Models
Docs
Pricing
Sign in
Download
Models
Download
Docs
Pricing
Sign in
frob
/
glm-5
:744b-a40b-ud-q4_K_XL
250
Downloads
Updated
1 month ago
Cancel
tools
thinking
glm-5:744b-a40b-ud-q4_K_XL
...
/
model
f8d1d6ee112c · 431GB
Metadata
general.architecture
glm-dsa
glm-dsa
general.file_type
Q4_K_M
Q4_K_M
glm-dsa.attention.head_count
64
64
glm-dsa.attention.head_count_kv
1
1
glm-dsa.attention.indexer.head_count
32
32
glm-dsa.attention.indexer.key_length
128
128
glm-dsa.attention.indexer.top_k
2048
2048
glm-dsa.attention.key_length
576
576
glm-dsa.attention.key_length_mla
256
256
glm-dsa.attention.kv_lora_rank
512
512
glm-dsa.attention.layer_norm_rms_epsilon
1e-05
1e-05
glm-dsa.attention.q_lora_rank
2048
2048
glm-dsa.attention.value_length
512
512
glm-dsa.attention.value_length_mla
256
256
glm-dsa.block_count
79
79
glm-dsa.context_length
202752
202752
glm-dsa.embedding_length
6144
6144
glm-dsa.expert_count
256
256
glm-dsa.expert_feed_forward_length
2048
2048
glm-dsa.expert_gating_func
2
2
glm-dsa.expert_group_count
1
1
glm-dsa.expert_group_used_count
1
1
glm-dsa.expert_shared_count
1
1
glm-dsa.expert_used_count
8
8
glm-dsa.expert_weights_norm
true
true
glm-dsa.expert_weights_scale
2.5
2.5
glm-dsa.feed_forward_length
12288
12288
glm-dsa.leading_dense_block_count
3
3
glm-dsa.nextn_predict_layers
1
1
glm-dsa.rope.dimension_count
64
64
glm-dsa.rope.freq_base
1e+06
1e+06
glm-dsa.vocab_size
154880
154880
tokenizer.ggml.bos_token_id
154822
154822
tokenizer.ggml.eom_token_id
154829
154829
tokenizer.ggml.eos_token_id
154820
154820
tokenizer.ggml.eot_token_id
154827
154827
tokenizer.ggml.merges
[Ġ Ġ, Ġ ĠĠĠ, ĠĠ ĠĠ, ĠĠĠ Ġ, i n, ...]
[Ġ Ġ, Ġ ĠĠĠ, ĠĠ ĠĠ, ĠĠĠ Ġ, i n, ...]
tokenizer.ggml.model
gpt2
gpt2
tokenizer.ggml.padding_token_id
154821
154821
tokenizer.ggml.pre
glm4
glm4
tokenizer.ggml.token_type
[1, 1, 1, 1, 1, ...]
[1, 1, 1, 1, 1, ...]
tokenizer.ggml.tokens
[!, ", #, $, %, ...]
[!, ", #, $, %, ...]
tokenizer.ggml.unknown_token_id
154820
154820
quantize.imatrix.chunks_count
85
85
quantize.imatrix.dataset
unsloth_calibration_GLM-5.txt
unsloth_calibration_GLM-5.txt
quantize.imatrix.entries_count
1002
1002
quantize.imatrix.file
GLM-5-GGUF/imatrix_unsloth.gguf
GLM-5-GGUF/imatrix_unsloth.gguf
split.count
0
0
split.no
0
0
split.tensors.count
1809
1809
Tensor
Name
Type
Shape
token_embd.weight
Q4_K
Q4_K
[6144, 154880]
blk.0
blk.0.attn_k_b.weight
Q8_0
Q8_0
[192, 512, 64]
blk.0.attn_kv_a_mqa.weight
Q8_0
Q8_0
[6144, 576]
blk.0.attn_kv_a_norm.weight
F32
F32
[512]
blk.0.attn_norm.weight
F32
F32
[6144]
blk.0.attn_output.weight
Q4_K
Q4_K
[16384, 6144]
blk.0.attn_q_a.weight
Q4_K
Q4_K
[6144, 2048]
blk.0.attn_q_a_norm.weight
F32
F32
[2048]
blk.0.attn_q_b.weight
Q8_0
Q8_0
[2048, 16384]
blk.0.attn_v_b.weight
Q8_0
Q8_0
[512, 256, 64]
blk.0.ffn_down.weight
Q4_K
Q4_K
[12288, 6144]
blk.0.ffn_gate.weight
Q4_K
Q4_K
[6144, 12288]
blk.0.ffn_norm.weight
F32
F32
[6144]
blk.0.ffn_up.weight
Q4_K
Q4_K
[6144, 12288]
blk.0.indexer.attn_k.weight
Q4_K
Q4_K
[6144, 128]
blk.0.indexer.attn_q_b.weight
Q4_K
Q4_K
[2048, 4096]
blk.0.indexer.k_norm.bias
F32
F32
[128]
blk.0.indexer.k_norm.weight
F32
F32
[128]
blk.0.indexer.proj.weight
Q4_K
Q4_K
[6144, 32]
blk.1
blk.1.attn_k_b.weight
Q8_0
Q8_0
[192, 512, 64]
blk.1.attn_kv_a_mqa.weight
Q8_0
Q8_0
[6144, 576]
blk.1.attn_kv_a_norm.weight
F32
F32
[512]
blk.1.attn_norm.weight
F32
F32
[6144]
blk.1.attn_output.weight
Q5_K
Q5_K
[16384, 6144]
blk.1.attn_q_a.weight
Q5_K
Q5_K
[6144, 2048]
blk.1.attn_q_a_norm.weight
F32
F32
[2048]
blk.1.attn_q_b.weight
Q8_0
Q8_0
[2048, 16384]
blk.1.attn_v_b.weight
Q8_0
Q8_0
[512, 256, 64]
blk.1.ffn_down.weight
Q5_K
Q5_K
[12288, 6144]
blk.1.ffn_gate.weight
Q4_K
Q4_K
[6144, 12288]
blk.1.ffn_norm.weight
F32
F32
[6144]
blk.1.ffn_up.weight
Q4_K
Q4_K
[6144, 12288]
blk.1.indexer.attn_k.weight
Q4_K
Q4_K
[6144, 128]
blk.1.indexer.attn_q_b.weight
Q4_K
Q4_K
[2048, 4096]
blk.1.indexer.k_norm.bias
F32
F32
[128]
blk.1.indexer.k_norm.weight
F32
F32
[128]
blk.1.indexer.proj.weight
Q4_K
Q4_K
[6144, 32]
blk.2
blk.2.attn_k_b.weight
Q8_0
Q8_0
[192, 512, 64]
blk.2.attn_kv_a_mqa.weight
Q8_0
Q8_0
[6144, 576]
blk.2.attn_kv_a_norm.weight
F32
F32
[512]
blk.2.attn_norm.weight
F32
F32
[6144]
blk.2.attn_output.weight
Q4_K
Q4_K
[16384, 6144]
blk.2.attn_q_a.weight
Q4_K
Q4_K
[6144, 2048]
blk.2.attn_q_a_norm.weight
F32
F32
[2048]
blk.2.attn_q_b.weight
Q8_0
Q8_0
[2048, 16384]
blk.2.attn_v_b.weight
Q8_0
Q8_0
[512, 256, 64]
blk.2.ffn_down.weight
Q4_K
Q4_K
[12288, 6144]
blk.2.ffn_gate.weight
Q4_K
Q4_K
[6144, 12288]
blk.2.ffn_norm.weight
F32
F32
[6144]
blk.2.ffn_up.weight
Q4_K
Q4_K
[6144, 12288]
blk.2.indexer.attn_k.weight
Q4_K
Q4_K
[6144, 128]
blk.2.indexer.attn_q_b.weight
Q4_K
Q4_K
[2048, 4096]
blk.2.indexer.k_norm.bias
F32
F32
[128]
blk.2.indexer.k_norm.weight
F32
F32
[128]
blk.2.indexer.proj.weight
Q4_K
Q4_K
[6144, 32]
blk.3
blk.3.attn_k_b.weight
Q8_0
Q8_0
[192, 512, 64]
blk.3.attn_kv_a_mqa.weight
Q8_0
Q8_0
[6144, 576]
blk.3.attn_kv_a_norm.weight
F32
F32
[512]
blk.3.attn_norm.weight
F32
F32
[6144]
blk.3.attn_output.weight
Q4_K
Q4_K
[16384, 6144]
blk.3.attn_q_a.weight
Q4_K
Q4_K
[6144, 2048]
blk.3.attn_q_a_norm.weight
F32
F32
[2048]
blk.3.attn_q_b.weight
Q8_0
Q8_0
[2048, 16384]
blk.3.attn_v_b.weight
Q8_0
Q8_0
[512, 256, 64]
blk.3.exp_probs_b.bias
F32
F32
[256]
blk.3.ffn_down_exps.weight
Q5_K
Q5_K
[2048, 6144, 256]
blk.3.ffn_down_shexp.weight
Q6_K
Q6_K
[2048, 6144]
blk.3.ffn_gate_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.3.ffn_gate_inp.weight
F32
F32
[6144, 256]
blk.3.ffn_gate_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.3.ffn_norm.weight
F32
F32
[6144]
blk.3.ffn_up_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.3.ffn_up_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.3.indexer.attn_k.weight
Q4_K
Q4_K
[6144, 128]
blk.3.indexer.attn_q_b.weight
Q4_K
Q4_K
[2048, 4096]
blk.3.indexer.k_norm.bias
F32
F32
[128]
blk.3.indexer.k_norm.weight
F32
F32
[128]
blk.3.indexer.proj.weight
Q4_K
Q4_K
[6144, 32]
blk.4
blk.4.attn_k_b.weight
Q8_0
Q8_0
[192, 512, 64]
blk.4.attn_kv_a_mqa.weight
Q8_0
Q8_0
[6144, 576]
blk.4.attn_kv_a_norm.weight
F32
F32
[512]
blk.4.attn_norm.weight
F32
F32
[6144]
blk.4.attn_output.weight
Q4_K
Q4_K
[16384, 6144]
blk.4.attn_q_a.weight
Q4_K
Q4_K
[6144, 2048]
blk.4.attn_q_a_norm.weight
F32
F32
[2048]
blk.4.attn_q_b.weight
Q8_0
Q8_0
[2048, 16384]
blk.4.attn_v_b.weight
Q8_0
Q8_0
[512, 256, 64]
blk.4.exp_probs_b.bias
F32
F32
[256]
blk.4.ffn_down_exps.weight
Q4_K
Q4_K
[2048, 6144, 256]
blk.4.ffn_down_shexp.weight
Q6_K
Q6_K
[2048, 6144]
blk.4.ffn_gate_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.4.ffn_gate_inp.weight
F32
F32
[6144, 256]
blk.4.ffn_gate_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.4.ffn_norm.weight
F32
F32
[6144]
blk.4.ffn_up_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.4.ffn_up_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.4.indexer.attn_k.weight
Q4_K
Q4_K
[6144, 128]
blk.4.indexer.attn_q_b.weight
Q4_K
Q4_K
[2048, 4096]
blk.4.indexer.k_norm.bias
F32
F32
[128]
blk.4.indexer.k_norm.weight
F32
F32
[128]
blk.4.indexer.proj.weight
Q4_K
Q4_K
[6144, 32]
blk.5
blk.5.attn_k_b.weight
Q8_0
Q8_0
[192, 512, 64]
blk.5.attn_kv_a_mqa.weight
Q8_0
Q8_0
[6144, 576]
blk.5.attn_kv_a_norm.weight
F32
F32
[512]
blk.5.attn_norm.weight
F32
F32
[6144]
blk.5.attn_output.weight
Q4_K
Q4_K
[16384, 6144]
blk.5.attn_q_a.weight
Q4_K
Q4_K
[6144, 2048]
blk.5.attn_q_a_norm.weight
F32
F32
[2048]
blk.5.attn_q_b.weight
Q6_K
Q6_K
[2048, 16384]
blk.5.attn_v_b.weight
Q8_0
Q8_0
[512, 256, 64]
blk.5.exp_probs_b.bias
F32
F32
[256]
blk.5.ffn_down_exps.weight
Q4_K
Q4_K
[2048, 6144, 256]
blk.5.ffn_down_shexp.weight
Q6_K
Q6_K
[2048, 6144]
blk.5.ffn_gate_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.5.ffn_gate_inp.weight
F32
F32
[6144, 256]
blk.5.ffn_gate_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.5.ffn_norm.weight
F32
F32
[6144]
blk.5.ffn_up_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.5.ffn_up_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.5.indexer.attn_k.weight
Q4_K
Q4_K
[6144, 128]
blk.5.indexer.attn_q_b.weight
Q4_K
Q4_K
[2048, 4096]
blk.5.indexer.k_norm.bias
F32
F32
[128]
blk.5.indexer.k_norm.weight
F32
F32
[128]
blk.5.indexer.proj.weight
Q4_K
Q4_K
[6144, 32]
blk.6
blk.6.attn_k_b.weight
Q8_0
Q8_0
[192, 512, 64]
blk.6.attn_kv_a_mqa.weight
Q8_0
Q8_0
[6144, 576]
blk.6.attn_kv_a_norm.weight
F32
F32
[512]
blk.6.attn_norm.weight
F32
F32
[6144]
blk.6.attn_output.weight
Q5_K
Q5_K
[16384, 6144]
blk.6.attn_q_a.weight
Q5_K
Q5_K
[6144, 2048]
blk.6.attn_q_a_norm.weight
F32
F32
[2048]
blk.6.attn_q_b.weight
Q8_0
Q8_0
[2048, 16384]
blk.6.attn_v_b.weight
Q8_0
Q8_0
[512, 256, 64]
blk.6.exp_probs_b.bias
F32
F32
[256]
blk.6.ffn_down_exps.weight
Q4_K
Q4_K
[2048, 6144, 256]
blk.6.ffn_down_shexp.weight
Q6_K
Q6_K
[2048, 6144]
blk.6.ffn_gate_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.6.ffn_gate_inp.weight
F32
F32
[6144, 256]
blk.6.ffn_gate_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.6.ffn_norm.weight
F32
F32
[6144]
blk.6.ffn_up_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.6.ffn_up_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.6.indexer.attn_k.weight
Q4_K
Q4_K
[6144, 128]
blk.6.indexer.attn_q_b.weight
Q4_K
Q4_K
[2048, 4096]
blk.6.indexer.k_norm.bias
F32
F32
[128]
blk.6.indexer.k_norm.weight
F32
F32
[128]
blk.6.indexer.proj.weight
Q4_K
Q4_K
[6144, 32]
blk.7
blk.7.attn_k_b.weight
Q8_0
Q8_0
[192, 512, 64]
blk.7.attn_kv_a_mqa.weight
Q8_0
Q8_0
[6144, 576]
blk.7.attn_kv_a_norm.weight
F32
F32
[512]
blk.7.attn_norm.weight
F32
F32
[6144]
blk.7.attn_output.weight
Q4_K
Q4_K
[16384, 6144]
blk.7.attn_q_a.weight
Q4_K
Q4_K
[6144, 2048]
blk.7.attn_q_a_norm.weight
F32
F32
[2048]
blk.7.attn_q_b.weight
Q8_0
Q8_0
[2048, 16384]
blk.7.attn_v_b.weight
Q8_0
Q8_0
[512, 256, 64]
blk.7.exp_probs_b.bias
F32
F32
[256]
blk.7.ffn_down_exps.weight
Q4_K
Q4_K
[2048, 6144, 256]
blk.7.ffn_down_shexp.weight
Q6_K
Q6_K
[2048, 6144]
blk.7.ffn_gate_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.7.ffn_gate_inp.weight
F32
F32
[6144, 256]
blk.7.ffn_gate_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.7.ffn_norm.weight
F32
F32
[6144]
blk.7.ffn_up_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.7.ffn_up_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.7.indexer.attn_k.weight
Q4_K
Q4_K
[6144, 128]
blk.7.indexer.attn_q_b.weight
Q4_K
Q4_K
[2048, 4096]
blk.7.indexer.k_norm.bias
F32
F32
[128]
blk.7.indexer.k_norm.weight
F32
F32
[128]
blk.7.indexer.proj.weight
Q4_K
Q4_K
[6144, 32]
blk.8
blk.8.attn_k_b.weight
Q8_0
Q8_0
[192, 512, 64]
blk.8.attn_kv_a_mqa.weight
Q8_0
Q8_0
[6144, 576]
blk.8.attn_kv_a_norm.weight
F32
F32
[512]
blk.8.attn_norm.weight
F32
F32
[6144]
blk.8.attn_output.weight
Q5_K
Q5_K
[16384, 6144]
blk.8.attn_q_a.weight
Q5_K
Q5_K
[6144, 2048]
blk.8.attn_q_a_norm.weight
F32
F32
[2048]
blk.8.attn_q_b.weight
Q6_K
Q6_K
[2048, 16384]
blk.8.attn_v_b.weight
Q8_0
Q8_0
[512, 256, 64]
blk.8.exp_probs_b.bias
F32
F32
[256]
blk.8.ffn_down_exps.weight
Q5_K
Q5_K
