 14336]
-
blk.10.attn_k.weightQ4_K[4096, 1024]
-
blk.10.attn_norm.weightF32[4096]
-
blk.10.attn_output.weightQ4_K[4096, 4096]
-
blk.10.attn_q.weightQ4_K[4096, 4096]
-
blk.10.attn_v.weightQ4_K[4096, 1024]
-
blk.10.ffn_down.weightQ4_K[14336, 4096]
-
blk.10.ffn_gate.weightQ4_K[4096, 14336]
-
blk.10.ffn_norm.weightF32[4096]
-
blk.10.ffn_up.weightQ4_K[4096, 14336]
-
blk.11.attn_k.weightQ4_K[4096, 1024]
-
blk.11.attn_norm.weightF32[4096]
-
blk.11.attn_output.weightQ4_K[4096, 4096]
-
blk.11.attn_q.weightQ4_K[4096, 4096]
-
blk.11.attn_v.weightQ4_K[4096, 1024]
-
blk.11.ffn_down.weightQ4_K[14336, 4096]
-
blk.11.ffn_gate.weightQ4_K[4096, 14336]
-
blk.11.ffn_norm.weightF32[4096]
-
blk.11.ffn_up.weightQ4_K[4096, 14336]
-
blk.12.attn_k.weightQ4_K[4096, 1024]
-
blk.12.attn_norm.weightF32[4096]
-
blk.12.attn_output.weightQ4_K[4096, 4096]
-
blk.12.attn_q.weightQ4_K[4096, 4096]
-
blk.12.attn_v.weightQ6_K[4096, 1024]
-
blk.12.ffn_down.weightQ6_K[14336, 4096]
-
blk.12.ffn_gate.weightQ4_K[4096, 14336]
-
blk.12.ffn_norm.weightF32[4096]
-
blk.12.ffn_up.weightQ4_K[4096, 14336]
-
blk.13.attn_norm.weightF32[4096]
-
blk.13.cross_attn_attn_gateF32[1]
-
blk.13.cross_attn_k_norm.weightF32[128]
-
blk.13.cross_attn_k_proj.weightQ4_K[4096, 1024]
-
blk.13.cross_attn_mlp_gateF32[1]
-
blk.13.cross_attn_o_proj.weightQ4_K[4096, 4096]
-
blk.13.cross_attn_q_norm.weightF32[128]
-
blk.13.cross_attn_q_proj.weightQ4_K[4096, 4096]
-
blk.13.cross_attn_v_proj.weightQ4_K[4096, 1024]
-
blk.13.ffn_down.weightQ4_K[14336, 4096]
-
blk.13.ffn_gate.weightQ4_K[4096, 14336]
-
blk.13.ffn_norm.weightF32[4096]
-
blk.13.ffn_up.weightQ4_K[4096, 14336]
-
blk.14.attn_k.weightQ4_K[4096, 1024]
-
blk.14.attn_norm.weightF32[4096]
-
blk.14.attn_output.weightQ4_K[4096, 4096]
-
blk.14.attn_q.weightQ4_K[4096, 4096]
-
blk.14.attn_v.weightQ4_K[4096, 1024]
-
blk.14.ffn_down.weightQ4_K[14336, 4096]
-
blk.14.ffn_gate.weightQ4_K[4096, 14336]
-
blk.14.ffn_norm.weightF32[4096]
-
blk.14.ffn_up.weightQ4_K[4096, 14336]
-
blk.15.attn_k.weightQ4_K[4096, 1024]
-
blk.15.attn_norm.weightF32[4096]
-
blk.15.attn_output.weightQ4_K[4096, 4096]
-
blk.15.attn_q.weightQ4_K[4096, 4096]
-
blk.15.attn_v.weightQ4_K[4096, 1024]
-
blk.15.ffn_down.weightQ6_K[14336, 4096]
-
blk.15.ffn_gate.weightQ4_K[4096, 14336]
-
blk.15.ffn_norm.weightF32[4096]
-
blk.15.ffn_up.weightQ4_K[4096, 14336]
-
blk.16.attn_k.weightQ4_K[4096, 1024]
-
blk.16.attn_norm.weightF32[4096]
-
blk.16.attn_output.weightQ4_K[4096, 4096]
