An open-source Mixture-of-Experts code language model that achieves performance comparable to GPT4-Turbo in code-specific tasks.
539 Pulls Updated 2 months ago
f51d8b174112 · 12GB
-
deepseek2.attention.head_count16
-
deepseek2.attention.head_count_kv16
-
deepseek2.attention.key_length192
-
deepseek2.attention.kv_lora_rank512
-
deepseek2.attention.layer_norm_rms_epsilon1e-06
-
deepseek2.attention.value_length128
-
deepseek2.block_count27
-
deepseek2.context_length163840
-
deepseek2.embedding_length2048
-
deepseek2.expert_count64
-
deepseek2.expert_feed_forward_length1408
-
deepseek2.expert_shared_count2
-
deepseek2.expert_used_count6
-
deepseek2.expert_weights_scale1
-
deepseek2.feed_forward_length10944
-
deepseek2.leading_dense_block_count1
-
deepseek2.rope.dimension_count64
-
deepseek2.rope.freq_base10000
-
deepseek2.rope.scaling.factor40
-
deepseek2.rope.scaling.original_context_length4096
-
deepseek2.rope.scaling.typeyarn
-
deepseek2.rope.scaling.yarn_log_multiplier0.0707
-
deepseek2.vocab_size102400
-
general.architecturedeepseek2
-
general.file_type9
-
general.name..
-
general.quantization_version2
-
general.typemodel
-
tokenizer.ggml.add_bos_tokentrue
-
tokenizer.ggml.add_eos_tokenfalse
-
tokenizer.ggml.bos_token_id100000
-
tokenizer.ggml.eos_token_id100001
-
tokenizer.ggml.merges[Ġ Ġ, Ġ t, Ġ a, i n, h e, ...]
-
tokenizer.ggml.modelgpt2
-
tokenizer.ggml.padding_token_id100001
-
tokenizer.ggml.predeepseek-llm
-
tokenizer.ggml.token_type[1, 1, 1, 1, 1, ...]
-
tokenizer.ggml.tokens[!, ", #, $, %, ...]
-
NameTypeShape
-
token_embd.weightQ5_1[2048, 102400]
-
blk.0.attn_kv_a_mqa.weightQ5_1[2048, 576]
-
blk.0.attn_kv_a_norm.weightF32[512]
-
blk.0.attn_kv_b.weightQ5_1[512, 4096]
-
blk.0.attn_norm.weightF32[2048]
-
blk.0.attn_output.weightQ5_1[2048, 2048]
-
blk.0.attn_q.weightQ5_1[2048, 3072]
-
blk.0.ffn_down.weightQ5_1[10944, 2048]
-
blk.0.ffn_gate.weightQ5_1[2048, 10944]
-
blk.0.ffn_norm.weightF32[2048]
-
blk.0.ffn_up.weightQ5_1[2048, 10944]
-
blk.1.attn_kv_a_mqa.weightQ5_1[2048, 576]
-
blk.1.attn_kv_a_norm.weightF32[512]
-
blk.1.attn_kv_b.weightQ5_1[512, 4096]
-
blk.1.attn_norm.weightF32[2048]
-
blk.1.attn_output.weightQ5_1[2048, 2048]
-
blk.1.attn_q.weightQ5_1[2048, 3072]
-
blk.1.ffn_down_exps.weightQ5_1[1408, 2048, 64]
-
blk.1.ffn_down_shexp.weightQ5_1[2816, 2048]
-
blk.1.ffn_gate_exps.weightQ5_1[2048, 1408, 64]
-
blk.1.ffn_gate_inp.weightF32[2048, 64]
-
blk.1.ffn_gate_shexp.weightQ5_1[2048, 2816]
-
blk.1.ffn_norm.weightF32[2048]
-
blk.1.ffn_up_exps.weightQ5_1[2048, 1408, 64]
-