[2048, 6144, 256]
blk.8.ffn_down_shexp.weight
Q8_0
Q8_0
[2048, 6144]
blk.8.ffn_gate_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.8.ffn_gate_inp.weight
F32
F32
[6144, 256]
blk.8.ffn_gate_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.8.ffn_norm.weight
F32
F32
[6144]
blk.8.ffn_up_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.8.ffn_up_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.8.indexer.attn_k.weight
Q4_K
Q4_K
[6144, 128]
blk.8.indexer.attn_q_b.weight
Q4_K
Q4_K
[2048, 4096]
blk.8.indexer.k_norm.bias
F32
F32
[128]
blk.8.indexer.k_norm.weight
F32
F32
[128]
blk.8.indexer.proj.weight
Q4_K
Q4_K
[6144, 32]
blk.9
blk.9.attn_k_b.weight
Q8_0
Q8_0
[192, 512, 64]
blk.9.attn_kv_a_mqa.weight
Q8_0
Q8_0
[6144, 576]
blk.9.attn_kv_a_norm.weight
F32
F32
[512]
blk.9.attn_norm.weight
F32
F32
[6144]
blk.9.attn_output.weight
Q5_K
Q5_K
[16384, 6144]
blk.9.attn_q_a.weight
Q5_K
Q5_K
[6144, 2048]
blk.9.attn_q_a_norm.weight
F32
F32
[2048]
blk.9.attn_q_b.weight
Q6_K
Q6_K
[2048, 16384]
blk.9.attn_v_b.weight
Q8_0
Q8_0
[512, 256, 64]
blk.9.exp_probs_b.bias
F32
F32
[256]
blk.9.ffn_down_exps.weight
Q4_K
Q4_K
[2048, 6144, 256]
blk.9.ffn_down_shexp.weight
Q6_K
Q6_K
[2048, 6144]
blk.9.ffn_gate_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.9.ffn_gate_inp.weight
F32
F32
[6144, 256]
blk.9.ffn_gate_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.9.ffn_norm.weight
F32
F32
[6144]
blk.9.ffn_up_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.9.ffn_up_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.9.indexer.attn_k.weight
Q4_K
Q4_K
[6144, 128]
blk.9.indexer.attn_q_b.weight
Q4_K
Q4_K
[2048, 4096]
blk.9.indexer.k_norm.bias
F32
F32
[128]
blk.9.indexer.k_norm.weight
F32
F32
[128]
blk.9.indexer.proj.weight
Q4_K
Q4_K
[6144, 32]
blk.10
blk.10.attn_k_b.weight
Q8_0
Q8_0
[192, 512, 64]
blk.10.attn_kv_a_mqa.weight
Q8_0
Q8_0
[6144, 576]
blk.10.attn_kv_a_norm.weight
F32
F32
[512]
blk.10.attn_norm.weight
F32
F32
[6144]
blk.10.attn_output.weight
Q4_K
Q4_K
[16384, 6144]
blk.10.attn_q_a.weight
Q4_K
Q4_K
[6144, 2048]
blk.10.attn_q_a_norm.weight
F32
F32
[2048]
blk.10.attn_q_b.weight
Q4_K
Q4_K
[2048, 16384]
blk.10.attn_v_b.weight
Q8_0
Q8_0
[512, 256, 64]
blk.10.exp_probs_b.bias
F32
F32
[256]
blk.10.ffn_down_exps.weight
Q4_K
Q4_K
[2048, 6144, 256]
blk.10.ffn_down_shexp.weight
Q6_K
Q6_K
[2048, 6144]
blk.10.ffn_gate_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.10.ffn_gate_inp.weight
F32
F32
[6144, 256]
blk.10.ffn_gate_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.10.ffn_norm.weight
F32
F32
[6144]
blk.10.ffn_up_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.10.ffn_up_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.10.indexer.attn_k.weight
Q4_K
Q4_K
[6144, 128]
blk.10.indexer.attn_q_b.weight
Q4_K
Q4_K
[2048, 4096]
blk.10.indexer.k_norm.bias
F32
F32
[128]
blk.10.indexer.k_norm.weight
F32
F32
[128]
blk.10.indexer.proj.weight
Q4_K
Q4_K
[6144, 32]
blk.11
blk.11.attn_k_b.weight
Q8_0
Q8_0
[192, 512, 64]
blk.11.attn_kv_a_mqa.weight
Q8_0
Q8_0
[6144, 576]
blk.11.attn_kv_a_norm.weight
F32
F32
[512]
blk.11.attn_norm.weight
F32
F32
[6144]
blk.11.attn_output.weight
Q4_K
Q4_K
[16384, 6144]
blk.11.attn_q_a.weight
Q4_K
Q4_K
[6144, 2048]
blk.11.attn_q_a_norm.weight
F32
F32
[2048]
blk.11.attn_q_b.weight
Q6_K
Q6_K
[2048, 16384]
blk.11.attn_v_b.weight
Q8_0
Q8_0
[512, 256, 64]
blk.11.exp_probs_b.bias
F32
F32
[256]
blk.11.ffn_down_exps.weight
Q4_K
Q4_K
[2048, 6144, 256]
blk.11.ffn_down_shexp.weight
Q6_K
Q6_K
[2048, 6144]
blk.11.ffn_gate_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.11.ffn_gate_inp.weight
F32
F32
[6144, 256]
blk.11.ffn_gate_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.11.ffn_norm.weight
F32
F32
[6144]
blk.11.ffn_up_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.11.ffn_up_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.11.indexer.attn_k.weight
Q4_K
Q4_K
[6144, 128]
blk.11.indexer.attn_q_b.weight
Q4_K
Q4_K
[2048, 4096]
blk.11.indexer.k_norm.bias
F32
F32
[128]
blk.11.indexer.k_norm.weight
F32
F32
[128]
blk.11.indexer.proj.weight
Q4_K
Q4_K
[6144, 32]
blk.12
blk.12.attn_k_b.weight
Q8_0
Q8_0
[192, 512, 64]
blk.12.attn_kv_a_mqa.weight
Q8_0
Q8_0
[6144, 576]
blk.12.attn_kv_a_norm.weight
F32
F32
[512]
blk.12.attn_norm.weight
F32
F32
[6144]
blk.12.attn_output.weight
Q4_K
Q4_K
[16384, 6144]
blk.12.attn_q_a.weight
Q4_K
Q4_K
[6144, 2048]
blk.12.attn_q_a_norm.weight
F32
F32
[2048]
blk.12.attn_q_b.weight
Q4_K
Q4_K
[2048, 16384]
blk.12.attn_v_b.weight
Q8_0
Q8_0
[512, 256, 64]
blk.12.exp_probs_b.bias
F32
F32
[256]
blk.12.ffn_down_exps.weight
Q5_K
Q5_K
[2048, 6144, 256]
blk.12.ffn_down_shexp.weight
Q8_0
Q8_0
[2048, 6144]
blk.12.ffn_gate_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.12.ffn_gate_inp.weight
F32
F32
[6144, 256]
blk.12.ffn_gate_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.12.ffn_norm.weight
F32
F32
[6144]
blk.12.ffn_up_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.12.ffn_up_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.12.indexer.attn_k.weight
Q4_K
Q4_K
[6144, 128]
blk.12.indexer.attn_q_b.weight
Q4_K
Q4_K
[2048, 4096]
blk.12.indexer.k_norm.bias
F32
F32
[128]
blk.12.indexer.k_norm.weight
F32
F32
[128]
blk.12.indexer.proj.weight
Q4_K
Q4_K
[6144, 32]
blk.13
blk.13.attn_k_b.weight
Q8_0
Q8_0
[192, 512, 64]
blk.13.attn_kv_a_mqa.weight
Q8_0
Q8_0
[6144, 576]
blk.13.attn_kv_a_norm.weight
F32
F32
[512]
blk.13.attn_norm.weight
F32
F32
[6144]
blk.13.attn_output.weight
Q4_K
Q4_K
[16384, 6144]
blk.13.attn_q_a.weight
Q4_K
Q4_K
[6144, 2048]
blk.13.attn_q_a_norm.weight
F32
F32
[2048]
blk.13.attn_q_b.weight
Q6_K
Q6_K
[2048, 16384]
blk.13.attn_v_b.weight
Q8_0
Q8_0
[512, 256, 64]
blk.13.exp_probs_b.bias
F32
F32
[256]
blk.13.ffn_down_exps.weight
Q4_K
Q4_K
[2048, 6144, 256]
blk.13.ffn_down_shexp.weight
Q6_K
Q6_K
[2048, 6144]
blk.13.ffn_gate_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.13.ffn_gate_inp.weight
F32
F32
[6144, 256]
blk.13.ffn_gate_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.13.ffn_norm.weight
F32
F32
[6144]
blk.13.ffn_up_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.13.ffn_up_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.13.indexer.attn_k.weight
Q4_K
Q4_K
[6144, 128]
blk.13.indexer.attn_q_b.weight
Q4_K
Q4_K
[2048, 4096]
blk.13.indexer.k_norm.bias
F32
F32
[128]
blk.13.indexer.k_norm.weight
F32
F32
[128]
blk.13.indexer.proj.weight
Q4_K
Q4_K
[6144, 32]
blk.14
blk.14.attn_k_b.weight
Q8_0
Q8_0
[192, 512, 64]
blk.14.attn_kv_a_mqa.weight
Q8_0
Q8_0
[6144, 576]
blk.14.attn_kv_a_norm.weight
F32
F32
[512]
blk.14.attn_norm.weight
F32
F32
[6144]
blk.14.attn_output.weight
Q4_K
Q4_K
[16384, 6144]
blk.14.attn_q_a.weight
Q4_K
Q4_K
[6144, 2048]
blk.14.attn_q_a_norm.weight
F32
F32
[2048]
blk.14.attn_q_b.weight
Q4_K
Q4_K
[2048, 16384]
blk.14.attn_v_b.weight
Q8_0
Q8_0
[512, 256, 64]
blk.14.exp_probs_b.bias
F32
F32
[256]
blk.14.ffn_down_exps.weight
Q4_K
Q4_K
[2048, 6144, 256]
blk.14.ffn_down_shexp.weight
Q6_K
Q6_K
[2048, 6144]
blk.14.ffn_gate_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.14.ffn_gate_inp.weight
F32
F32
[6144, 256]
blk.14.ffn_gate_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.14.ffn_norm.weight
F32
F32
[6144]
blk.14.ffn_up_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.14.ffn_up_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.14.indexer.attn_k.weight
Q4_K
Q4_K
[6144, 128]
blk.14.indexer.attn_q_b.weight
Q4_K
Q4_K
[2048, 4096]
blk.14.indexer.k_norm.bias
F32
F32
[128]
blk.14.indexer.k_norm.weight
F32
F32
[128]
blk.14.indexer.proj.weight
Q4_K
Q4_K
[6144, 32]
blk.15
blk.15.attn_k_b.weight
Q8_0
Q8_0
[192, 512, 64]
blk.15.attn_kv_a_mqa.weight
Q8_0
Q8_0
[6144, 576]
blk.15.attn_kv_a_norm.weight
F32
F32
[512]
blk.15.attn_norm.weight
F32
F32
[6144]
blk.15.attn_output.weight
Q4_K
Q4_K
[16384, 6144]
blk.15.attn_q_a.weight
Q4_K
Q4_K
[6144, 2048]
blk.15.attn_q_a_norm.weight
F32
F32
[2048]
blk.15.attn_q_b.weight
Q4_K
Q4_K
[2048, 16384]
blk.15.attn_v_b.weight
Q8_0
Q8_0
[512, 256, 64]
blk.15.exp_probs_b.bias
F32
F32
[256]
blk.15.ffn_down_exps.weight
Q4_K
Q4_K
[2048, 6144, 256]
blk.15.ffn_down_shexp.weight
Q6_K
Q6_K
[2048, 6144]
blk.15.ffn_gate_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.15.ffn_gate_inp.weight
F32
F32
[6144, 256]
blk.15.ffn_gate_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.15.ffn_norm.weight
F32
F32
[6144]
blk.15.ffn_up_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.15.ffn_up_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.15.indexer.attn_k.weight
Q4_K
Q4_K
[6144, 128]
blk.15.indexer.attn_q_b.weight
Q4_K
Q4_K
[2048, 4096]
blk.15.indexer.k_norm.bias
F32
F32
[128]
blk.15.indexer.k_norm.weight
F32
F32
[128]
blk.15.indexer.proj.weight
Q4_K
Q4_K
[6144, 32]
blk.16
blk.16.attn_k_b.weight
Q8_0
Q8_0
[192, 512, 64]
blk.16.attn_kv_a_mqa.weight
Q8_0
Q8_0
[6144, 576]
blk.16.attn_kv_a_norm.weight
F32
F32
[512]
blk.16.attn_norm.weight
F32
F32
[6144]
blk.16.attn_output.weight
Q4_K
Q4_K
[16384, 6144]
blk.16.attn_q_a.weight
Q4_K
Q4_K
[6144, 2048]
blk.16.attn_q_a_norm.weight
F32
F32
[2048]
blk.16.attn_q_b.weight
Q6_K
Q6_K
[2048, 16384]
blk.16.attn_v_b.weight
Q8_0
Q8_0
[512, 256, 64]
blk.16.exp_probs_b.bias
F32
F32
[256]
blk.16.ffn_down_exps.weight
Q4_K
Q4_K
[2048, 6144, 256]
blk.16.ffn_down_shexp.weight
Q6_K
Q6_K
[2048, 6144]
blk.16.ffn_gate_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.16.ffn_gate_inp.weight
F32
F32
[6144, 256]
blk.16.ffn_gate_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.16.ffn_norm.weight
F32
F32
[6144]
blk.16.ffn_up_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.16.ffn_up_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.16.indexer.attn_k.weight
Q4_K
Q4_K
[6144, 128]
blk.16.indexer.attn_q_b.weight
Q4_K
Q4_K
[2048, 4096]
blk.16.indexer.k_norm.bias
F32
F32
[128]
blk.16.indexer.k_norm.weight
F32
F32
[128]
blk.16.indexer.proj.weight
Q4_K
Q4_K
[6144, 32]
blk.17
blk.17.attn_k_b.weight
Q8_0
Q8_0
[192, 512, 64]
blk.17.attn_kv_a_mqa.weight
Q8_0
Q8_0
[6144, 576]
blk.17.attn_kv_a_norm.weight
F32
F32
[512]
blk.17.attn_norm.weight
F32
F32
[6144]
blk.17.attn_output.weight
Q4_K
Q4_K
[16384, 6144]
blk.17.attn_q_a.weight
Q4_K
Q4_K
[6144, 2048]
blk.17.attn_q_a_norm.weight
F32
F32
[2048]
blk.17.attn_q_b.weight
Q4_K
Q4_K
[2048, 16384]
blk.17.attn_v_b.weight
Q8_0
Q8_0
[512, 256, 64]
blk.17.exp_probs_b.bias
F32
F32
[256]
blk.17.ffn_down_exps.weight
Q4_K
Q4_K
[2048, 6144, 256]
blk.17.ffn_down_shexp.weight
Q6_K
Q6_K
[2048, 6144]
blk.17.ffn_gate_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.17.ffn_gate_inp.weight
F32
F32
[6144, 256]
blk.17.ffn_gate_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.17.ffn_norm.weight
F32
F32
[6144]
blk.17.ffn_up_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.17.ffn_up_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.17.indexer.attn_k.weight
Q4_K
Q4_K
[6144, 128]
blk.17.indexer.attn_q_b.weight
Q4_K
Q4_K
[2048, 4096]
blk.17.indexer.k_norm.bias
F32
F32
[128]
blk.17.indexer.k_norm.weight
F32
F32