-
blk.16.attn_q.weightQ4_K[4096, 4096]
-
blk.16.attn_v.weightQ6_K[4096, 1024]
-
blk.16.ffn_down.weightQ6_K[14336, 4096]
-
blk.16.ffn_gate.weightQ4_K[4096, 14336]
-
blk.16.ffn_norm.weightF32[4096]
-
blk.16.ffn_up.weightQ4_K[4096, 14336]
-
blk.17.attn_k.weightQ4_K[4096, 1024]
-
blk.17.attn_norm.weightF32[4096]
-
blk.17.attn_output.weightQ4_K[4096, 4096]
-
blk.17.attn_q.weightQ4_K[4096, 4096]
-
blk.17.attn_v.weightQ4_K[4096, 1024]
-
blk.17.ffn_down.weightQ4_K[14336, 4096]
-
blk.17.ffn_gate.weightQ4_K[4096, 14336]
-
blk.17.ffn_norm.weightF32[4096]
-
blk.17.ffn_up.weightQ4_K[4096, 14336]
-
blk.18.attn_norm.weightF32[4096]
-
blk.18.cross_attn_attn_gateF32[1]
-
blk.18.cross_attn_k_norm.weightF32[128]
-
blk.18.cross_attn_k_proj.weightQ4_K[4096, 1024]
-
blk.18.cross_attn_mlp_gateF32[1]
-
blk.18.cross_attn_o_proj.weightQ4_K[4096, 4096]
-
blk.18.cross_attn_q_norm.weightF32[128]
-
blk.18.cross_attn_q_proj.weightQ4_K[4096, 4096]
-
blk.18.cross_attn_v_proj.weightQ4_K[4096, 1024]
-
blk.18.ffn_down.weightQ4_K[14336, 4096]
-
blk.18.ffn_gate.weightQ4_K[4096, 14336]
-
blk.18.ffn_norm.weightF32[4096]
-
blk.18.ffn_up.weightQ4_K[4096, 14336]
-
blk.19.attn_k.weightQ4_K[4096, 1024]
-
blk.19.attn_norm.weightF32[4096]
-
blk.19.attn_output.weightQ4_K[4096, 4096]
-
blk.19.attn_q.weightQ4_K[4096, 4096]
-
blk.19.attn_v.weightQ6_K[4096, 1024]
-
blk.19.ffn_down.weightQ6_K[14336, 4096]
-
blk.19.ffn_gate.weightQ4_K[4096, 14336]
-
blk.19.ffn_norm.weightF32[4096]
-
blk.19.ffn_up.weightQ4_K[4096, 14336]
-
blk.20.attn_k.weightQ4_K[4096, 1024]
-
blk.20.attn_norm.weightF32[4096]
-
blk.20.attn_output.weightQ4_K[4096, 4096]
-
blk.20.attn_q.weightQ4_K[4096, 4096]
-
blk.20.attn_v.weightQ4_K[4096, 1024]
-
blk.20.ffn_down.weightQ4_K[14336, 4096]
-
blk.20.ffn_gate.weightQ4_K[4096, 14336]
-
blk.20.ffn_norm.weightF32[4096]
-
blk.20.ffn_up.weightQ4_K[4096, 14336]
-
blk.21.attn_k.weightQ4_K[4096, 1024]
-
blk.21.attn_norm.weightF32[4096]
-
blk.21.attn_output.weightQ4_K[4096, 4096]
-
blk.21.attn_q.weightQ4_K[4096, 4096]
-
blk.21.attn_v.weightQ4_K[4096, 1024]
-
blk.21.ffn_down.weightQ4_K[14336, 4096]
-
blk.21.ffn_gate.weightQ4_K[4096, 14336]
-
blk.21.ffn_norm.weightF32[4096]
-
blk.21.ffn_up.weightQ4_K[4096, 14336]
-
blk.22.attn_k.weightQ4_K[4096, 1024]
-
blk.22.attn_norm.weightF32[4096]
-
blk.22.attn_output.weightQ4_K[4096, 4096]
-
blk.22.attn_q.weightQ4_K[4096, 4096]
-
blk.22.attn_v.weightQ6_K[4096, 1024]
-
blk.22.ffn_down.weightQ6_K[14336, 4096]