blk.1.ffn_up_shexp.weightQ5_1[2048, 2816]
-
blk.2.attn_kv_a_mqa.weightQ5_1[2048, 576]
-
blk.2.attn_kv_a_norm.weightF32[512]
-
blk.2.attn_kv_b.weightQ5_1[512, 4096]
-
blk.2.attn_norm.weightF32[2048]
-
blk.2.attn_output.weightQ5_1[2048, 2048]
-
blk.2.attn_q.weightQ5_1[2048, 3072]
-
blk.2.ffn_down_exps.weightQ5_1[1408, 2048, 64]
-
blk.2.ffn_down_shexp.weightQ5_1[2816, 2048]
-
blk.2.ffn_gate_exps.weightQ5_1[2048, 1408, 64]
-
blk.2.ffn_gate_inp.weightF32[2048, 64]
-
blk.2.ffn_gate_shexp.weightQ5_1[2048, 2816]
-
blk.2.ffn_norm.weightF32[2048]
-
blk.2.ffn_up_exps.weightQ5_1[2048, 1408, 64]
-
blk.2.ffn_up_shexp.weightQ5_1[2048, 2816]
-
blk.3.attn_kv_a_mqa.weightQ5_1[2048, 576]
-
blk.3.attn_kv_a_norm.weightF32[512]
-
blk.3.attn_kv_b.weightQ5_1[512, 4096]
-
blk.3.attn_norm.weightF32[2048]
-
blk.3.attn_output.weightQ5_1[2048, 2048]
-
blk.3.attn_q.weightQ5_1[2048, 3072]
-
blk.3.ffn_down_exps.weightQ5_1[1408, 2048, 64]
-
blk.3.ffn_down_shexp.weightQ5_1[2816, 2048]
-
blk.3.ffn_gate_exps.weightQ5_1[2048, 1408, 64]
-
blk.3.ffn_gate_inp.weightF32[2048, 64]
-
blk.3.ffn_gate_shexp.weightQ5_1[2048, 2816]
-
blk.3.ffn_norm.weightF32[2048]
-
blk.3.ffn_up_exps.weightQ5_1[2048, 1408, 64]
-
blk.3.ffn_up_shexp.weightQ5_1[2048, 2816]
-
blk.4.attn_kv_a_mqa.weightQ5_1[2048, 576]
-
blk.4.attn_kv_a_norm.weightF32[512]
-
blk.4.attn_kv_b.weightQ5_1[512, 4096]
-
blk.4.attn_norm.weightF32[2048]
-
blk.4.attn_output.weightQ5_1[2048, 2048]
-
blk.4.attn_q.weightQ5_1[2048, 3072]
-
blk.4.ffn_down_exps.weightQ5_1[1408, 2048, 64]
-
blk.4.ffn_down_shexp.weightQ5_1[2816, 2048]
-
blk.4.ffn_gate_exps.weightQ5_1[2048, 1408, 64]
-
blk.4.ffn_gate_inp.weightF32[2048, 64]
-
blk.4.ffn_gate_shexp.weightQ5_1[2048, 2816]
-
blk.4.ffn_norm.weightF32[2048]
-
blk.4.ffn_up_exps.weightQ5_1[2048, 1408, 64]
-
blk.4.ffn_up_shexp.weightQ5_1[2048, 2816]
-
blk.5.attn_kv_a_mqa.weightQ5_1[2048, 576]
-
blk.5.attn_kv_a_norm.weightF32[512]
-
blk.5.attn_kv_b.weightQ5_1[512, 4096]
-
blk.5.attn_norm.weightF32[2048]
-
blk.5.attn_output.weightQ5_1[2048, 2048]
-
blk.5.attn_q.weightQ5_1[2048, 3072]
-
blk.5.ffn_down_exps.weightQ5_1[1408, 2048, 64]
-
blk.5.ffn_down_shexp.weightQ5_1[2816, 2048]
-
blk.5.ffn_gate_exps.weightQ5_1[2048, 1408, 64]
-
blk.5.ffn_gate_inp.weightF32[2048, 64]
-
blk.5.ffn_gate_shexp.weightQ5_1[2048, 2816]
-
blk.5.ffn_norm.weightF32[2048]
-
blk.5.ffn_up_exps.weightQ5_1[2048, 1408, 64]
-
blk.5.ffn_up_shexp.weightQ5_1[2048, 2816]
-
blk.6.attn_kv_a_mqa.weightQ5_1[2048, 576]
-
blk.6.attn_kv_a_norm.weightF32[512]
-
blk.6.attn_kv_b.weightQ5_1[512, 4096]
-