[128]
blk.17.indexer.proj.weight
Q4_K
Q4_K
[6144, 32]
blk.18
blk.18.attn_k_b.weight
Q8_0
Q8_0
[192, 512, 64]
blk.18.attn_kv_a_mqa.weight
Q8_0
Q8_0
[6144, 576]
blk.18.attn_kv_a_norm.weight
F32
F32
[512]
blk.18.attn_norm.weight
F32
F32
[6144]
blk.18.attn_output.weight
Q4_K
Q4_K
[16384, 6144]
blk.18.attn_q_a.weight
Q4_K
Q4_K
[6144, 2048]
blk.18.attn_q_a_norm.weight
F32
F32
[2048]
blk.18.attn_q_b.weight
Q4_K
Q4_K
[2048, 16384]
blk.18.attn_v_b.weight
Q8_0
Q8_0
[512, 256, 64]
blk.18.exp_probs_b.bias
F32
F32
[256]
blk.18.ffn_down_exps.weight
Q4_K
Q4_K
[2048, 6144, 256]
blk.18.ffn_down_shexp.weight
Q6_K
Q6_K
[2048, 6144]
blk.18.ffn_gate_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.18.ffn_gate_inp.weight
F32
F32
[6144, 256]
blk.18.ffn_gate_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.18.ffn_norm.weight
F32
F32
[6144]
blk.18.ffn_up_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.18.ffn_up_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.18.indexer.attn_k.weight
Q4_K
Q4_K
[6144, 128]
blk.18.indexer.attn_q_b.weight
Q4_K
Q4_K
[2048, 4096]
blk.18.indexer.k_norm.bias
F32
F32
[128]
blk.18.indexer.k_norm.weight
F32
F32
[128]
blk.18.indexer.proj.weight
Q4_K
Q4_K
[6144, 32]
blk.19
blk.19.attn_k_b.weight
Q8_0
Q8_0
[192, 512, 64]
blk.19.attn_kv_a_mqa.weight
Q8_0
Q8_0
[6144, 576]
blk.19.attn_kv_a_norm.weight
F32
F32
[512]
blk.19.attn_norm.weight
F32
F32
[6144]
blk.19.attn_output.weight
Q4_K
Q4_K
[16384, 6144]
blk.19.attn_q_a.weight
Q4_K
Q4_K
[6144, 2048]
blk.19.attn_q_a_norm.weight
F32
F32
[2048]
blk.19.attn_q_b.weight
Q4_K
Q4_K
[2048, 16384]
blk.19.attn_v_b.weight
Q8_0
Q8_0
[512, 256, 64]
blk.19.exp_probs_b.bias
F32
F32
[256]
blk.19.ffn_down_exps.weight
Q4_K
Q4_K
[2048, 6144, 256]
blk.19.ffn_down_shexp.weight
Q6_K
Q6_K
[2048, 6144]
blk.19.ffn_gate_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.19.ffn_gate_inp.weight
F32
F32
[6144, 256]
blk.19.ffn_gate_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.19.ffn_norm.weight
F32
F32
[6144]
blk.19.ffn_up_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.19.ffn_up_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.19.indexer.attn_k.weight
Q4_K
Q4_K
[6144, 128]
blk.19.indexer.attn_q_b.weight
Q4_K
Q4_K
[2048, 4096]
blk.19.indexer.k_norm.bias
F32
F32
[128]
blk.19.indexer.k_norm.weight
F32
F32
[128]
blk.19.indexer.proj.weight
Q4_K
Q4_K
[6144, 32]
blk.20
blk.20.attn_k_b.weight
Q8_0
Q8_0
[192, 512, 64]
blk.20.attn_kv_a_mqa.weight
Q8_0
Q8_0
[6144, 576]
blk.20.attn_kv_a_norm.weight
F32
F32
[512]
blk.20.attn_norm.weight
F32
F32
[6144]
blk.20.attn_output.weight
Q4_K
Q4_K
[16384, 6144]
blk.20.attn_q_a.weight
Q4_K
Q4_K
[6144, 2048]
blk.20.attn_q_a_norm.weight
F32
F32
[2048]
blk.20.attn_q_b.weight
Q4_K
Q4_K
[2048, 16384]
blk.20.attn_v_b.weight
Q8_0
Q8_0
[512, 256, 64]
blk.20.exp_probs_b.bias
F32
F32
[256]
blk.20.ffn_down_exps.weight
Q4_K
Q4_K
[2048, 6144, 256]
blk.20.ffn_down_shexp.weight
Q6_K
Q6_K
[2048, 6144]
blk.20.ffn_gate_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.20.ffn_gate_inp.weight
F32
F32
[6144, 256]
blk.20.ffn_gate_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.20.ffn_norm.weight
F32
F32
[6144]
blk.20.ffn_up_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.20.ffn_up_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.20.indexer.attn_k.weight
Q4_K
Q4_K
[6144, 128]
blk.20.indexer.attn_q_b.weight
Q4_K
Q4_K
[2048, 4096]
blk.20.indexer.k_norm.bias
F32
F32
[128]
blk.20.indexer.k_norm.weight
F32
F32
[128]
blk.20.indexer.proj.weight
Q4_K
Q4_K
[6144, 32]
blk.21
blk.21.attn_k_b.weight
Q8_0
Q8_0
[192, 512, 64]
blk.21.attn_kv_a_mqa.weight
Q8_0
Q8_0
[6144, 576]
blk.21.attn_kv_a_norm.weight
F32
F32
[512]
blk.21.attn_norm.weight
F32
F32
[6144]
blk.21.attn_output.weight
Q4_K
Q4_K
[16384, 6144]
blk.21.attn_q_a.weight
Q4_K
Q4_K
[6144, 2048]
blk.21.attn_q_a_norm.weight
F32
F32
[2048]
blk.21.attn_q_b.weight
Q6_K
Q6_K
[2048, 16384]
blk.21.attn_v_b.weight
Q8_0
Q8_0
[512, 256, 64]
blk.21.exp_probs_b.bias
F32
F32
[256]
blk.21.ffn_down_exps.weight
Q4_K
Q4_K
[2048, 6144, 256]
blk.21.ffn_down_shexp.weight
Q6_K
Q6_K
[2048, 6144]
blk.21.ffn_gate_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.21.ffn_gate_inp.weight
F32
F32
[6144, 256]
blk.21.ffn_gate_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.21.ffn_norm.weight
F32
F32
[6144]
blk.21.ffn_up_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.21.ffn_up_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.21.indexer.attn_k.weight
Q4_K
Q4_K
[6144, 128]
blk.21.indexer.attn_q_b.weight
Q4_K
Q4_K
[2048, 4096]
blk.21.indexer.k_norm.bias
F32
F32
[128]
blk.21.indexer.k_norm.weight
F32
F32
[128]
blk.21.indexer.proj.weight
Q4_K
Q4_K
[6144, 32]
blk.22
blk.22.attn_k_b.weight
Q8_0
Q8_0
[192, 512, 64]
blk.22.attn_kv_a_mqa.weight
Q8_0
Q8_0
[6144, 576]
blk.22.attn_kv_a_norm.weight
F32
F32
[512]
blk.22.attn_norm.weight
F32
F32
[6144]
blk.22.attn_output.weight
Q4_K
Q4_K
[16384, 6144]
blk.22.attn_q_a.weight
Q4_K
Q4_K
[6144, 2048]
blk.22.attn_q_a_norm.weight
F32
F32
[2048]
blk.22.attn_q_b.weight
Q4_K
Q4_K
[2048, 16384]
blk.22.attn_v_b.weight
Q8_0
Q8_0
[512, 256, 64]
blk.22.exp_probs_b.bias
F32
F32
[256]
blk.22.ffn_down_exps.weight
Q5_K
Q5_K
[2048, 6144, 256]
blk.22.ffn_down_shexp.weight
Q8_0
Q8_0
[2048, 6144]
blk.22.ffn_gate_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.22.ffn_gate_inp.weight
F32
F32
[6144, 256]
blk.22.ffn_gate_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.22.ffn_norm.weight
F32
F32
[6144]
blk.22.ffn_up_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.22.ffn_up_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.22.indexer.attn_k.weight
Q4_K
Q4_K
[6144, 128]
blk.22.indexer.attn_q_b.weight
Q4_K
Q4_K
[2048, 4096]
blk.22.indexer.k_norm.bias
F32
F32
[128]
blk.22.indexer.k_norm.weight
F32
F32
[128]
blk.22.indexer.proj.weight
Q4_K
Q4_K
[6144, 32]
blk.23
blk.23.attn_k_b.weight
Q8_0
Q8_0
[192, 512, 64]
blk.23.attn_kv_a_mqa.weight
Q8_0
Q8_0
[6144, 576]
blk.23.attn_kv_a_norm.weight
F32
F32
[512]
blk.23.attn_norm.weight
F32
F32
[6144]
blk.23.attn_output.weight
Q4_K
Q4_K
[16384, 6144]
blk.23.attn_q_a.weight
Q4_K
Q4_K
[6144, 2048]
blk.23.attn_q_a_norm.weight
F32
F32
[2048]
blk.23.attn_q_b.weight
Q4_K
Q4_K
[2048, 16384]
blk.23.attn_v_b.weight
Q8_0
Q8_0
[512, 256, 64]
blk.23.exp_probs_b.bias
F32
F32
[256]
blk.23.ffn_down_exps.weight
Q4_K
Q4_K
[2048, 6144, 256]
blk.23.ffn_down_shexp.weight
Q6_K
Q6_K
[2048, 6144]
blk.23.ffn_gate_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.23.ffn_gate_inp.weight
F32
F32
[6144, 256]
blk.23.ffn_gate_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.23.ffn_norm.weight
F32
F32
[6144]
blk.23.ffn_up_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.23.ffn_up_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.23.indexer.attn_k.weight
Q4_K
Q4_K
[6144, 128]
blk.23.indexer.attn_q_b.weight
Q4_K
Q4_K
[2048, 4096]
blk.23.indexer.k_norm.bias
F32
F32
[128]
blk.23.indexer.k_norm.weight
F32
F32
[128]
blk.23.indexer.proj.weight
Q4_K
Q4_K
[6144, 32]
blk.24
blk.24.attn_k_b.weight
Q8_0
Q8_0
[192, 512, 64]
blk.24.attn_kv_a_mqa.weight
Q8_0
Q8_0
[6144, 576]
blk.24.attn_kv_a_norm.weight
F32
F32
[512]
blk.24.attn_norm.weight
F32
F32
[6144]
blk.24.attn_output.weight
Q4_K
Q4_K
[16384, 6144]
blk.24.attn_q_a.weight
Q4_K
Q4_K
[6144, 2048]
blk.24.attn_q_a_norm.weight
F32
F32
[2048]
blk.24.attn_q_b.weight
Q4_K
Q4_K
[2048, 16384]
blk.24.attn_v_b.weight
Q8_0
Q8_0
[512, 256, 64]
blk.24.exp_probs_b.bias
F32
F32
[256]
blk.24.ffn_down_exps.weight
Q4_K
Q4_K
[2048, 6144, 256]
blk.24.ffn_down_shexp.weight
Q6_K
Q6_K
[2048, 6144]
blk.24.ffn_gate_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.24.ffn_gate_inp.weight
F32
F32
[6144, 256]
blk.24.ffn_gate_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.24.ffn_norm.weight
F32
F32
[6144]
blk.24.ffn_up_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.24.ffn_up_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.24.indexer.attn_k.weight
Q4_K
Q4_K
[6144, 128]
blk.24.indexer.attn_q_b.weight
Q4_K
Q4_K
[2048, 4096]
blk.24.indexer.k_norm.bias
F32
F32
[128]
blk.24.indexer.k_norm.weight
F32
F32
[128]
blk.24.indexer.proj.weight
Q4_K
Q4_K
[6144, 32]
blk.25
blk.25.attn_k_b.weight
Q8_0
Q8_0
[192, 512, 64]
blk.25.attn_kv_a_mqa.weight
Q8_0
Q8_0
[6144, 576]
blk.25.attn_kv_a_norm.weight
F32
F32
[512]
blk.25.attn_norm.weight
F32
F32
[6144]
blk.25.attn_output.weight
Q4_K
Q4_K
[16384, 6144]
blk.25.attn_q_a.weight
Q4_K
Q4_K
[6144, 2048]
blk.25.attn_q_a_norm.weight
F32
F32
[2048]
blk.25.attn_q_b.weight
Q4_K
Q4_K
[2048, 16384]
blk.25.attn_v_b.weight
Q8_0
Q8_0
[512, 256, 64]
blk.25.exp_probs_b.bias
F32
F32
[256]
blk.25.ffn_down_exps.weight
Q4_K
Q4_K
[2048, 6144, 256]
blk.25.ffn_down_shexp.weight
Q6_K
Q6_K
[2048, 6144]
blk.25.ffn_gate_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.25.ffn_gate_inp.weight
F32
F32
[6144, 256]
blk.25.ffn_gate_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.25.ffn_norm.weight
F32
F32
[6144]
blk.25.ffn_up_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.25.ffn_up_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.25.indexer.attn_k.weight
Q4_K
Q4_K
[6144, 128]
blk.25.indexer.attn_q_b.weight
Q4_K
Q4_K
[2048, 4096]
blk.25.indexer.k_norm.bias
F32
F32
[128]
blk.25.indexer.k_norm.weight
F32
F32
[128]
blk.25.indexer.proj.weight
Q4_K
Q4_K
[6144, 32]
blk.26
blk.26.attn_k_b.weight
Q8_0
Q8_0
[192, 512, 64]
blk.26.attn_kv_a_mqa.weight
Q8_0
Q8_0
[6144, 576]
blk.26.attn_kv_a_norm.weight
F32
F32
[512]
blk.26.attn_norm.weight
F32
F32
[6144]
blk.26.attn_output.weight
Q4_K
Q4_K
[16384, 6144]
blk.26.attn_q_a.weight
Q4_K
Q4_K
[6144, 2048]
blk.26.attn_q_a_norm.weight
F32
F32
[2048]
blk.26.attn_q_b.weight
Q4_K
Q4_K
[2048, 16384]
blk.26.attn_v_b.weight
Q8_0
Q8_0
[512, 256, 64]
blk.26.exp_probs_b.bias
F32
F32
[256]
blk.26.ffn_down_exps.weight
Q4_K
Q4_K
[2048, 6144, 256]
blk.26.ffn_down_shexp.weight
Q6_K
Q6_K
[2048, 6144]
blk.26.ffn_gate_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.26.ffn_gate_inp.weight
F32
F32
[6144, 256]
blk.26.ffn_gate_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.26.ffn_norm.weight
F32
F32
[6144]
blk.26.ffn_up_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.26.ffn_up_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.26.indexer.attn_k.weight
Q4_K
Q4_K
[6144, 128]
blk.26.indexer.attn_q_b.weight
Q4_K
Q4_K
[2048, 4096]
blk.26.indexer.k_norm.bias
F32
F32
[128]
blk.26.indexer.k_norm.weight
F32
F32
[128]
blk.26.indexer.proj.weight
Q4_K
Q4_K
[6144, 32]
blk.27
blk.27.attn_k_b.weight
Q8_0
Q8_0
[192, 512, 64]
blk.27.attn_kv_a_mqa.weight
Q8_0
Q8_0
[6144, 576]
blk.27.attn_kv_a_norm.weight
F32
F32
[512]
blk.27.attn_norm.weight
F32
F32
[6144]
blk.27.attn_output.weight
Q4_K
Q4_K
[16384, 6144]
blk.27.attn_q_a.weight
Q4_K
Q4_K
[6144, 2048]
blk.27.attn_q_a_norm.weight
F32
F32
[2048]
blk.27.attn_q_b.weight
Q4_K
Q4_K
[2048, 16384]
blk.27.attn_v_b.weight
Q8_0
Q8_0
[512, 256, 64]
blk.27.exp_probs_b.bias
F32
F32
[256]
blk.27.ffn_down_exps.weight
Q4_K
Q4_K
[2048, 6144, 256]