-
blk.22.ffn_gate.weightQ4_K[4096, 14336]
-
blk.22.ffn_norm.weightF32[4096]
-
blk.22.ffn_up.weightQ4_K[4096, 14336]
-
blk.23.attn_norm.weightF32[4096]
-
blk.23.cross_attn_attn_gateF32[1]
-
blk.23.cross_attn_k_norm.weightF32[128]
-
blk.23.cross_attn_k_proj.weightQ4_K[4096, 1024]
-
blk.23.cross_attn_mlp_gateF32[1]
-
blk.23.cross_attn_o_proj.weightQ4_K[4096, 4096]
-
blk.23.cross_attn_q_norm.weightF32[128]
-
blk.23.cross_attn_q_proj.weightQ4_K[4096, 4096]
-
blk.23.cross_attn_v_proj.weightQ4_K[4096, 1024]
-
blk.23.ffn_down.weightQ4_K[14336, 4096]
-
blk.23.ffn_gate.weightQ4_K[4096, 14336]
-
blk.23.ffn_norm.weightF32[4096]
-
blk.23.ffn_up.weightQ4_K[4096, 14336]
-
blk.24.attn_k.weightQ4_K[4096, 1024]
-
blk.24.attn_norm.weightF32[4096]
-
blk.24.attn_output.weightQ4_K[4096, 4096]
-
blk.24.attn_q.weightQ4_K[4096, 4096]
-
blk.24.attn_v.weightQ4_K[4096, 1024]
-
blk.24.ffn_down.weightQ4_K[14336, 4096]
-
blk.24.ffn_gate.weightQ4_K[4096, 14336]
-
blk.24.ffn_norm.weightF32[4096]
-
blk.24.ffn_up.weightQ4_K[4096, 14336]
-
blk.25.attn_k.weightQ4_K[4096, 1024]
-
blk.25.attn_norm.weightF32[4096]
-
blk.25.attn_output.weightQ4_K[4096, 4096]
-
blk.25.attn_q.weightQ4_K[4096, 4096]
-
blk.25.attn_v.weightQ4_K[4096, 1024]
-
blk.25.ffn_down.weightQ6_K[14336, 4096]
-
blk.25.ffn_gate.weightQ4_K[4096, 14336]
-
blk.25.ffn_norm.weightF32[4096]
-
blk.25.ffn_up.weightQ4_K[4096, 14336]
-
blk.26.attn_k.weightQ4_K[4096, 1024]
-
blk.26.attn_norm.weightF32[4096]
-
blk.26.attn_output.weightQ4_K[4096, 4096]
-
blk.26.attn_q.weightQ4_K[4096, 4096]
-
blk.26.attn_v.weightQ6_K[4096, 1024]
-
blk.26.ffn_down.weightQ4_K[14336, 4096]
-
blk.26.ffn_gate.weightQ4_K[4096, 14336]
-
blk.26.ffn_norm.weightF32[4096]
-
blk.26.ffn_up.weightQ4_K[4096, 14336]
-
blk.27.attn_k.weightQ4_K[4096, 1024]
-
blk.27.attn_norm.weightF32[4096]
-
blk.27.attn_output.weightQ4_K[4096, 4096]
-
blk.27.attn_q.weightQ4_K[4096, 4096]
-
blk.27.attn_v.weightQ4_K[4096, 1024]
-
blk.27.ffn_down.weightQ4_K[14336, 4096]
-
blk.27.ffn_gate.weightQ4_K[4096, 14336]
-
blk.27.ffn_norm.weightF32[4096]
-
blk.27.ffn_up.weightQ4_K[4096, 14336]
-
blk.28.attn_norm.weightF32[4096]
-
blk.28.cross_attn_attn_gateF32[1]
-
blk.28.cross_attn_k_norm.weightF32[128]
-
blk.28.cross_attn_k_proj.weightQ4_K[4096, 1024]
-
blk.28.cross_attn_mlp_gateF32[1]
-
blk.28.cross_attn_o_proj.weightQ4_K[4096, 4096]
-
blk.28.cross_attn_q_norm.weightF32[128]
-
blk.28.cross_attn_q_proj.weightQ4_K[4096, 4096]