blk.6.attn_norm.weightF32[2048]
-
blk.6.attn_output.weightQ5_1[2048, 2048]
-
blk.6.attn_q.weightQ5_1[2048, 3072]
-
blk.6.ffn_down_exps.weightQ5_1[1408, 2048, 64]
-
blk.6.ffn_down_shexp.weightQ5_1[2816, 2048]
-
blk.6.ffn_gate_exps.weightQ5_1[2048, 1408, 64]
-
blk.6.ffn_gate_inp.weightF32[2048, 64]
-
blk.6.ffn_gate_shexp.weightQ5_1[2048, 2816]
-
blk.6.ffn_norm.weightF32[2048]
-
blk.6.ffn_up_exps.weightQ5_1[2048, 1408, 64]
-
blk.6.ffn_up_shexp.weightQ5_1[2048, 2816]
-
blk.7.attn_kv_a_mqa.weightQ5_1[2048, 576]
-
blk.7.attn_kv_a_norm.weightF32[512]
-
blk.7.attn_kv_b.weightQ5_1[512, 4096]
-
blk.7.attn_norm.weightF32[2048]
-
blk.7.attn_output.weightQ5_1[2048, 2048]
-
blk.7.attn_q.weightQ5_1[2048, 3072]
-
blk.7.ffn_down_exps.weightQ5_1[1408, 2048, 64]
-
blk.7.ffn_down_shexp.weightQ5_1[2816, 2048]
-
blk.7.ffn_gate_exps.weightQ5_1[2048, 1408, 64]
-
blk.7.ffn_gate_inp.weightF32[2048, 64]
-
blk.7.ffn_gate_shexp.weightQ5_1[2048, 2816]
-
blk.7.ffn_norm.weightF32[2048]
-
blk.7.ffn_up_exps.weightQ5_1[2048, 1408, 64]
-
blk.7.ffn_up_shexp.weightQ5_1[2048, 2816]
-
blk.8.attn_kv_a_mqa.weightQ5_1[2048, 576]
-
blk.8.attn_kv_a_norm.weightF32[512]
-
blk.8.attn_kv_b.weightQ5_1[512, 4096]
-
blk.8.attn_norm.weightF32[2048]
-
blk.8.attn_output.weightQ5_1[2048, 2048]
-
blk.8.attn_q.weightQ5_1[2048, 3072]
-
blk.8.ffn_down_exps.weightQ5_1[1408, 2048, 64]
-
blk.8.ffn_down_shexp.weightQ5_1[2816, 2048]
-
blk.8.ffn_gate_exps.weightQ5_1[2048, 1408, 64]
-
blk.8.ffn_gate_inp.weightF32[2048, 64]
-
blk.8.ffn_gate_shexp.weightQ5_1[2048, 2816]
-
blk.8.ffn_norm.weightF32[2048]
-
blk.8.ffn_up_exps.weightQ5_1[2048, 1408, 64]
-
blk.8.ffn_up_shexp.weightQ5_1[2048, 2816]
-
blk.9.attn_kv_a_mqa.weightQ5_1[2048, 576]
-
blk.9.attn_kv_a_norm.weightF32[512]
-
blk.9.attn_kv_b.weightQ5_1[512, 4096]
-
blk.9.attn_norm.weightF32[2048]
-
blk.9.attn_output.weightQ5_1[2048, 2048]
-
blk.9.attn_q.weightQ5_1[2048, 3072]
-
blk.9.ffn_down_exps.weightQ5_1[1408, 2048, 64]
-
blk.9.ffn_down_shexp.weightQ5_1[2816, 2048]
-
blk.9.ffn_gate_exps.weightQ5_1[2048, 1408, 64]
-
blk.9.ffn_gate_inp.weightF32[2048, 64]
-
blk.9.ffn_gate_shexp.weightQ5_1[2048, 2816]
-
blk.9.ffn_norm.weightF32[2048]
-
blk.9.ffn_up_exps.weightQ5_1[2048, 1408, 64]
-
blk.9.ffn_up_shexp.weightQ5_1[2048, 2816]
-
blk.10.attn_kv_a_mqa.weightQ5_1[2048, 576]
-
blk.10.attn_kv_a_norm.weightF32[512]
-
blk.10.attn_kv_b.weightQ5_1[512, 4096]
-
blk.10.attn_norm.weightF32[2048]
-
blk.10.attn_output.weightQ5_1[2048, 2048]
-
blk.10.attn_q.weightQ5_1[2048, 3072]
-
blk.10.ffn_down_exps.weightQ5_1[1408, 2048, 64]