blk.27.ffn_down_shexp.weight
Q6_K
Q6_K
[2048, 6144]
blk.27.ffn_gate_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.27.ffn_gate_inp.weight
F32
F32
[6144, 256]
blk.27.ffn_gate_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.27.ffn_norm.weight
F32
F32
[6144]
blk.27.ffn_up_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.27.ffn_up_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.27.indexer.attn_k.weight
Q4_K
Q4_K
[6144, 128]
blk.27.indexer.attn_q_b.weight
Q4_K
Q4_K
[2048, 4096]
blk.27.indexer.k_norm.bias
F32
F32
[128]
blk.27.indexer.k_norm.weight
F32
F32
[128]
blk.27.indexer.proj.weight
Q4_K
Q4_K
[6144, 32]
blk.28
blk.28.attn_k_b.weight
Q8_0
Q8_0
[192, 512, 64]
blk.28.attn_kv_a_mqa.weight
Q8_0
Q8_0
[6144, 576]
blk.28.attn_kv_a_norm.weight
F32
F32
[512]
blk.28.attn_norm.weight
F32
F32
[6144]
blk.28.attn_output.weight
Q4_K
Q4_K
[16384, 6144]
blk.28.attn_q_a.weight
Q4_K
Q4_K
[6144, 2048]
blk.28.attn_q_a_norm.weight
F32
F32
[2048]
blk.28.attn_q_b.weight
Q4_K
Q4_K
[2048, 16384]
blk.28.attn_v_b.weight
Q8_0
Q8_0
[512, 256, 64]
blk.28.exp_probs_b.bias
F32
F32
[256]
blk.28.ffn_down_exps.weight
Q4_K
Q4_K
[2048, 6144, 256]
blk.28.ffn_down_shexp.weight
Q6_K
Q6_K
[2048, 6144]
blk.28.ffn_gate_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.28.ffn_gate_inp.weight
F32
F32
[6144, 256]
blk.28.ffn_gate_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.28.ffn_norm.weight
F32
F32
[6144]
blk.28.ffn_up_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.28.ffn_up_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.28.indexer.attn_k.weight
Q4_K
Q4_K
[6144, 128]
blk.28.indexer.attn_q_b.weight
Q4_K
Q4_K
[2048, 4096]
blk.28.indexer.k_norm.bias
F32
F32
[128]
blk.28.indexer.k_norm.weight
F32
F32
[128]
blk.28.indexer.proj.weight
Q4_K
Q4_K
[6144, 32]
blk.29
blk.29.attn_k_b.weight
Q8_0
Q8_0
[192, 512, 64]
blk.29.attn_kv_a_mqa.weight
Q8_0
Q8_0
[6144, 576]
blk.29.attn_kv_a_norm.weight
F32
F32
[512]
blk.29.attn_norm.weight
F32
F32
[6144]
blk.29.attn_output.weight
Q4_K
Q4_K
[16384, 6144]
blk.29.attn_q_a.weight
Q4_K
Q4_K
[6144, 2048]
blk.29.attn_q_a_norm.weight
F32
F32
[2048]
blk.29.attn_q_b.weight
Q4_K
Q4_K
[2048, 16384]
blk.29.attn_v_b.weight
Q8_0
Q8_0
[512, 256, 64]
blk.29.exp_probs_b.bias
F32
F32
[256]
blk.29.ffn_down_exps.weight
Q5_K
Q5_K
[2048, 6144, 256]
blk.29.ffn_down_shexp.weight
Q8_0
Q8_0
[2048, 6144]
blk.29.ffn_gate_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.29.ffn_gate_inp.weight
F32
F32
[6144, 256]
blk.29.ffn_gate_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.29.ffn_norm.weight
F32
F32
[6144]
blk.29.ffn_up_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.29.ffn_up_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.29.indexer.attn_k.weight
Q4_K
Q4_K
[6144, 128]
blk.29.indexer.attn_q_b.weight
Q4_K
Q4_K
[2048, 4096]
blk.29.indexer.k_norm.bias
F32
F32
[128]
blk.29.indexer.k_norm.weight
F32
F32
[128]
blk.29.indexer.proj.weight
Q4_K
Q4_K
[6144, 32]
blk.30
blk.30.attn_k_b.weight
Q8_0
Q8_0
[192, 512, 64]
blk.30.attn_kv_a_mqa.weight
Q8_0
Q8_0
[6144, 576]
blk.30.attn_kv_a_norm.weight
F32
F32
[512]
blk.30.attn_norm.weight
F32
F32
[6144]
blk.30.attn_output.weight
Q4_K
Q4_K
[16384, 6144]
blk.30.attn_q_a.weight
Q4_K
Q4_K
[6144, 2048]
blk.30.attn_q_a_norm.weight
F32
F32
[2048]
blk.30.attn_q_b.weight
Q4_K
Q4_K
[2048, 16384]
blk.30.attn_v_b.weight
Q8_0
Q8_0
[512, 256, 64]
blk.30.exp_probs_b.bias
F32
F32
[256]
blk.30.ffn_down_exps.weight
Q4_K
Q4_K
[2048, 6144, 256]
blk.30.ffn_down_shexp.weight
Q6_K
Q6_K
[2048, 6144]
blk.30.ffn_gate_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.30.ffn_gate_inp.weight
F32
F32
[6144, 256]
blk.30.ffn_gate_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.30.ffn_norm.weight
F32
F32
[6144]
blk.30.ffn_up_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.30.ffn_up_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.30.indexer.attn_k.weight
Q4_K
Q4_K
[6144, 128]
blk.30.indexer.attn_q_b.weight
Q4_K
Q4_K
[2048, 4096]
blk.30.indexer.k_norm.bias
F32
F32
[128]
blk.30.indexer.k_norm.weight
F32
F32
[128]
blk.30.indexer.proj.weight
Q4_K
Q4_K
[6144, 32]
blk.31
blk.31.attn_k_b.weight
Q8_0
Q8_0
[192, 512, 64]
blk.31.attn_kv_a_mqa.weight
Q8_0
Q8_0
[6144, 576]
blk.31.attn_kv_a_norm.weight
F32
F32
[512]
blk.31.attn_norm.weight
F32
F32
[6144]
blk.31.attn_output.weight
Q4_K
Q4_K
[16384, 6144]
blk.31.attn_q_a.weight
Q4_K
Q4_K
[6144, 2048]
blk.31.attn_q_a_norm.weight
F32
F32
[2048]
blk.31.attn_q_b.weight
Q4_K
Q4_K
[2048, 16384]
blk.31.attn_v_b.weight
Q8_0
Q8_0
[512, 256, 64]
blk.31.exp_probs_b.bias
F32
F32
[256]
blk.31.ffn_down_exps.weight
Q4_K
Q4_K
[2048, 6144, 256]
blk.31.ffn_down_shexp.weight
Q6_K
Q6_K
[2048, 6144]
blk.31.ffn_gate_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.31.ffn_gate_inp.weight
F32
F32
[6144, 256]
blk.31.ffn_gate_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.31.ffn_norm.weight
F32
F32
[6144]
blk.31.ffn_up_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.31.ffn_up_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.31.indexer.attn_k.weight
Q4_K
Q4_K
[6144, 128]
blk.31.indexer.attn_q_b.weight
Q4_K
Q4_K
[2048, 4096]
blk.31.indexer.k_norm.bias
F32
F32
[128]
blk.31.indexer.k_norm.weight
F32
F32
[128]
blk.31.indexer.proj.weight
Q4_K
Q4_K
[6144, 32]
blk.32
blk.32.attn_k_b.weight
Q8_0
Q8_0
[192, 512, 64]
blk.32.attn_kv_a_mqa.weight
Q8_0
Q8_0
[6144, 576]
blk.32.attn_kv_a_norm.weight
F32
F32
[512]
blk.32.attn_norm.weight
F32
F32
[6144]
blk.32.attn_output.weight
Q4_K
Q4_K
[16384, 6144]
blk.32.attn_q_a.weight
Q4_K
Q4_K
[6144, 2048]
blk.32.attn_q_a_norm.weight
F32
F32
[2048]
blk.32.attn_q_b.weight
Q4_K
Q4_K
[2048, 16384]
blk.32.attn_v_b.weight
Q8_0
Q8_0
[512, 256, 64]
blk.32.exp_probs_b.bias
F32
F32
[256]
blk.32.ffn_down_exps.weight
Q4_K
Q4_K
[2048, 6144, 256]
blk.32.ffn_down_shexp.weight
Q6_K
Q6_K
[2048, 6144]
blk.32.ffn_gate_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.32.ffn_gate_inp.weight
F32
F32
[6144, 256]
blk.32.ffn_gate_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.32.ffn_norm.weight
F32
F32
[6144]
blk.32.ffn_up_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.32.ffn_up_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.32.indexer.attn_k.weight
Q4_K
Q4_K
[6144, 128]
blk.32.indexer.attn_q_b.weight
Q4_K
Q4_K
[2048, 4096]
blk.32.indexer.k_norm.bias
F32
F32
[128]
blk.32.indexer.k_norm.weight
F32
F32
[128]
blk.32.indexer.proj.weight
Q4_K
Q4_K
[6144, 32]
blk.33
blk.33.attn_k_b.weight
Q8_0
Q8_0
[192, 512, 64]
blk.33.attn_kv_a_mqa.weight
Q8_0
Q8_0
[6144, 576]
blk.33.attn_kv_a_norm.weight
F32
F32
[512]
blk.33.attn_norm.weight
F32
F32
[6144]
blk.33.attn_output.weight
Q4_K
Q4_K
[16384, 6144]
blk.33.attn_q_a.weight
Q4_K
Q4_K
[6144, 2048]
blk.33.attn_q_a_norm.weight
F32
F32
[2048]
blk.33.attn_q_b.weight
Q4_K
Q4_K
[2048, 16384]
blk.33.attn_v_b.weight
Q8_0
Q8_0
[512, 256, 64]
blk.33.exp_probs_b.bias
F32
F32
[256]
blk.33.ffn_down_exps.weight
Q4_K
Q4_K
[2048, 6144, 256]
blk.33.ffn_down_shexp.weight
Q6_K
Q6_K
[2048, 6144]
blk.33.ffn_gate_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.33.ffn_gate_inp.weight
F32
F32
[6144, 256]
blk.33.ffn_gate_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.33.ffn_norm.weight
F32
F32
[6144]
blk.33.ffn_up_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.33.ffn_up_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.33.indexer.attn_k.weight
Q4_K
Q4_K
[6144, 128]
blk.33.indexer.attn_q_b.weight
Q4_K
Q4_K
[2048, 4096]
blk.33.indexer.k_norm.bias
F32
F32
[128]
blk.33.indexer.k_norm.weight
F32
F32
[128]
blk.33.indexer.proj.weight
Q4_K
Q4_K
[6144, 32]
blk.34
blk.34.attn_k_b.weight
Q8_0
Q8_0
[192, 512, 64]
blk.34.attn_kv_a_mqa.weight
Q8_0
Q8_0
[6144, 576]
blk.34.attn_kv_a_norm.weight
F32
F32
[512]
blk.34.attn_norm.weight
F32
F32
[6144]
blk.34.attn_output.weight
Q4_K
Q4_K
[16384, 6144]
blk.34.attn_q_a.weight
Q4_K
Q4_K
[6144, 2048]
blk.34.attn_q_a_norm.weight
F32
F32
[2048]
blk.34.attn_q_b.weight
Q4_K
Q4_K
[2048, 16384]
blk.34.attn_v_b.weight
Q8_0
Q8_0
[512, 256, 64]
blk.34.exp_probs_b.bias
F32
F32
[256]
blk.34.ffn_down_exps.weight
Q4_K
Q4_K
[2048, 6144, 256]
blk.34.ffn_down_shexp.weight
Q6_K
Q6_K
[2048, 6144]
blk.34.ffn_gate_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.34.ffn_gate_inp.weight
F32
F32
[6144, 256]
blk.34.ffn_gate_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.34.ffn_norm.weight
F32
F32
[6144]
blk.34.ffn_up_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.34.ffn_up_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.34.indexer.attn_k.weight
Q4_K
Q4_K
[6144, 128]
blk.34.indexer.attn_q_b.weight
Q4_K
Q4_K
[2048, 4096]
blk.34.indexer.k_norm.bias
F32
F32
[128]
blk.34.indexer.k_norm.weight
F32
F32
[128]
blk.34.indexer.proj.weight
Q4_K
Q4_K
[6144, 32]
blk.35
blk.35.attn_k_b.weight
Q8_0
Q8_0
[192, 512, 64]
blk.35.attn_kv_a_mqa.weight
Q8_0
Q8_0
[6144, 576]
blk.35.attn_kv_a_norm.weight
F32
F32
[512]
blk.35.attn_norm.weight
F32
F32
[6144]
blk.35.attn_output.weight
Q4_K
Q4_K
[16384, 6144]
blk.35.attn_q_a.weight
Q4_K
Q4_K
[6144, 2048]
blk.35.attn_q_a_norm.weight
F32
F32
[2048]
blk.35.attn_q_b.weight
Q4_K
Q4_K
[2048, 16384]
blk.35.attn_v_b.weight
Q8_0
Q8_0
[512, 256, 64]
blk.35.exp_probs_b.bias
F32
F32
[256]
blk.35.ffn_down_exps.weight
Q4_K
Q4_K
[2048, 6144, 256]
blk.35.ffn_down_shexp.weight
Q6_K
Q6_K
[2048, 6144]
blk.35.ffn_gate_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.35.ffn_gate_inp.weight
F32
F32
[6144, 256]
blk.35.ffn_gate_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.35.ffn_norm.weight
F32
F32
[6144]
blk.35.ffn_up_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.35.ffn_up_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.35.indexer.attn_k.weight
Q4_K
Q4_K
[6144, 128]
blk.35.indexer.attn_q_b.weight
Q4_K
Q4_K
[2048, 4096]
blk.35.indexer.k_norm.bias
F32
F32
[128]
blk.35.indexer.k_norm.weight
F32
F32
[128]
blk.35.indexer.proj.weight
Q4_K
Q4_K
[6144, 32]
blk.36
blk.36.attn_k_b.weight
Q8_0
Q8_0
[192, 512, 64]
blk.36.attn_kv_a_mqa.weight
Q8_0
Q8_0
[6144, 576]
blk.36.attn_kv_a_norm.weight
F32
F32
[512]
blk.36.attn_norm.weight
F32
F32
[6144]
blk.36.attn_output.weight
Q4_K
Q4_K
[16384, 6144]
blk.36.attn_q_a.weight
Q4_K
Q4_K
[6144, 2048]
blk.36.attn_q_a_norm.weight
F32
F32
[2048]
blk.36.attn_q_b.weight
Q6_K
Q6_K
[2048, 16384]
blk.36.attn_v_b.weight
Q8_0
Q8_0
[512, 256, 64]
blk.36.exp_probs_b.bias
F32
F32
[256]
blk.36.ffn_down_exps.weight
Q4_K
Q4_K
[2048, 6144, 256]
blk.36.ffn_down_shexp.weight
Q6_K
Q6_K
[2048, 6144]
blk.36.ffn_gate_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.36.ffn_gate_inp.weight
F32
F32
[6144, 256]
blk.36.ffn_gate_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.36.ffn_norm.weight
F32
F32
[6144]
blk.36.ffn_up_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.36.ffn_up_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.36.indexer.attn_k.weight
Q4_K
Q4_K
[6144, 128]
blk.36.indexer.attn_q_b.weight
Q4_K
Q4_K
[2048, 4096]
blk.36.indexer.k_norm.bias
F32
F32
[128]