-
blk.28.cross_attn_v_proj.weightQ4_K[4096, 1024]
-
blk.28.ffn_down.weightQ6_K[14336, 4096]
-
blk.28.ffn_gate.weightQ4_K[4096, 14336]
-
blk.28.ffn_norm.weightF32[4096]
-
blk.28.ffn_up.weightQ4_K[4096, 14336]
-
blk.29.attn_k.weightQ4_K[4096, 1024]
-
blk.29.attn_norm.weightF32[4096]
-
blk.29.attn_output.weightQ4_K[4096, 4096]
-
blk.29.attn_q.weightQ4_K[4096, 4096]
-
blk.29.attn_v.weightQ4_K[4096, 1024]
-
blk.29.ffn_down.weightQ4_K[14336, 4096]
-
blk.29.ffn_gate.weightQ4_K[4096, 14336]
-
blk.29.ffn_norm.weightF32[4096]
-
blk.29.ffn_up.weightQ4_K[4096, 14336]
-
blk.30.attn_k.weightQ4_K[4096, 1024]
-
blk.30.attn_norm.weightF32[4096]
-
blk.30.attn_output.weightQ4_K[4096, 4096]
-
blk.30.attn_q.weightQ4_K[4096, 4096]
-
blk.30.attn_v.weightQ6_K[4096, 1024]
-
blk.30.ffn_down.weightQ4_K[14336, 4096]
-
blk.30.ffn_gate.weightQ4_K[4096, 14336]
-
blk.30.ffn_norm.weightF32[4096]
-
blk.30.ffn_up.weightQ4_K[4096, 14336]
-
blk.31.attn_k.weightQ4_K[4096, 1024]
-
blk.31.attn_norm.weightF32[4096]
-
blk.31.attn_output.weightQ4_K[4096, 4096]
-
blk.31.attn_q.weightQ4_K[4096, 4096]
-
blk.31.attn_v.weightQ4_K[4096, 1024]
-
blk.31.ffn_down.weightQ6_K[14336, 4096]
-
blk.31.ffn_gate.weightQ4_K[4096, 14336]
-
blk.31.ffn_norm.weightF32[4096]
-
blk.31.ffn_up.weightQ4_K[4096, 14336]
-
blk.32.attn_k.weightQ4_K[4096, 1024]
-
blk.32.attn_norm.weightF32[4096]
-
blk.32.attn_output.weightQ4_K[4096, 4096]
-
blk.32.attn_q.weightQ4_K[4096, 4096]
-
blk.32.attn_v.weightQ4_K[4096, 1024]
-
blk.32.ffn_down.weightQ4_K[14336, 4096]
-
blk.32.ffn_gate.weightQ4_K[4096, 14336]
-
blk.32.ffn_norm.weightF32[4096]
-
blk.32.ffn_up.weightQ4_K[4096, 14336]
-
blk.33.attn_norm.weightF32[4096]
-
blk.33.cross_attn_attn_gateF32[1]
-
blk.33.cross_attn_k_norm.weightF32[128]
-
blk.33.cross_attn_k_proj.weightQ4_K[4096, 1024]
-
blk.33.cross_attn_mlp_gateF32[1]
-
blk.33.cross_attn_o_proj.weightQ4_K[4096, 4096]
-
blk.33.cross_attn_q_norm.weightF32[128]
-
blk.33.cross_attn_q_proj.weightQ4_K[4096, 4096]
-
blk.33.cross_attn_v_proj.weightQ4_K[4096, 1024]
-
blk.33.ffn_down.weightQ4_K[14336, 4096]
-
blk.33.ffn_gate.weightQ4_K[4096, 14336]
-
blk.33.ffn_norm.weightF32[4096]
-
blk.33.ffn_up.weightQ4_K[4096, 14336]
-
blk.34.attn_k.weightQ4_K[4096, 1024]
-
blk.34.attn_norm.weightF32[4096]
-
blk.34.attn_output.weightQ4_K[4096, 4096]
-
blk.34.attn_q.weightQ4_K[4096, 4096]
-
blk.34.attn_v.weightQ6_K[4096, 1024]
-
blk.34.ffn_down.weightQ6_K[14336, 4096]
-
blk.34.ffn_gate.weightQ4_K[4096, 14336]