-
blk.10.ffn_down_shexp.weightQ5_1[2816, 2048]
-
blk.10.ffn_gate_exps.weightQ5_1[2048, 1408, 64]
-
blk.10.ffn_gate_inp.weightF32[2048, 64]
-
blk.10.ffn_gate_shexp.weightQ5_1[2048, 2816]
-
blk.10.ffn_norm.weightF32[2048]
-
blk.10.ffn_up_exps.weightQ5_1[2048, 1408, 64]
-
blk.10.ffn_up_shexp.weightQ5_1[2048, 2816]
-
blk.11.attn_kv_a_mqa.weightQ5_1[2048, 576]
-
blk.11.attn_kv_a_norm.weightF32[512]
-
blk.11.attn_kv_b.weightQ5_1[512, 4096]
-
blk.11.attn_norm.weightF32[2048]
-
blk.11.attn_output.weightQ5_1[2048, 2048]
-
blk.11.attn_q.weightQ5_1[2048, 3072]
-
blk.11.ffn_down_exps.weightQ5_1[1408, 2048, 64]
-
blk.11.ffn_down_shexp.weightQ5_1[2816, 2048]
-
blk.11.ffn_gate_exps.weightQ5_1[2048, 1408, 64]
-
blk.11.ffn_gate_inp.weightF32[2048, 64]
-
blk.11.ffn_gate_shexp.weightQ5_1[2048, 2816]
-
blk.11.ffn_norm.weightF32[2048]
-
blk.11.ffn_up_exps.weightQ5_1[2048, 1408, 64]
-
blk.11.ffn_up_shexp.weightQ5_1[2048, 2816]
-
blk.12.attn_kv_a_mqa.weightQ5_1[2048, 576]
-
blk.12.attn_kv_a_norm.weightF32[512]
-
blk.12.attn_kv_b.weightQ5_1[512, 4096]
-
blk.12.attn_norm.weightF32[2048]
-
blk.12.attn_output.weightQ5_1[2048, 2048]
-
blk.12.attn_q.weightQ5_1[2048, 3072]
-
blk.12.ffn_down_exps.weightQ5_1[1408, 2048, 64]
-
blk.12.ffn_down_shexp.weightQ5_1[2816, 2048]
-
blk.12.ffn_gate_exps.weightQ5_1[2048, 1408, 64]
-
blk.12.ffn_gate_inp.weightF32[2048, 64]
-
blk.12.ffn_gate_shexp.weightQ5_1[2048, 2816]
-
blk.12.ffn_norm.weightF32[2048]
-
blk.12.ffn_up_exps.weightQ5_1[2048, 1408, 64]
-
blk.12.ffn_up_shexp.weightQ5_1[2048, 2816]
-
blk.13.attn_kv_a_mqa.weightQ5_1[2048, 576]
-
blk.13.attn_kv_a_norm.weightF32[512]
-
blk.13.attn_kv_b.weightQ5_1[512, 4096]
-
blk.13.attn_norm.weightF32[2048]
-
blk.13.attn_output.weightQ5_1[2048, 2048]
-
blk.13.attn_q.weightQ5_1[2048, 3072]
-
blk.13.ffn_down_exps.weightQ5_1[1408, 2048, 64]
-
blk.13.ffn_down_shexp.weightQ5_1[2816, 2048]
-
blk.13.ffn_gate_exps.weightQ5_1[2048, 1408, 64]
-
blk.13.ffn_gate_inp.weightF32[2048, 64]
-
blk.13.ffn_gate_shexp.weightQ5_1[2048, 2816]
-
blk.13.ffn_norm.weightF32[2048]
-
blk.13.ffn_up_exps.weightQ5_1[2048, 1408, 64]
-
blk.13.ffn_up_shexp.weightQ5_1[2048, 2816]
-
blk.14.attn_kv_a_mqa.weightQ5_1[2048, 576]
-
blk.14.attn_kv_a_norm.weightF32[512]
-
blk.14.attn_kv_b.weightQ5_1[512, 4096]
-
blk.14.attn_norm.weightF32[2048]
-
blk.14.attn_output.weightQ5_1[2048, 2048]
-
blk.14.attn_q.weightQ5_1[2048, 3072]
-
blk.14.ffn_down_exps.weightQ5_1[1408, 2048, 64]
-
blk.14.ffn_down_shexp.weightQ5_1[2816, 2048]
-
blk.14.ffn_gate_exps.weightQ5_1[2048, 1408, 64]
-