blk.36.indexer.k_norm.weight
F32
F32
[128]
blk.36.indexer.proj.weight
Q4_K
Q4_K
[6144, 32]
blk.37
blk.37.attn_k_b.weight
Q8_0
Q8_0
[192, 512, 64]
blk.37.attn_kv_a_mqa.weight
Q8_0
Q8_0
[6144, 576]
blk.37.attn_kv_a_norm.weight
F32
F32
[512]
blk.37.attn_norm.weight
F32
F32
[6144]
blk.37.attn_output.weight
Q4_K
Q4_K
[16384, 6144]
blk.37.attn_q_a.weight
Q4_K
Q4_K
[6144, 2048]
blk.37.attn_q_a_norm.weight
F32
F32
[2048]
blk.37.attn_q_b.weight
Q4_K
Q4_K
[2048, 16384]
blk.37.attn_v_b.weight
Q8_0
Q8_0
[512, 256, 64]
blk.37.exp_probs_b.bias
F32
F32
[256]
blk.37.ffn_down_exps.weight
Q4_K
Q4_K
[2048, 6144, 256]
blk.37.ffn_down_shexp.weight
Q6_K
Q6_K
[2048, 6144]
blk.37.ffn_gate_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.37.ffn_gate_inp.weight
F32
F32
[6144, 256]
blk.37.ffn_gate_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.37.ffn_norm.weight
F32
F32
[6144]
blk.37.ffn_up_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.37.ffn_up_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.37.indexer.attn_k.weight
Q4_K
Q4_K
[6144, 128]
blk.37.indexer.attn_q_b.weight
Q4_K
Q4_K
[2048, 4096]
blk.37.indexer.k_norm.bias
F32
F32
[128]
blk.37.indexer.k_norm.weight
F32
F32
[128]
blk.37.indexer.proj.weight
Q4_K
Q4_K
[6144, 32]
blk.38
blk.38.attn_k_b.weight
Q8_0
Q8_0
[192, 512, 64]
blk.38.attn_kv_a_mqa.weight
Q8_0
Q8_0
[6144, 576]
blk.38.attn_kv_a_norm.weight
F32
F32
[512]
blk.38.attn_norm.weight
F32
F32
[6144]
blk.38.attn_output.weight
Q4_K
Q4_K
[16384, 6144]
blk.38.attn_q_a.weight
Q4_K
Q4_K
[6144, 2048]
blk.38.attn_q_a_norm.weight
F32
F32
[2048]
blk.38.attn_q_b.weight
Q4_K
Q4_K
[2048, 16384]
blk.38.attn_v_b.weight
Q8_0
Q8_0
[512, 256, 64]
blk.38.exp_probs_b.bias
F32
F32
[256]
blk.38.ffn_down_exps.weight
Q4_K
Q4_K
[2048, 6144, 256]
blk.38.ffn_down_shexp.weight
Q6_K
Q6_K
[2048, 6144]
blk.38.ffn_gate_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.38.ffn_gate_inp.weight
F32
F32
[6144, 256]
blk.38.ffn_gate_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.38.ffn_norm.weight
F32
F32
[6144]
blk.38.ffn_up_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.38.ffn_up_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.38.indexer.attn_k.weight
Q4_K
Q4_K
[6144, 128]
blk.38.indexer.attn_q_b.weight
Q4_K
Q4_K
[2048, 4096]
blk.38.indexer.k_norm.bias
F32
F32
[128]
blk.38.indexer.k_norm.weight
F32
F32
[128]
blk.38.indexer.proj.weight
Q4_K
Q4_K
[6144, 32]
blk.39
blk.39.attn_k_b.weight
Q8_0
Q8_0
[192, 512, 64]
blk.39.attn_kv_a_mqa.weight
Q8_0
Q8_0
[6144, 576]
blk.39.attn_kv_a_norm.weight
F32
F32
[512]
blk.39.attn_norm.weight
F32
F32
[6144]
blk.39.attn_output.weight
Q4_K
Q4_K
[16384, 6144]
blk.39.attn_q_a.weight
Q4_K
Q4_K
[6144, 2048]
blk.39.attn_q_a_norm.weight
F32
F32
[2048]
blk.39.attn_q_b.weight
Q4_K
Q4_K
[2048, 16384]
blk.39.attn_v_b.weight
Q8_0
Q8_0
[512, 256, 64]
blk.39.exp_probs_b.bias
F32
F32
[256]
blk.39.ffn_down_exps.weight
Q4_K
Q4_K
[2048, 6144, 256]
blk.39.ffn_down_shexp.weight
Q6_K
Q6_K
[2048, 6144]
blk.39.ffn_gate_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.39.ffn_gate_inp.weight
F32
F32
[6144, 256]
blk.39.ffn_gate_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.39.ffn_norm.weight
F32
F32
[6144]
blk.39.ffn_up_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.39.ffn_up_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.39.indexer.attn_k.weight
Q4_K
Q4_K
[6144, 128]
blk.39.indexer.attn_q_b.weight
Q4_K
Q4_K
[2048, 4096]
blk.39.indexer.k_norm.bias
F32
F32
[128]
blk.39.indexer.k_norm.weight
F32
F32
[128]
blk.39.indexer.proj.weight
Q4_K
Q4_K
[6144, 32]
blk.40
blk.40.attn_k_b.weight
Q8_0
Q8_0
[192, 512, 64]
blk.40.attn_kv_a_mqa.weight
Q8_0
Q8_0
[6144, 576]
blk.40.attn_kv_a_norm.weight
F32
F32
[512]
blk.40.attn_norm.weight
F32
F32
[6144]
blk.40.attn_output.weight
Q4_K
Q4_K
[16384, 6144]
blk.40.attn_q_a.weight
Q4_K
Q4_K
[6144, 2048]
blk.40.attn_q_a_norm.weight
F32
F32
[2048]
blk.40.attn_q_b.weight
Q4_K
Q4_K
[2048, 16384]
blk.40.attn_v_b.weight
Q8_0
Q8_0
[512, 256, 64]
blk.40.exp_probs_b.bias
F32
F32
[256]
blk.40.ffn_down_exps.weight
Q4_K
Q4_K
[2048, 6144, 256]
blk.40.ffn_down_shexp.weight
Q6_K
Q6_K
[2048, 6144]
blk.40.ffn_gate_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.40.ffn_gate_inp.weight
F32
F32
[6144, 256]
blk.40.ffn_gate_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.40.ffn_norm.weight
F32
F32
[6144]
blk.40.ffn_up_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.40.ffn_up_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.40.indexer.attn_k.weight
Q4_K
Q4_K
[6144, 128]
blk.40.indexer.attn_q_b.weight
Q4_K
Q4_K
[2048, 4096]
blk.40.indexer.k_norm.bias
F32
F32
[128]
blk.40.indexer.k_norm.weight
F32
F32
[128]
blk.40.indexer.proj.weight
Q4_K
Q4_K
[6144, 32]
blk.41
blk.41.attn_k_b.weight
Q8_0
Q8_0
[192, 512, 64]
blk.41.attn_kv_a_mqa.weight
Q8_0
Q8_0
[6144, 576]
blk.41.attn_kv_a_norm.weight
F32
F32
[512]
blk.41.attn_norm.weight
F32
F32
[6144]
blk.41.attn_output.weight
Q4_K
Q4_K
[16384, 6144]
blk.41.attn_q_a.weight
Q4_K
Q4_K
[6144, 2048]
blk.41.attn_q_a_norm.weight
F32
F32
[2048]
blk.41.attn_q_b.weight
Q4_K
Q4_K
[2048, 16384]
blk.41.attn_v_b.weight
Q8_0
Q8_0
[512, 256, 64]
blk.41.exp_probs_b.bias
F32
F32
[256]
blk.41.ffn_down_exps.weight
Q4_K
Q4_K
[2048, 6144, 256]
blk.41.ffn_down_shexp.weight
Q6_K
Q6_K
[2048, 6144]
blk.41.ffn_gate_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.41.ffn_gate_inp.weight
F32
F32
[6144, 256]
blk.41.ffn_gate_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.41.ffn_norm.weight
F32
F32
[6144]
blk.41.ffn_up_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.41.ffn_up_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.41.indexer.attn_k.weight
Q4_K
Q4_K
[6144, 128]
blk.41.indexer.attn_q_b.weight
Q4_K
Q4_K
[2048, 4096]
blk.41.indexer.k_norm.bias
F32
F32
[128]
blk.41.indexer.k_norm.weight
F32
F32
[128]
blk.41.indexer.proj.weight
Q4_K
Q4_K
[6144, 32]
blk.42
blk.42.attn_k_b.weight
Q8_0
Q8_0
[192, 512, 64]
blk.42.attn_kv_a_mqa.weight
Q8_0
Q8_0
[6144, 576]
blk.42.attn_kv_a_norm.weight
F32
F32
[512]
blk.42.attn_norm.weight
F32
F32
[6144]
blk.42.attn_output.weight
Q4_K
Q4_K
[16384, 6144]
blk.42.attn_q_a.weight
Q4_K
Q4_K
[6144, 2048]
blk.42.attn_q_a_norm.weight
F32
F32
[2048]
blk.42.attn_q_b.weight
Q4_K
Q4_K
[2048, 16384]
blk.42.attn_v_b.weight
Q8_0
Q8_0
[512, 256, 64]
blk.42.exp_probs_b.bias
F32
F32
[256]
blk.42.ffn_down_exps.weight
Q4_K
Q4_K
[2048, 6144, 256]
blk.42.ffn_down_shexp.weight
Q6_K
Q6_K
[2048, 6144]
blk.42.ffn_gate_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.42.ffn_gate_inp.weight
F32
F32
[6144, 256]
blk.42.ffn_gate_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.42.ffn_norm.weight
F32
F32
[6144]
blk.42.ffn_up_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.42.ffn_up_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.42.indexer.attn_k.weight
Q4_K
Q4_K
[6144, 128]
blk.42.indexer.attn_q_b.weight
Q4_K
Q4_K
[2048, 4096]
blk.42.indexer.k_norm.bias
F32
F32
[128]
blk.42.indexer.k_norm.weight
F32
F32
[128]
blk.42.indexer.proj.weight
Q4_K
Q4_K
[6144, 32]
blk.43
blk.43.attn_k_b.weight
Q8_0
Q8_0
[192, 512, 64]
blk.43.attn_kv_a_mqa.weight
Q8_0
Q8_0
[6144, 576]
blk.43.attn_kv_a_norm.weight
F32
F32
[512]
blk.43.attn_norm.weight
F32
F32
[6144]
blk.43.attn_output.weight
Q4_K
Q4_K
[16384, 6144]
blk.43.attn_q_a.weight
Q4_K
Q4_K
[6144, 2048]
blk.43.attn_q_a_norm.weight
F32
F32
[2048]
blk.43.attn_q_b.weight
Q4_K
Q4_K
[2048, 16384]
blk.43.attn_v_b.weight
Q8_0
Q8_0
[512, 256, 64]
blk.43.exp_probs_b.bias
F32
F32
[256]
blk.43.ffn_down_exps.weight
Q4_K
Q4_K
[2048, 6144, 256]
blk.43.ffn_down_shexp.weight
Q6_K
Q6_K
[2048, 6144]
blk.43.ffn_gate_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.43.ffn_gate_inp.weight
F32
F32
[6144, 256]
blk.43.ffn_gate_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.43.ffn_norm.weight
F32
F32
[6144]
blk.43.ffn_up_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.43.ffn_up_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.43.indexer.attn_k.weight
Q4_K
Q4_K
[6144, 128]
blk.43.indexer.attn_q_b.weight
Q4_K
Q4_K
[2048, 4096]
blk.43.indexer.k_norm.bias
F32
F32
[128]
blk.43.indexer.k_norm.weight
F32
F32
[128]
blk.43.indexer.proj.weight
Q4_K
Q4_K
[6144, 32]
blk.44
blk.44.attn_k_b.weight
Q8_0
Q8_0
[192, 512, 64]
blk.44.attn_kv_a_mqa.weight
Q8_0
Q8_0
[6144, 576]
blk.44.attn_kv_a_norm.weight
F32
F32
[512]
blk.44.attn_norm.weight
F32
F32
[6144]
blk.44.attn_output.weight
Q4_K
Q4_K
[16384, 6144]
blk.44.attn_q_a.weight
Q4_K
Q4_K
[6144, 2048]
blk.44.attn_q_a_norm.weight
F32
F32
[2048]
blk.44.attn_q_b.weight
Q4_K
Q4_K
[2048, 16384]
blk.44.attn_v_b.weight
Q8_0
Q8_0
[512, 256, 64]
blk.44.exp_probs_b.bias
F32
F32
[256]
blk.44.ffn_down_exps.weight
Q4_K
Q4_K
[2048, 6144, 256]
blk.44.ffn_down_shexp.weight
Q6_K
Q6_K
[2048, 6144]
blk.44.ffn_gate_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.44.ffn_gate_inp.weight
F32
F32
[6144, 256]
blk.44.ffn_gate_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.44.ffn_norm.weight
F32
F32
[6144]
blk.44.ffn_up_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.44.ffn_up_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.44.indexer.attn_k.weight
Q4_K
Q4_K
[6144, 128]
blk.44.indexer.attn_q_b.weight
Q4_K
Q4_K
[2048, 4096]
blk.44.indexer.k_norm.bias
F32
F32
[128]
blk.44.indexer.k_norm.weight
F32
F32
[128]
blk.44.indexer.proj.weight
Q4_K
Q4_K
[6144, 32]
blk.45
blk.45.attn_k_b.weight
Q8_0
Q8_0
[192, 512, 64]
blk.45.attn_kv_a_mqa.weight
Q8_0
Q8_0
[6144, 576]
blk.45.attn_kv_a_norm.weight
F32
F32
[512]
blk.45.attn_norm.weight
F32
F32
[6144]
blk.45.attn_output.weight
Q4_K
Q4_K
[16384, 6144]
blk.45.attn_q_a.weight
Q4_K
Q4_K
[6144, 2048]
blk.45.attn_q_a_norm.weight
F32
F32
[2048]
blk.45.attn_q_b.weight
Q6_K
Q6_K
[2048, 16384]
blk.45.attn_v_b.weight
Q8_0
Q8_0
[512, 256, 64]
blk.45.exp_probs_b.bias
F32
F32
[256]
blk.45.ffn_down_exps.weight
Q4_K
Q4_K
[2048, 6144, 256]
blk.45.ffn_down_shexp.weight
Q6_K
Q6_K
[2048, 6144]
blk.45.ffn_gate_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.45.ffn_gate_inp.weight
F32
F32
[6144, 256]
blk.45.ffn_gate_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.45.ffn_norm.weight
F32
F32
[6144]
blk.45.ffn_up_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.45.ffn_up_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.45.indexer.attn_k.weight
Q4_K
Q4_K
[6144, 128]
blk.45.indexer.attn_q_b.weight
Q4_K
Q4_K
[2048, 4096]
blk.45.indexer.k_norm.bias
F32
F32
[128]
blk.45.indexer.k_norm.weight
F32
F32
[128]
blk.45.indexer.proj.weight
Q4_K
Q4_K
[6144, 32]
blk.46
blk.46.attn_k_b.weight
Q8_0
Q8_0
[192, 512, 64]
blk.46.attn_kv_a_mqa.weight
Q8_0
Q8_0
[6144, 576]
blk.46.attn_kv_a_norm.weight
F32
F32
[512]
blk.46.attn_norm.weight
F32
F32
[6144]
blk.46.attn_output.weight
Q4_K
Q4_K
[16384, 6144]
blk.46.attn_q_a.weight
Q4_K
Q4_K
[6144, 2048]
blk.46.attn_q_a_norm.weight
F32
F32
[2048]
blk.46.attn_q_b.weight
Q4_K
Q4_K
[2048, 16384]
blk.46.attn_v_b.weight
Q8_0
Q8_0
[512, 256, 64]
blk.46.exp_probs_b.bias
F32
F32
[256]
blk.46.ffn_down_exps.weight