-
blk.34.ffn_norm.weightF32[4096]
-
blk.34.ffn_up.weightQ4_K[4096, 14336]
-
blk.35.attn_k.weightQ4_K[4096, 1024]
-
blk.35.attn_norm.weightF32[4096]
-
blk.35.attn_output.weightQ4_K[4096, 4096]
-
blk.35.attn_q.weightQ4_K[4096, 4096]
-
blk.35.attn_v.weightQ6_K[4096, 1024]
-
blk.35.ffn_down.weightQ6_K[14336, 4096]
-
blk.35.ffn_gate.weightQ4_K[4096, 14336]
-
blk.35.ffn_norm.weightF32[4096]
-
blk.35.ffn_up.weightQ4_K[4096, 14336]
-
blk.36.attn_k.weightQ4_K[4096, 1024]
-
blk.36.attn_norm.weightF32[4096]
-
blk.36.attn_output.weightQ4_K[4096, 4096]
-
blk.36.attn_q.weightQ4_K[4096, 4096]
-
blk.36.attn_v.weightQ6_K[4096, 1024]
-
blk.36.ffn_down.weightQ6_K[14336, 4096]
-
blk.36.ffn_gate.weightQ4_K[4096, 14336]
-
blk.36.ffn_norm.weightF32[4096]
-
blk.36.ffn_up.weightQ4_K[4096, 14336]
-
blk.37.attn_k.weightQ4_K[4096, 1024]
-
blk.37.attn_norm.weightF32[4096]
-
blk.37.attn_output.weightQ4_K[4096, 4096]
-
blk.37.attn_q.weightQ4_K[4096, 4096]
-
blk.37.attn_v.weightQ6_K[4096, 1024]
-
blk.37.ffn_down.weightQ6_K[14336, 4096]
-
blk.37.ffn_gate.weightQ4_K[4096, 14336]
-
blk.37.ffn_norm.weightF32[4096]
-
blk.37.ffn_up.weightQ4_K[4096, 14336]
-
blk.38.attn_norm.weightF32[4096]
-
blk.38.cross_attn_attn_gateF32[1]
-
blk.38.cross_attn_k_norm.weightF32[128]
-
blk.38.cross_attn_k_proj.weightQ4_K[4096, 1024]
-
blk.38.cross_attn_mlp_gateF32[1]
-
blk.38.cross_attn_o_proj.weightQ4_K[4096, 4096]
-
blk.38.cross_attn_q_norm.weightF32[128]
-
blk.38.cross_attn_q_proj.weightQ4_K[4096, 4096]
-
blk.38.cross_attn_v_proj.weightQ4_K[4096, 1024]
-
blk.38.ffn_down.weightQ6_K[14336, 4096]
-
blk.38.ffn_gate.weightQ4_K[4096, 14336]
-
blk.38.ffn_norm.weightF32[4096]
-
blk.38.ffn_up.weightQ4_K[4096, 14336]
-
blk.39.attn_k.weightQ4_K[4096, 1024]
-
blk.39.attn_norm.weightF32[4096]
-
blk.39.attn_output.weightQ4_K[4096, 4096]
-
blk.39.attn_q.weightQ4_K[4096, 4096]
-
blk.39.attn_v.weightQ6_K[4096, 1024]
-
blk.39.ffn_down.weightQ6_K[14336, 4096]
-
blk.39.ffn_gate.weightQ4_K[4096, 14336]
-
blk.39.ffn_norm.weightF32[4096]
-
blk.39.ffn_up.weightQ4_K[4096, 14336]
-
output.weightQ6_K[4096, 128256]
-
rope_freqs.weightF32[64]
-
output_norm.weightF32[4096]
Metadata
Tensor
blk.0
blk.1
blk.2
blk.3
blk.4
blk.5
blk.6
blk.7
blk.8
blk.9
blk.10
blk.11
blk.12
blk.13
blk.14
blk.15
blk.16
blk.17
blk.18
blk.19
blk.20
blk.21
blk.22
blk.23
blk.24
blk.25
blk.26
blk.27
blk.28
blk.29
blk.30
blk.31
blk.32
blk.33
blk.34
blk.35
blk.36
blk.37
blk.38
blk.39