blk.14.ffn_gate_inp.weightF32[2048, 64]
-
blk.14.ffn_gate_shexp.weightQ5_1[2048, 2816]
-
blk.14.ffn_norm.weightF32[2048]
-
blk.14.ffn_up_exps.weightQ5_1[2048, 1408, 64]
-
blk.14.ffn_up_shexp.weightQ5_1[2048, 2816]
-
blk.15.attn_kv_a_mqa.weightQ5_1[2048, 576]
-
blk.15.attn_kv_a_norm.weightF32[512]
-
blk.15.attn_kv_b.weightQ5_1[512, 4096]
-
blk.15.attn_norm.weightF32[2048]
-
blk.15.attn_output.weightQ5_1[2048, 2048]
-
blk.15.attn_q.weightQ5_1[2048, 3072]
-
blk.15.ffn_down_exps.weightQ5_1[1408, 2048, 64]
-
blk.15.ffn_down_shexp.weightQ5_1[2816, 2048]
-
blk.15.ffn_gate_exps.weightQ5_1[2048, 1408, 64]
-
blk.15.ffn_gate_inp.weightF32[2048, 64]
-
blk.15.ffn_gate_shexp.weightQ5_1[2048, 2816]
-
blk.15.ffn_norm.weightF32[2048]
-
blk.15.ffn_up_exps.weightQ5_1[2048, 1408, 64]
-
blk.15.ffn_up_shexp.weightQ5_1[2048, 2816]
-
blk.16.attn_kv_a_mqa.weightQ5_1[2048, 576]
-
blk.16.attn_kv_a_norm.weightF32[512]
-
blk.16.attn_kv_b.weightQ5_1[512, 4096]
-
blk.16.attn_norm.weightF32[2048]
-
blk.16.attn_output.weightQ5_1[2048, 2048]
-
blk.16.attn_q.weightQ5_1[2048, 3072]
-
blk.16.ffn_down_exps.weightQ5_1[1408, 2048, 64]
-
blk.16.ffn_down_shexp.weightQ5_1[2816, 2048]
-
blk.16.ffn_gate_exps.weightQ5_1[2048, 1408, 64]
-
blk.16.ffn_gate_inp.weightF32[2048, 64]
-
blk.16.ffn_gate_shexp.weightQ5_1[2048, 2816]
-
blk.16.ffn_norm.weightF32[2048]
-
blk.16.ffn_up_exps.weightQ5_1[2048, 1408, 64]
-
blk.16.ffn_up_shexp.weightQ5_1[2048, 2816]
-
blk.17.attn_kv_a_mqa.weightQ5_1[2048, 576]
-
blk.17.attn_kv_a_norm.weightF32[512]
-
blk.17.attn_kv_b.weightQ5_1[512, 4096]
-
blk.17.attn_norm.weightF32[2048]
-
blk.17.attn_output.weightQ5_1[2048, 2048]
-
blk.17.attn_q.weightQ5_1[2048, 3072]
-
blk.17.ffn_down_exps.weightQ5_1[1408, 2048, 64]
-
blk.17.ffn_down_shexp.weightQ5_1[2816, 2048]
-
blk.17.ffn_gate_exps.weightQ5_1[2048, 1408, 64]
-
blk.17.ffn_gate_inp.weightF32[2048, 64]
-
blk.17.ffn_gate_shexp.weightQ5_1[2048, 2816]
-
blk.17.ffn_norm.weightF32[2048]
-
blk.17.ffn_up_exps.weightQ5_1[2048, 1408, 64]
-
blk.17.ffn_up_shexp.weightQ5_1[2048, 2816]
-
blk.18.attn_kv_a_mqa.weightQ5_1[2048, 576]
-
blk.18.attn_kv_a_norm.weightF32[512]
-
blk.18.attn_kv_b.weightQ5_1[512, 4096]
-
blk.18.attn_norm.weightF32[2048]
-
blk.18.attn_output.weightQ5_1[2048, 2048]
-
blk.18.attn_q.weightQ5_1[2048, 3072]
-
blk.18.ffn_down_exps.weightQ5_1[1408, 2048, 64]
-
blk.18.ffn_down_shexp.weightQ5_1[2816, 2048]
-
blk.18.ffn_gate_exps.weightQ5_1[2048, 1408, 64]
-
blk.18.ffn_gate_inp.weightF32[2048, 64]
-
blk.18.ffn_gate_shexp.weightQ5_1[2048, 2816]
-
blk.18.ffn_norm.weightF32[2048]