Q4_K
Q4_K
[2048, 6144, 256]
blk.46.ffn_down_shexp.weight
Q6_K
Q6_K
[2048, 6144]
blk.46.ffn_gate_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.46.ffn_gate_inp.weight
F32
F32
[6144, 256]
blk.46.ffn_gate_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.46.ffn_norm.weight
F32
F32
[6144]
blk.46.ffn_up_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.46.ffn_up_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.46.indexer.attn_k.weight
Q4_K
Q4_K
[6144, 128]
blk.46.indexer.attn_q_b.weight
Q4_K
Q4_K
[2048, 4096]
blk.46.indexer.k_norm.bias
F32
F32
[128]
blk.46.indexer.k_norm.weight
F32
F32
[128]
blk.46.indexer.proj.weight
Q4_K
Q4_K
[6144, 32]
blk.47
blk.47.attn_k_b.weight
Q8_0
Q8_0
[192, 512, 64]
blk.47.attn_kv_a_mqa.weight
Q8_0
Q8_0
[6144, 576]
blk.47.attn_kv_a_norm.weight
F32
F32
[512]
blk.47.attn_norm.weight
F32
F32
[6144]
blk.47.attn_output.weight
Q4_K
Q4_K
[16384, 6144]
blk.47.attn_q_a.weight
Q4_K
Q4_K
[6144, 2048]
blk.47.attn_q_a_norm.weight
F32
F32
[2048]
blk.47.attn_q_b.weight
F16
F16
[2048, 16384]
blk.47.attn_v_b.weight
Q8_0
Q8_0
[512, 256, 64]
blk.47.exp_probs_b.bias
F32
F32
[256]
blk.47.ffn_down_exps.weight
Q4_K
Q4_K
[2048, 6144, 256]
blk.47.ffn_down_shexp.weight
Q6_K
Q6_K
[2048, 6144]
blk.47.ffn_gate_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.47.ffn_gate_inp.weight
F32
F32
[6144, 256]
blk.47.ffn_gate_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.47.ffn_norm.weight
F32
F32
[6144]
blk.47.ffn_up_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.47.ffn_up_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.47.indexer.attn_k.weight
Q4_K
Q4_K
[6144, 128]
blk.47.indexer.attn_q_b.weight
Q4_K
Q4_K
[2048, 4096]
blk.47.indexer.k_norm.bias
F32
F32
[128]
blk.47.indexer.k_norm.weight
F32
F32
[128]
blk.47.indexer.proj.weight
Q4_K
Q4_K
[6144, 32]
blk.48
blk.48.attn_k_b.weight
Q8_0
Q8_0
[192, 512, 64]
blk.48.attn_kv_a_mqa.weight
Q8_0
Q8_0
[6144, 576]
blk.48.attn_kv_a_norm.weight
F32
F32
[512]
blk.48.attn_norm.weight
F32
F32
[6144]
blk.48.attn_output.weight
Q4_K
Q4_K
[16384, 6144]
blk.48.attn_q_a.weight
Q4_K
Q4_K
[6144, 2048]
blk.48.attn_q_a_norm.weight
F32
F32
[2048]
blk.48.attn_q_b.weight
Q6_K
Q6_K
[2048, 16384]
blk.48.attn_v_b.weight
Q8_0
Q8_0
[512, 256, 64]
blk.48.exp_probs_b.bias
F32
F32
[256]
blk.48.ffn_down_exps.weight
Q4_K
Q4_K
[2048, 6144, 256]
blk.48.ffn_down_shexp.weight
Q6_K
Q6_K
[2048, 6144]
blk.48.ffn_gate_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.48.ffn_gate_inp.weight
F32
F32
[6144, 256]
blk.48.ffn_gate_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.48.ffn_norm.weight
F32
F32
[6144]
blk.48.ffn_up_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.48.ffn_up_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.48.indexer.attn_k.weight
Q4_K
Q4_K
[6144, 128]
blk.48.indexer.attn_q_b.weight
Q4_K
Q4_K
[2048, 4096]
blk.48.indexer.k_norm.bias
F32
F32
[128]
blk.48.indexer.k_norm.weight
F32
F32
[128]
blk.48.indexer.proj.weight
Q4_K
Q4_K
[6144, 32]
blk.49
blk.49.attn_k_b.weight
Q8_0
Q8_0
[192, 512, 64]
blk.49.attn_kv_a_mqa.weight
Q8_0
Q8_0
[6144, 576]
blk.49.attn_kv_a_norm.weight
F32
F32
[512]
blk.49.attn_norm.weight
F32
F32
[6144]
blk.49.attn_output.weight
Q4_K
Q4_K
[16384, 6144]
blk.49.attn_q_a.weight
Q4_K
Q4_K
[6144, 2048]
blk.49.attn_q_a_norm.weight
F32
F32
[2048]
blk.49.attn_q_b.weight
Q6_K
Q6_K
[2048, 16384]
blk.49.attn_v_b.weight
Q8_0
Q8_0
[512, 256, 64]
blk.49.exp_probs_b.bias
F32
F32
[256]
blk.49.ffn_down_exps.weight
Q4_K
Q4_K
[2048, 6144, 256]
blk.49.ffn_down_shexp.weight
Q6_K
Q6_K
[2048, 6144]
blk.49.ffn_gate_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.49.ffn_gate_inp.weight
F32
F32
[6144, 256]
blk.49.ffn_gate_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.49.ffn_norm.weight
F32
F32
[6144]
blk.49.ffn_up_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.49.ffn_up_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.49.indexer.attn_k.weight
Q4_K
Q4_K
[6144, 128]
blk.49.indexer.attn_q_b.weight
Q4_K
Q4_K
[2048, 4096]
blk.49.indexer.k_norm.bias
F32
F32
[128]
blk.49.indexer.k_norm.weight
F32
F32
[128]
blk.49.indexer.proj.weight
Q4_K
Q4_K
[6144, 32]
blk.50
blk.50.attn_k_b.weight
Q8_0
Q8_0
[192, 512, 64]
blk.50.attn_kv_a_mqa.weight
Q8_0
Q8_0
[6144, 576]
blk.50.attn_kv_a_norm.weight
F32
F32
[512]
blk.50.attn_norm.weight
F32
F32
[6144]
blk.50.attn_output.weight
Q4_K
Q4_K
[16384, 6144]
blk.50.attn_q_a.weight
Q4_K
Q4_K
[6144, 2048]
blk.50.attn_q_a_norm.weight
F32
F32
[2048]
blk.50.attn_q_b.weight
Q4_K
Q4_K
[2048, 16384]
blk.50.attn_v_b.weight
Q8_0
Q8_0
[512, 256, 64]
blk.50.exp_probs_b.bias
F32
F32
[256]
blk.50.ffn_down_exps.weight
Q4_K
Q4_K
[2048, 6144, 256]
blk.50.ffn_down_shexp.weight
Q6_K
Q6_K
[2048, 6144]
blk.50.ffn_gate_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.50.ffn_gate_inp.weight
F32
F32
[6144, 256]
blk.50.ffn_gate_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.50.ffn_norm.weight
F32
F32
[6144]
blk.50.ffn_up_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.50.ffn_up_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.50.indexer.attn_k.weight
Q4_K
Q4_K
[6144, 128]
blk.50.indexer.attn_q_b.weight
Q4_K
Q4_K
[2048, 4096]
blk.50.indexer.k_norm.bias
F32
F32
[128]
blk.50.indexer.k_norm.weight
F32
F32
[128]
blk.50.indexer.proj.weight
Q4_K
Q4_K
[6144, 32]
blk.51
blk.51.attn_k_b.weight
Q8_0
Q8_0
[192, 512, 64]
blk.51.attn_kv_a_mqa.weight
Q8_0
Q8_0
[6144, 576]
blk.51.attn_kv_a_norm.weight
F32
F32
[512]
blk.51.attn_norm.weight
F32
F32
[6144]
blk.51.attn_output.weight
Q4_K
Q4_K
[16384, 6144]
blk.51.attn_q_a.weight
Q4_K
Q4_K
[6144, 2048]
blk.51.attn_q_a_norm.weight
F32
F32
[2048]
blk.51.attn_q_b.weight
Q6_K
Q6_K
[2048, 16384]
blk.51.attn_v_b.weight
Q8_0
Q8_0
[512, 256, 64]
blk.51.exp_probs_b.bias
F32
F32
[256]
blk.51.ffn_down_exps.weight
Q4_K
Q4_K
[2048, 6144, 256]
blk.51.ffn_down_shexp.weight
Q6_K
Q6_K
[2048, 6144]
blk.51.ffn_gate_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.51.ffn_gate_inp.weight
F32
F32
[6144, 256]
blk.51.ffn_gate_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.51.ffn_norm.weight
F32
F32
[6144]
blk.51.ffn_up_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.51.ffn_up_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.51.indexer.attn_k.weight
Q4_K
Q4_K
[6144, 128]
blk.51.indexer.attn_q_b.weight
Q4_K
Q4_K
[2048, 4096]
blk.51.indexer.k_norm.bias
F32
F32
[128]
blk.51.indexer.k_norm.weight
F32
F32
[128]
blk.51.indexer.proj.weight
Q4_K
Q4_K
[6144, 32]
blk.52
blk.52.attn_k_b.weight
Q8_0
Q8_0
[192, 512, 64]
blk.52.attn_kv_a_mqa.weight
Q8_0
Q8_0
[6144, 576]
blk.52.attn_kv_a_norm.weight
F32
F32
[512]
blk.52.attn_norm.weight
F32
F32
[6144]
blk.52.attn_output.weight
Q4_K
Q4_K
[16384, 6144]
blk.52.attn_q_a.weight
Q4_K
Q4_K
[6144, 2048]
blk.52.attn_q_a_norm.weight
F32
F32
[2048]
blk.52.attn_q_b.weight
Q6_K
Q6_K
[2048, 16384]
blk.52.attn_v_b.weight
Q8_0
Q8_0
[512, 256, 64]
blk.52.exp_probs_b.bias
F32
F32
[256]
blk.52.ffn_down_exps.weight
Q4_K
Q4_K
[2048, 6144, 256]
blk.52.ffn_down_shexp.weight
Q6_K
Q6_K
[2048, 6144]
blk.52.ffn_gate_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.52.ffn_gate_inp.weight
F32
F32
[6144, 256]
blk.52.ffn_gate_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.52.ffn_norm.weight
F32
F32
[6144]
blk.52.ffn_up_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.52.ffn_up_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.52.indexer.attn_k.weight
Q4_K
Q4_K
[6144, 128]
blk.52.indexer.attn_q_b.weight
Q4_K
Q4_K
[2048, 4096]
blk.52.indexer.k_norm.bias
F32
F32
[128]
blk.52.indexer.k_norm.weight
F32
F32
[128]
blk.52.indexer.proj.weight
Q4_K
Q4_K
[6144, 32]
blk.53
blk.53.attn_k_b.weight
Q8_0
Q8_0
[192, 512, 64]
blk.53.attn_kv_a_mqa.weight
Q8_0
Q8_0
[6144, 576]
blk.53.attn_kv_a_norm.weight
F32
F32
[512]
blk.53.attn_norm.weight
F32
F32
[6144]
blk.53.attn_output.weight
Q4_K
Q4_K
[16384, 6144]
blk.53.attn_q_a.weight
Q4_K
Q4_K
[6144, 2048]
blk.53.attn_q_a_norm.weight
F32
F32
[2048]
blk.53.attn_q_b.weight
Q4_K
Q4_K
[2048, 16384]
blk.53.attn_v_b.weight
Q8_0
Q8_0
[512, 256, 64]
blk.53.exp_probs_b.bias
F32
F32
[256]
blk.53.ffn_down_exps.weight
Q4_K
Q4_K
[2048, 6144, 256]
blk.53.ffn_down_shexp.weight
Q6_K
Q6_K
[2048, 6144]
blk.53.ffn_gate_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.53.ffn_gate_inp.weight
F32
F32
[6144, 256]
blk.53.ffn_gate_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.53.ffn_norm.weight
F32
F32
[6144]
blk.53.ffn_up_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.53.ffn_up_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.53.indexer.attn_k.weight
Q4_K
Q4_K
[6144, 128]
blk.53.indexer.attn_q_b.weight
Q4_K
Q4_K
[2048, 4096]
blk.53.indexer.k_norm.bias
F32
F32
[128]
blk.53.indexer.k_norm.weight
F32
F32
[128]
blk.53.indexer.proj.weight
Q4_K
Q4_K
[6144, 32]
blk.54
blk.54.attn_k_b.weight
Q8_0
Q8_0
[192, 512, 64]
blk.54.attn_kv_a_mqa.weight
Q8_0
Q8_0
[6144, 576]
blk.54.attn_kv_a_norm.weight
F32
F32
[512]
blk.54.attn_norm.weight
F32
F32
[6144]
blk.54.attn_output.weight
Q4_K
Q4_K
[16384, 6144]
blk.54.attn_q_a.weight
Q4_K
Q4_K
[6144, 2048]
blk.54.attn_q_a_norm.weight
F32
F32
[2048]
blk.54.attn_q_b.weight
Q4_K
Q4_K
[2048, 16384]
blk.54.attn_v_b.weight
Q8_0
Q8_0
[512, 256, 64]
blk.54.exp_probs_b.bias
F32
F32
[256]
blk.54.ffn_down_exps.weight
Q4_K
Q4_K
[2048, 6144, 256]
blk.54.ffn_down_shexp.weight
Q6_K
Q6_K
[2048, 6144]
blk.54.ffn_gate_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.54.ffn_gate_inp.weight
F32
F32
[6144, 256]
blk.54.ffn_gate_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.54.ffn_norm.weight
F32
F32
[6144]
blk.54.ffn_up_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.54.ffn_up_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.54.indexer.attn_k.weight
Q4_K
Q4_K
[6144, 128]
blk.54.indexer.attn_q_b.weight
Q4_K
Q4_K
[2048, 4096]
blk.54.indexer.k_norm.bias
F32
F32
[128]
blk.54.indexer.k_norm.weight
F32
F32
[128]
blk.54.indexer.proj.weight
Q4_K
Q4_K
[6144, 32]
blk.55
blk.55.attn_k_b.weight
Q8_0
Q8_0
[192, 512, 64]
blk.55.attn_kv_a_mqa.weight
Q8_0
Q8_0
[6144, 576]
blk.55.attn_kv_a_norm.weight
F32
F32
[512]
blk.55.attn_norm.weight
F32
F32
[6144]
blk.55.attn_output.weight
Q4_K
Q4_K
[16384, 6144]
blk.55.attn_q_a.weight
Q4_K
Q4_K
[6144, 2048]
blk.55.attn_q_a_norm.weight
F32
F32
[2048]
blk.55.attn_q_b.weight
Q4_K
Q4_K
[2048, 16384]
blk.55.attn_v_b.weight
Q8_0
Q8_0
[512, 256, 64]
blk.55.exp_probs_b.bias
F32
F32
[256]
blk.55.ffn_down_exps.weight
Q4_K
Q4_K
[2048, 6144, 256]
blk.55.ffn_down_shexp.weight
Q6_K
Q6_K
[2048, 6144]
blk.55.ffn_gate_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.55.ffn_gate_inp.weight
F32
F32
[6144, 256]
blk.55.ffn_gate_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.55.ffn_norm.weight
F32
F32
[6144]
blk.55.ffn_up_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.55.ffn_up_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.55.indexer.attn_k.weight
Q4_K
Q4_K
[6144, 128]
blk.55.indexer.attn_q_b.weight
Q4_K
Q4_K
[2048, 4096]
blk.55.indexer.k_norm.bias
F32
F32
[128]
blk.55.indexer.k_norm.weight
F32
F32
[128]
blk.55.indexer.proj.weight
Q4_K
Q4_K
[6144, 32]
blk.56
blk.56.attn_k_b.weight
Q8_0
Q8_0
[192, 512, 64]
blk.56.attn_kv_a_mqa.weight
Q8_0
Q8_0
[6144, 576]
blk.56.attn_kv_a_norm.weight
F32
F32
[512]
blk.56.attn_norm.weight
F32
F32
[6144]
blk.56.attn_output.weight
Q4_K
Q4_K
[16384, 6144]
blk.56.attn_q_a.weight
Q4_K
Q4_K
[6144, 2048]
blk.56.attn_q_a_norm.weight
F32
F32
[2048]
blk.56.attn_q_b.weight
Q4_K
Q4_K
[2048, 16384]
blk.56.attn_v_b.weight
Q8_0
Q8_0
[512, 256, 64]
blk.56.exp_probs_b.bias
F32
F32
[256]
blk.56.ffn_down_exps.weight
Q4_K
Q4_K
[2048, 6144, 256]
blk.56.ffn_down_shexp.weight
Q6_K
Q6_K
[2048, 6144]
blk.56.ffn_gate_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.56.ffn_gate_inp.weight
F32
F32
[6144, 256]
blk.56.ffn_gate_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.56.ffn_norm.weight
F32
F32
[6144]
blk.56.ffn_up_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.56.ffn_up_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.56.indexer.attn_k.weight
Q4_K
Q4_K
[6144, 128]
blk.56.indexer.attn_q_b.weight
Q4_K
Q4_K
[2048, 4096]
blk.56.indexer.k_norm.bias
F32
F32
[128]
blk.56.indexer.k_norm.weight
F32
F32
[128]
blk.56.indexer.proj.weight
Q4_K
Q4_K
[6144, 32]
blk.57
blk.57.attn_k_b.weight
Q8_0
Q8_0
[192, 512, 64]
blk.57.attn_kv_a_mqa.weight
Q8_0
Q8_0
[6144, 576]
blk.57.attn_kv_a_norm.weight
F32
F32
[512]
blk.57.attn_norm.weight
F32
F32
[6144]
blk.57.attn_output.weight
Q4_K
Q4_K
[16384, 6144]
blk.57.attn_q_a.weight
Q4_K
Q4_K
[6144, 2048]
blk.57.attn_q_a_norm.weight
F32
F32
[2048]
blk.57.attn_q_b.weight
Q6_K
Q6_K
[2048, 16384]
blk.57.attn_v_b.weight
Q8_0
Q8_0
[512, 256, 64]
blk.57.exp_probs_b.bias
F32
F32
[256]
blk.57.ffn_down_exps.weight
Q4_K
Q4_K
[2048, 6144, 256]
blk.57.ffn_down_shexp.weight
Q6_K
Q6_K
[2048, 6144]
blk.57.ffn_gate_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.57.ffn_gate_inp.weight
F32
F32
[6144, 256]
blk.57.ffn_gate_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.57.ffn_norm.weight
F32
F32
[6144]
blk.57.ffn_up_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.57.ffn_up_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.57.indexer.attn_k.weight
Q4_K
Q4_K
[6144, 128]
blk.57.indexer.attn_q_b.weight
Q4_K
Q4_K
[2048, 4096]
blk.57.indexer.k_norm.bias
F32
F32
[128]
blk.57.indexer.k_norm.weight
F32
F32
[128]
blk.57.indexer.proj.weight
Q4_K
Q4_K
[6144, 32]
blk.58
blk.58.attn_k_b.weight
Q8_0
Q8_0
[192, 512, 64]
blk.58.attn_kv_a_mqa.weight
Q8_0
Q8_0
[6144, 576]
blk.58.attn_kv_a_norm.weight
F32
F32
[512]
blk.58.attn_norm.weight
F32
F32
[6144]
blk.58.attn_output.weight
Q4_K
Q4_K
[16384, 6144]
blk.58.attn_q_a.weight
Q4_K
Q4_K
[6144, 2048]
blk.58.attn_q_a_norm.weight
F32
F32
[2048]
blk.58.attn_q_b.weight
Q6_K
Q6_K
[2048, 16384]
blk.58.attn_v_b.weight
Q8_0
Q8_0
[512, 256, 64]
blk.58.exp_probs_b.bias
F32
F32
[256]
blk.58.ffn_down_exps.weight
Q4_K
Q4_K
[2048, 6144, 256]
blk.58.ffn_down_shexp.weight
Q6_K
Q6_K
[2048, 6144]
blk.58.ffn_gate_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.58.ffn_gate_inp.weight
F32
F32
[6144, 256]
blk.58.ffn_gate_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.58.ffn_norm.weight
F32
F32
[6144]
blk.58.ffn_up_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.58.ffn_up_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.58.indexer.attn_k.weight
Q4_K
Q4_K
[6144, 128]
blk.58.indexer.attn_q_b.weight
Q4_K
Q4_K
[2048, 4096]
blk.58.indexer.k_norm.bias
F32
F32
[128]
blk.58.indexer.k_norm.weight
F32
F32
[128]
blk.58.indexer.proj.weight
Q4_K
Q4_K
[6144, 32]
blk.59
blk.59.attn_k_b.weight
Q8_0
Q8_0
[192, 512, 64]
blk.59.attn_kv_a_mqa.weight
Q8_0
Q8_0
[6144, 576]
blk.59.attn_kv_a_norm.weight
F32
F32
[512]
blk.59.attn_norm.weight
F32
F32
[6144]
blk.59.attn_output.weight
Q4_K
Q4_K
[16384, 6144]
blk.59.attn_q_a.weight
Q4_K
Q4_K
[6144, 2048]
blk.59.attn_q_a_norm.weight
F32
F32
[2048]
blk.59.attn_q_b.weight
Q6_K
Q6_K
[2048, 16384]
blk.59.attn_v_b.weight
Q8_0
Q8_0
[512, 256, 64]
blk.59.exp_probs_b.bias
F32
F32
[256]
blk.59.ffn_down_exps.weight
Q4_K
Q4_K
[2048, 6144, 256]
blk.59.ffn_down_shexp.weight
Q6_K
Q6_K
[2048, 6144]
blk.59.ffn_gate_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.59.ffn_gate_inp.weight
F32
F32
[6144, 256]
blk.59.ffn_gate_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.59.ffn_norm.weight
F32
F32
[6144]
blk.59.ffn_up_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.59.ffn_up_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.59.indexer.attn_k.weight
Q4_K
Q4_K
[6144, 128]
blk.59.indexer.attn_q_b.weight
Q4_K
Q4_K
[2048, 4096]
blk.59.indexer.k_norm.bias
F32
F32
[128]
blk.59.indexer.k_norm.weight
F32
F32
[128]
blk.59.indexer.proj.weight
Q4_K
Q4_K
[6144, 32]
blk.60
blk.60.attn_k_b.weight
Q8_0
Q8_0
[192, 512, 64]
blk.60.attn_kv_a_mqa.weight
Q8_0
Q8_0
[6144, 576]
blk.60.attn_kv_a_norm.weight
F32
F32
[512]
blk.60.attn_norm.weight
F32
F32
[6144]
blk.60.attn_output.weight
Q4_K
Q4_K
[16384, 6144]
blk.60.attn_q_a.weight
Q4_K
Q4_K
[6144, 2048]
blk.60.attn_q_a_norm.weight
F32
F32
[2048]
blk.60.attn_q_b.weight
Q4_K
Q4_K
[2048, 16384]
blk.60.attn_v_b.weight
Q8_0
Q8_0
[512, 256, 64]
blk.60.exp_probs_b.bias
F32
F32
[256]
blk.60.ffn_down_exps.weight
Q4_K
Q4_K
[2048, 6144, 256]
blk.60.ffn_down_shexp.weight
Q6_K
Q6_K
[2048, 6144]
blk.60.ffn_gate_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.60.ffn_gate_inp.weight
F32
F32
[6144, 256]
blk.60.ffn_gate_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.60.ffn_norm.weight
F32
F32
[6144]
blk.60.ffn_up_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.60.ffn_up_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.60.indexer.attn_k.weight
Q4_K
Q4_K
[6144, 128]
blk.60.indexer.attn_q_b.weight
Q4_K
Q4_K
[2048, 4096]
blk.60.indexer.k_norm.bias
F32
F32
[128]
blk.60.indexer.k_norm.weight
F32
F32
[128]
blk.60.indexer.proj.weight
Q4_K
Q4_K
[6144, 32]
blk.61
blk.61.attn_k_b.weight
Q8_0
Q8_0
[192, 512, 64]
blk.61.attn_kv_a_mqa.weight
Q8_0
Q8_0
[6144, 576]
blk.61.attn_kv_a_norm.weight
F32
F32
[512]
blk.61.attn_norm.weight
F32
F32
[6144]
blk.61.attn_output.weight
Q4_K
Q4_K
[16384, 6144]
blk.61.attn_q_a.weight
Q4_K
Q4_K
[6144, 2048]
blk.61.attn_q_a_norm.weight
F32
F32
[2048]
blk.61.attn_q_b.weight
Q6_K
Q6_K
[2048, 16384]
blk.61.attn_v_b.weight
Q8_0
Q8_0
[512, 256, 64]
blk.61.exp_probs_b.bias
F32
F32
[256]
blk.61.ffn_down_exps.weight
Q4_K
Q4_K
[2048, 6144, 256]
blk.61.ffn_down_shexp.weight
Q6_K
Q6_K
[2048, 6144]
blk.61.ffn_gate_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.61.ffn_gate_inp.weight
F32
F32
[6144, 256]
blk.61.ffn_gate_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.61.ffn_norm.weight
F32
F32
[6144]
blk.61.ffn_up_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.61.ffn_up_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.61.indexer.attn_k.weight
Q4_K
Q4_K
[6144, 128]
blk.61.indexer.attn_q_b.weight
Q4_K
Q4_K
[2048, 4096]
blk.61.indexer.k_norm.bias
F32
F32
[128]
blk.61.indexer.k_norm.weight
F32
F32
[128]
blk.61.indexer.proj.weight
Q4_K
Q4_K
[6144, 32]
blk.62
blk.62.attn_k_b.weight
Q8_0
Q8_0
[192, 512, 64]
blk.62.attn_kv_a_mqa.weight
Q8_0
Q8_0
[6144, 576]
blk.62.attn_kv_a_norm.weight
F32
F32
[512]
blk.62.attn_norm.weight
F32
F32
[6144]
blk.62.attn_output.weight
Q4_K
Q4_K
[16384, 6144]
blk.62.attn_q_a.weight
Q4_K
Q4_K
[6144, 2048]
blk.62.attn_q_a_norm.weight
F32
F32
[2048]
blk.62.attn_q_b.weight
Q8_0
Q8_0
[2048, 16384]
blk.62.attn_v_b.weight
Q8_0
Q8_0
[512, 256, 64]
blk.62.exp_probs_b.bias
F32
F32
[256]
blk.62.ffn_down_exps.weight
Q4_K
Q4_K
[2048, 6144, 256]
blk.62.ffn_down_shexp.weight
Q6_K
Q6_K
[2048, 6144]
blk.62.ffn_gate_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.62.ffn_gate_inp.weight
F32
F32
[6144, 256]
blk.62.ffn_gate_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.62.ffn_norm.weight
F32
F32
[6144]
blk.62.ffn_up_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.62.ffn_up_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.62.indexer.attn_k.weight
Q4_K
Q4_K
[6144, 128]
blk.62.indexer.attn_q_b.weight
Q4_K
Q4_K
[2048, 4096]
blk.62.indexer.k_norm.bias
F32
F32
[128]
blk.62.indexer.k_norm.weight
F32
F32
[128]
blk.62.indexer.proj.weight
Q4_K
Q4_K
[6144, 32]
blk.63
blk.63.attn_k_b.weight
Q8_0
Q8_0
[192, 512, 64]
blk.63.attn_kv_a_mqa.weight
Q8_0
Q8_0
[6144, 576]
blk.63.attn_kv_a_norm.weight
F32
F32
[512]
blk.63.attn_norm.weight
F32
F32
[6144]
blk.63.attn_output.weight
Q4_K
Q4_K
[16384, 6144]
blk.63.attn_q_a.weight
Q4_K
Q4_K
[6144, 2048]
blk.63.attn_q_a_norm.weight
F32
F32
[2048]
blk.63.attn_q_b.weight
Q6_K
Q6_K
[2048, 16384]
blk.63.attn_v_b.weight
Q8_0
Q8_0
[512, 256, 64]
blk.63.exp_probs_b.bias
F32
F32
[256]
blk.63.ffn_down_exps.weight
Q4_K
Q4_K
[2048, 6144, 256]
blk.63.ffn_down_shexp.weight
Q6_K
Q6_K
[2048, 6144]
blk.63.ffn_gate_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.63.ffn_gate_inp.weight
F32
F32
[6144, 256]
blk.63.ffn_gate_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.63.ffn_norm.weight
F32
F32
[6144]
blk.63.ffn_up_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.63.ffn_up_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.63.indexer.attn_k.weight
Q4_K
Q4_K
[6144, 128]
blk.63.indexer.attn_q_b.weight
Q4_K
Q4_K
[2048, 4096]
blk.63.indexer.k_norm.bias
F32
F32
[128]
blk.63.indexer.k_norm.weight
F32
F32
[128]
blk.63.indexer.proj.weight
Q4_K
Q4_K
[6144, 32]
blk.64
blk.64.attn_k_b.weight
Q8_0
Q8_0
[192, 512, 64]
blk.64.attn_kv_a_mqa.weight
Q8_0
Q8_0
[6144, 576]
blk.64.attn_kv_a_norm.weight
F32
F32
[512]
blk.64.attn_norm.weight
F32
F32
[6144]
blk.64.attn_output.weight
Q4_K
Q4_K
[16384, 6144]
blk.64.attn_q_a.weight
Q4_K
Q4_K
[6144, 2048]
blk.64.attn_q_a_norm.weight
F32
F32
[2048]
blk.64.attn_q_b.weight
Q6_K
Q6_K
[2048, 16384]
blk.64.attn_v_b.weight
Q8_0
Q8_0
[512, 256, 64]
blk.64.exp_probs_b.bias
F32
F32
[256]
blk.64.ffn_down_exps.weight
Q4_K
Q4_K
[2048, 6144, 256]
blk.64.ffn_down_shexp.weight
Q6_K
Q6_K
[2048, 6144]
blk.64.ffn_gate_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.64.ffn_gate_inp.weight
F32
F32
[6144, 256]
blk.64.ffn_gate_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.64.ffn_norm.weight
F32
F32
[6144]
blk.64.ffn_up_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.64.ffn_up_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.64.indexer.attn_k.weight
Q4_K
Q4_K
[6144, 128]
blk.64.indexer.attn_q_b.weight
Q4_K
Q4_K
[2048, 4096]
blk.64.indexer.k_norm.bias
F32
F32
[128]
blk.64.indexer.k_norm.weight
F32
F32
[128]
blk.64.indexer.proj.weight
Q4_K
Q4_K
[6144, 32]
blk.65
blk.65.attn_k_b.weight
Q8_0
Q8_0
[192, 512, 64]
blk.65.attn_kv_a_mqa.weight
Q8_0
Q8_0
[6144, 576]
blk.65.attn_kv_a_norm.weight
F32
F32
[512]
blk.65.attn_norm.weight
F32
F32
[6144]
blk.65.attn_output.weight
Q4_K
Q4_K
[16384, 6144]
blk.65.attn_q_a.weight
Q4_K
Q4_K
[6144, 2048]
blk.65.attn_q_a_norm.weight
F32
F32
[2048]
blk.65.attn_q_b.weight
Q6_K
Q6_K
[2048, 16384]
blk.65.attn_v_b.weight
Q8_0
Q8_0
[512, 256, 64]
blk.65.exp_probs_b.bias
F32
F32