-
blk.18.ffn_up_exps.weightQ5_1[2048, 1408, 64]
-
blk.18.ffn_up_shexp.weightQ5_1[2048, 2816]
-
blk.19.attn_kv_a_mqa.weightQ5_1[2048, 576]
-
blk.19.attn_kv_a_norm.weightF32[512]
-
blk.19.attn_kv_b.weightQ5_1[512, 4096]
-
blk.19.attn_norm.weightF32[2048]
-
blk.19.attn_output.weightQ5_1[2048, 2048]
-
blk.19.attn_q.weightQ5_1[2048, 3072]
-
blk.19.ffn_down_exps.weightQ5_1[1408, 2048, 64]
-
blk.19.ffn_down_shexp.weightQ5_1[2816, 2048]
-
blk.19.ffn_gate_exps.weightQ5_1[2048, 1408, 64]
-
blk.19.ffn_gate_inp.weightF32[2048, 64]
-
blk.19.ffn_gate_shexp.weightQ5_1[2048, 2816]
-
blk.19.ffn_norm.weightF32[2048]
-
blk.19.ffn_up_exps.weightQ5_1[2048, 1408, 64]
-
blk.19.ffn_up_shexp.weightQ5_1[2048, 2816]
-
blk.20.attn_kv_a_mqa.weightQ5_1[2048, 576]
-
blk.20.attn_kv_a_norm.weightF32[512]
-
blk.20.attn_kv_b.weightQ5_1[512, 4096]
-
blk.20.attn_norm.weightF32[2048]
-
blk.20.attn_output.weightQ5_1[2048, 2048]
-
blk.20.attn_q.weightQ5_1[2048, 3072]
-
blk.20.ffn_down_exps.weightQ5_1[1408, 2048, 64]
-
blk.20.ffn_down_shexp.weightQ5_1[2816, 2048]
-
blk.20.ffn_gate_exps.weightQ5_1[2048, 1408, 64]
-
blk.20.ffn_gate_inp.weightF32[2048, 64]
-
blk.20.ffn_gate_shexp.weightQ5_1[2048, 2816]
-
blk.20.ffn_norm.weightF32[2048]
-
blk.20.ffn_up_exps.weightQ5_1[2048, 1408, 64]
-
blk.20.ffn_up_shexp.weightQ5_1[2048, 2816]
-
blk.21.attn_kv_a_mqa.weightQ5_1[2048, 576]
-
blk.21.attn_kv_a_norm.weightF32[512]
-
blk.21.attn_kv_b.weightQ5_1[512, 4096]
-
blk.21.attn_norm.weightF32[2048]
-
blk.21.attn_output.weightQ5_1[2048, 2048]
-
blk.21.attn_q.weightQ5_1[2048, 3072]
-
blk.21.ffn_down_exps.weightQ5_1[1408, 2048, 64]
-
blk.21.ffn_down_shexp.weightQ5_1[2816, 2048]
-
blk.21.ffn_gate_exps.weightQ5_1[2048, 1408, 64]
-
blk.21.ffn_gate_inp.weightF32[2048, 64]
-
blk.21.ffn_gate_shexp.weightQ5_1[2048, 2816]
-
blk.21.ffn_norm.weightF32[2048]
-
blk.21.ffn_up_exps.weightQ5_1[2048, 1408, 64]
-
blk.21.ffn_up_shexp.weightQ5_1[2048, 2816]
-
blk.22.attn_kv_a_mqa.weightQ5_1[2048, 576]
-
blk.22.attn_kv_a_norm.weightF32[512]
-
blk.22.attn_kv_b.weightQ5_1[512, 4096]
-
blk.22.attn_norm.weightF32[2048]
-
blk.22.attn_output.weightQ5_1[2048, 2048]
-
blk.22.attn_q.weightQ5_1[2048, 3072]
-
blk.22.ffn_down_exps.weightQ5_1[1408, 2048, 64]
-
blk.22.ffn_down_shexp.weightQ5_1[2816, 2048]
-
blk.22.ffn_gate_exps.weightQ5_1[2048, 1408, 64]
-
blk.22.ffn_gate_inp.weightF32[2048, 64]
-
blk.22.ffn_gate_shexp.weightQ5_1[2048, 2816]
-
blk.22.ffn_norm.weightF32[2048]
-
blk.22.ffn_up_exps.weightQ5_1[2048, 1408, 64]
-
blk.22.ffn_up_shexp.weightQ5_1[2048, 2816]
-
blk.23.attn_kv_a_mqa.weightQ5_1[2048, 576]
-
blk.23.attn_kv_a_norm.weightF32[512]
-
blk.23.attn_kv_b.weightQ5_1[512, 4096]
-
blk.23.attn_norm.weightF32[2048]
-
blk.23.attn_output.weightQ5_1[2048, 2048]
-
blk.23.attn_q.weightQ5_1[2048, 3072]
-
blk.23.ffn_down_exps.weightQ5_1[1408, 2048, 64]
-
blk.23.ffn_down_shexp.weightQ5_1[2816, 2048]
-
blk.23.ffn_gate_exps.weightQ5_1[2048, 1408, 64]
-
blk.23.ffn_gate_inp.weightF32[2048, 64]
-
blk.23.ffn_gate_shexp.weightQ5_1[2048, 2816]
-
blk.23.ffn_norm.weightF32[2048]
-
blk.23.ffn_up_exps.weightQ5_1[2048, 1408, 64]
-
blk.23.ffn_up_shexp.weightQ5_1[2048, 2816]
-
blk.24.attn_kv_a_mqa.weightQ5_1[2048, 576]
-
blk.24.attn_kv_a_norm.weightF32[512]
-
blk.24.attn_kv_b.weightQ5_1[512, 4096]
-
blk.24.attn_norm.weightF32[2048]
-
blk.24.attn_output.weightQ5_1[2048, 2048]
-
blk.24.attn_q.weightQ5_1[2048, 3072]
-
blk.24.ffn_down_exps.weightQ5_1[1408, 2048, 64]
-
blk.24.ffn_down_shexp.weightQ5_1[2816, 2048]
-
blk.24.ffn_gate_exps.weightQ5_1[2048, 1408, 64]
-
blk.24.ffn_gate_inp.weightF32[2048, 64]
-
blk.24.ffn_gate_shexp.weightQ5_1[2048, 2816]
-
blk.24.ffn_norm.weightF32[2048]
-
blk.24.ffn_up_exps.weightQ5_1[2048, 1408, 64]
-
blk.24.ffn_up_shexp.weightQ5_1[2048, 2816]
-
blk.25.attn_kv_a_mqa.weightQ5_1[2048, 576]
-
blk.25.attn_kv_a_norm.weightF32[512]
-
blk.25.attn_kv_b.weightQ5_1[512, 4096]
-
blk.25.attn_norm.weightF32[2048]
-
blk.25.attn_output.weightQ5_1[2048, 2048]
-
blk.25.attn_q.weightQ5_1[2048, 3072]
-
blk.25.ffn_down_exps.weightQ5_1[1408, 2048, 64]
-
blk.25.ffn_down_shexp.weightQ5_1[2816, 2048]
-
blk.25.ffn_gate_exps.weightQ5_1[2048, 1408, 64]
-
blk.25.ffn_gate_inp.weightF32[2048, 64]
-
blk.25.ffn_gate_shexp.weightQ5_1[2048, 2816]
-
blk.25.ffn_norm.weightF32[2048]
-
blk.25.ffn_up_exps.weightQ5_1[2048, 1408, 64]
-
blk.25.ffn_up_shexp.weightQ5_1[2048, 2816]
-
blk.26.attn_kv_a_mqa.weightQ5_1[2048, 576]
-
blk.26.attn_kv_a_norm.weightF32[512]
-
blk.26.attn_kv_b.weightQ5_1[512, 4096]
-
blk.26.attn_norm.weightF32[2048]
-
blk.26.attn_output.weightQ5_1[2048, 2048]
-
blk.26.attn_q.weightQ5_1[2048, 3072]
-
blk.26.ffn_down_exps.weightQ5_1[1408, 2048, 64]
-
blk.26.ffn_down_shexp.weightQ5_1[2816, 2048]
-
blk.26.ffn_gate_exps.weightQ5_1[2048, 1408, 64]
-
blk.26.ffn_gate_inp.weightF32[2048, 64]
-
blk.26.ffn_gate_shexp.weightQ5_1[2048, 2816]
-
blk.26.ffn_norm.weightF32[2048]
-
blk.26.ffn_up_exps.weightQ5_1[2048, 1408, 64]
-
blk.26.ffn_up_shexp.weightQ5_1[2048, 2816]
-
output.weightQ6_K[2048, 102400]
-
output_norm.weightF32[2048]
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