[256]
blk.65.ffn_down_exps.weight
Q4_K
Q4_K
[2048, 6144, 256]
blk.65.ffn_down_shexp.weight
Q6_K
Q6_K
[2048, 6144]
blk.65.ffn_gate_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.65.ffn_gate_inp.weight
F32
F32
[6144, 256]
blk.65.ffn_gate_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.65.ffn_norm.weight
F32
F32
[6144]
blk.65.ffn_up_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.65.ffn_up_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.65.indexer.attn_k.weight
Q4_K
Q4_K
[6144, 128]
blk.65.indexer.attn_q_b.weight
Q4_K
Q4_K
[2048, 4096]
blk.65.indexer.k_norm.bias
F32
F32
[128]
blk.65.indexer.k_norm.weight
F32
F32
[128]
blk.65.indexer.proj.weight
Q4_K
Q4_K
[6144, 32]
blk.66
blk.66.attn_k_b.weight
Q8_0
Q8_0
[192, 512, 64]
blk.66.attn_kv_a_mqa.weight
Q8_0
Q8_0
[6144, 576]
blk.66.attn_kv_a_norm.weight
F32
F32
[512]
blk.66.attn_norm.weight
F32
F32
[6144]
blk.66.attn_output.weight
Q4_K
Q4_K
[16384, 6144]
blk.66.attn_q_a.weight
Q4_K
Q4_K
[6144, 2048]
blk.66.attn_q_a_norm.weight
F32
F32
[2048]
blk.66.attn_q_b.weight
Q6_K
Q6_K
[2048, 16384]
blk.66.attn_v_b.weight
Q8_0
Q8_0
[512, 256, 64]
blk.66.exp_probs_b.bias
F32
F32
[256]
blk.66.ffn_down_exps.weight
Q4_K
Q4_K
[2048, 6144, 256]
blk.66.ffn_down_shexp.weight
Q6_K
Q6_K
[2048, 6144]
blk.66.ffn_gate_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.66.ffn_gate_inp.weight
F32
F32
[6144, 256]
blk.66.ffn_gate_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.66.ffn_norm.weight
F32
F32
[6144]
blk.66.ffn_up_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.66.ffn_up_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.66.indexer.attn_k.weight
Q4_K
Q4_K
[6144, 128]
blk.66.indexer.attn_q_b.weight
Q4_K
Q4_K
[2048, 4096]
blk.66.indexer.k_norm.bias
F32
F32
[128]
blk.66.indexer.k_norm.weight
F32
F32
[128]
blk.66.indexer.proj.weight
Q4_K
Q4_K
[6144, 32]
blk.67
blk.67.attn_k_b.weight
Q8_0
Q8_0
[192, 512, 64]
blk.67.attn_kv_a_mqa.weight
Q8_0
Q8_0
[6144, 576]
blk.67.attn_kv_a_norm.weight
F32
F32
[512]
blk.67.attn_norm.weight
F32
F32
[6144]
blk.67.attn_output.weight
Q4_K
Q4_K
[16384, 6144]
blk.67.attn_q_a.weight
Q4_K
Q4_K
[6144, 2048]
blk.67.attn_q_a_norm.weight
F32
F32
[2048]
blk.67.attn_q_b.weight
Q8_0
Q8_0
[2048, 16384]
blk.67.attn_v_b.weight
Q8_0
Q8_0
[512, 256, 64]
blk.67.exp_probs_b.bias
F32
F32
[256]
blk.67.ffn_down_exps.weight
Q4_K
Q4_K
[2048, 6144, 256]
blk.67.ffn_down_shexp.weight
Q6_K
Q6_K
[2048, 6144]
blk.67.ffn_gate_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.67.ffn_gate_inp.weight
F32
F32
[6144, 256]
blk.67.ffn_gate_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.67.ffn_norm.weight
F32
F32
[6144]
blk.67.ffn_up_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.67.ffn_up_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.67.indexer.attn_k.weight
Q4_K
Q4_K
[6144, 128]
blk.67.indexer.attn_q_b.weight
Q4_K
Q4_K
[2048, 4096]
blk.67.indexer.k_norm.bias
F32
F32
[128]
blk.67.indexer.k_norm.weight
F32
F32
[128]
blk.67.indexer.proj.weight
Q4_K
Q4_K
[6144, 32]
blk.68
blk.68.attn_k_b.weight
Q8_0
Q8_0
[192, 512, 64]
blk.68.attn_kv_a_mqa.weight
Q8_0
Q8_0
[6144, 576]
blk.68.attn_kv_a_norm.weight
F32
F32
[512]
blk.68.attn_norm.weight
F32
F32
[6144]
blk.68.attn_output.weight
Q4_K
Q4_K
[16384, 6144]
blk.68.attn_q_a.weight
Q4_K
Q4_K
[6144, 2048]
blk.68.attn_q_a_norm.weight
F32
F32
[2048]
blk.68.attn_q_b.weight
Q8_0
Q8_0
[2048, 16384]
blk.68.attn_v_b.weight
Q8_0
Q8_0
[512, 256, 64]
blk.68.exp_probs_b.bias
F32
F32
[256]
blk.68.ffn_down_exps.weight
Q4_K
Q4_K
[2048, 6144, 256]
blk.68.ffn_down_shexp.weight
Q6_K
Q6_K
[2048, 6144]
blk.68.ffn_gate_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.68.ffn_gate_inp.weight
F32
F32
[6144, 256]
blk.68.ffn_gate_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.68.ffn_norm.weight
F32
F32
[6144]
blk.68.ffn_up_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.68.ffn_up_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.68.indexer.attn_k.weight
Q4_K
Q4_K
[6144, 128]
blk.68.indexer.attn_q_b.weight
Q4_K
Q4_K
[2048, 4096]
blk.68.indexer.k_norm.bias
F32
F32
[128]
blk.68.indexer.k_norm.weight
F32
F32
[128]
blk.68.indexer.proj.weight
Q4_K
Q4_K
[6144, 32]
blk.69
blk.69.attn_k_b.weight
Q8_0
Q8_0
[192, 512, 64]
blk.69.attn_kv_a_mqa.weight
Q8_0
Q8_0
[6144, 576]
blk.69.attn_kv_a_norm.weight
F32
F32
[512]
blk.69.attn_norm.weight
F32
F32
[6144]
blk.69.attn_output.weight
Q4_K
Q4_K
[16384, 6144]
blk.69.attn_q_a.weight
Q4_K
Q4_K
[6144, 2048]
blk.69.attn_q_a_norm.weight
F32
F32
[2048]
blk.69.attn_q_b.weight
Q4_K
Q4_K
[2048, 16384]
blk.69.attn_v_b.weight
Q8_0
Q8_0
[512, 256, 64]
blk.69.exp_probs_b.bias
F32
F32
[256]
blk.69.ffn_down_exps.weight
Q4_K
Q4_K
[2048, 6144, 256]
blk.69.ffn_down_shexp.weight
Q6_K
Q6_K
[2048, 6144]
blk.69.ffn_gate_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.69.ffn_gate_inp.weight
F32
F32
[6144, 256]
blk.69.ffn_gate_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.69.ffn_norm.weight
F32
F32
[6144]
blk.69.ffn_up_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.69.ffn_up_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.69.indexer.attn_k.weight
Q4_K
Q4_K
[6144, 128]
blk.69.indexer.attn_q_b.weight
Q4_K
Q4_K
[2048, 4096]
blk.69.indexer.k_norm.bias
F32
F32
[128]
blk.69.indexer.k_norm.weight
F32
F32
[128]
blk.69.indexer.proj.weight
Q4_K
Q4_K
[6144, 32]
blk.70
blk.70.attn_k_b.weight
Q8_0
Q8_0
[192, 512, 64]
blk.70.attn_kv_a_mqa.weight
Q8_0
Q8_0
[6144, 576]
blk.70.attn_kv_a_norm.weight
F32
F32
[512]
blk.70.attn_norm.weight
F32
F32
[6144]
blk.70.attn_output.weight
Q4_K
Q4_K
[16384, 6144]
blk.70.attn_q_a.weight
Q4_K
Q4_K
[6144, 2048]
blk.70.attn_q_a_norm.weight
F32
F32
[2048]
blk.70.attn_q_b.weight
Q4_K
Q4_K
[2048, 16384]
blk.70.attn_v_b.weight
Q8_0
Q8_0
[512, 256, 64]
blk.70.exp_probs_b.bias
F32
F32
[256]
blk.70.ffn_down_exps.weight
Q4_K
Q4_K
[2048, 6144, 256]
blk.70.ffn_down_shexp.weight
Q6_K
Q6_K
[2048, 6144]
blk.70.ffn_gate_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.70.ffn_gate_inp.weight
F32
F32
[6144, 256]
blk.70.ffn_gate_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.70.ffn_norm.weight
F32
F32
[6144]
blk.70.ffn_up_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.70.ffn_up_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.70.indexer.attn_k.weight
Q4_K
Q4_K
[6144, 128]
blk.70.indexer.attn_q_b.weight
Q4_K
Q4_K
[2048, 4096]
blk.70.indexer.k_norm.bias
F32
F32
[128]
blk.70.indexer.k_norm.weight
F32
F32
[128]
blk.70.indexer.proj.weight
Q4_K
Q4_K
[6144, 32]
blk.71
blk.71.attn_k_b.weight
Q8_0
Q8_0
[192, 512, 64]
blk.71.attn_kv_a_mqa.weight
Q8_0
Q8_0
[6144, 576]
blk.71.attn_kv_a_norm.weight
F32
F32
[512]
blk.71.attn_norm.weight
F32
F32
[6144]
blk.71.attn_output.weight
Q4_K
Q4_K
[16384, 6144]
blk.71.attn_q_a.weight
Q4_K
Q4_K
[6144, 2048]
blk.71.attn_q_a_norm.weight
F32
F32
[2048]
blk.71.attn_q_b.weight
Q6_K
Q6_K
[2048, 16384]
blk.71.attn_v_b.weight
Q8_0
Q8_0
[512, 256, 64]
blk.71.exp_probs_b.bias
F32
F32
[256]
blk.71.ffn_down_exps.weight
Q4_K
Q4_K
[2048, 6144, 256]
blk.71.ffn_down_shexp.weight
Q6_K
Q6_K
[2048, 6144]
blk.71.ffn_gate_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.71.ffn_gate_inp.weight
F32
F32
[6144, 256]
blk.71.ffn_gate_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.71.ffn_norm.weight
F32
F32
[6144]
blk.71.ffn_up_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.71.ffn_up_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.71.indexer.attn_k.weight
Q4_K
Q4_K
[6144, 128]
blk.71.indexer.attn_q_b.weight
Q4_K
Q4_K
[2048, 4096]
blk.71.indexer.k_norm.bias
F32
F32
[128]
blk.71.indexer.k_norm.weight
F32
F32
[128]
blk.71.indexer.proj.weight
Q4_K
Q4_K
[6144, 32]
blk.72
blk.72.attn_k_b.weight
Q8_0
Q8_0
[192, 512, 64]
blk.72.attn_kv_a_mqa.weight
Q8_0
Q8_0
[6144, 576]
blk.72.attn_kv_a_norm.weight
F32
F32
[512]
blk.72.attn_norm.weight
F32
F32
[6144]
blk.72.attn_output.weight
Q4_K
Q4_K
[16384, 6144]
blk.72.attn_q_a.weight
Q4_K
Q4_K
[6144, 2048]
blk.72.attn_q_a_norm.weight
F32
F32
[2048]
blk.72.attn_q_b.weight
Q8_0
Q8_0
[2048, 16384]
blk.72.attn_v_b.weight
Q8_0
Q8_0
[512, 256, 64]
blk.72.exp_probs_b.bias
F32
F32
[256]
blk.72.ffn_down_exps.weight
Q5_K
Q5_K
[2048, 6144, 256]
blk.72.ffn_down_shexp.weight
Q6_K
Q6_K
[2048, 6144]
blk.72.ffn_gate_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.72.ffn_gate_inp.weight
F32
F32
[6144, 256]
blk.72.ffn_gate_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.72.ffn_norm.weight
F32
F32
[6144]
blk.72.ffn_up_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.72.ffn_up_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.72.indexer.attn_k.weight
Q4_K
Q4_K
[6144, 128]
blk.72.indexer.attn_q_b.weight
Q4_K
Q4_K
[2048, 4096]
blk.72.indexer.k_norm.bias
F32
F32
[128]
blk.72.indexer.k_norm.weight
F32
F32
[128]
blk.72.indexer.proj.weight
Q4_K
Q4_K
[6144, 32]
blk.73
blk.73.attn_k_b.weight
Q8_0
Q8_0
[192, 512, 64]
blk.73.attn_kv_a_mqa.weight
Q8_0
Q8_0
[6144, 576]
blk.73.attn_kv_a_norm.weight
F32
F32
[512]
blk.73.attn_norm.weight
F32
F32
[6144]
blk.73.attn_output.weight
Q4_K
Q4_K
[16384, 6144]
blk.73.attn_q_a.weight
Q4_K
Q4_K
[6144, 2048]
blk.73.attn_q_a_norm.weight
F32
F32
[2048]
blk.73.attn_q_b.weight
Q6_K
Q6_K
[2048, 16384]
blk.73.attn_v_b.weight
Q8_0
Q8_0
[512, 256, 64]
blk.73.exp_probs_b.bias
F32
F32
[256]
blk.73.ffn_down_exps.weight
Q5_K
Q5_K
[2048, 6144, 256]
blk.73.ffn_down_shexp.weight
Q6_K
Q6_K
[2048, 6144]
blk.73.ffn_gate_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.73.ffn_gate_inp.weight
F32
F32
[6144, 256]
blk.73.ffn_gate_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.73.ffn_norm.weight
F32
F32
[6144]
blk.73.ffn_up_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.73.ffn_up_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.73.indexer.attn_k.weight
Q4_K
Q4_K
[6144, 128]
blk.73.indexer.attn_q_b.weight
Q4_K
Q4_K
[2048, 4096]
blk.73.indexer.k_norm.bias
F32
F32
[128]
blk.73.indexer.k_norm.weight
F32
F32
[128]
blk.73.indexer.proj.weight
Q4_K
Q4_K
[6144, 32]
blk.74
blk.74.attn_k_b.weight
Q8_0
Q8_0
[192, 512, 64]
blk.74.attn_kv_a_mqa.weight
Q8_0
Q8_0
[6144, 576]
blk.74.attn_kv_a_norm.weight
F32
F32
[512]
blk.74.attn_norm.weight
F32
F32
[6144]
blk.74.attn_output.weight
Q4_K
Q4_K
[16384, 6144]
blk.74.attn_q_a.weight
Q4_K
Q4_K
[6144, 2048]
blk.74.attn_q_a_norm.weight
F32
F32
[2048]
blk.74.attn_q_b.weight
Q4_K
Q4_K
[2048, 16384]
blk.74.attn_v_b.weight
Q8_0
Q8_0
[512, 256, 64]
blk.74.exp_probs_b.bias
F32
F32
[256]
blk.74.ffn_down_exps.weight
Q5_K
Q5_K
[2048, 6144, 256]
blk.74.ffn_down_shexp.weight
Q6_K
Q6_K
[2048, 6144]
blk.74.ffn_gate_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.74.ffn_gate_inp.weight
F32
F32
[6144, 256]
blk.74.ffn_gate_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.74.ffn_norm.weight
F32
F32
[6144]
blk.74.ffn_up_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.74.ffn_up_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.74.indexer.attn_k.weight
Q4_K
Q4_K
[6144, 128]
blk.74.indexer.attn_q_b.weight
Q4_K
Q4_K
[2048, 4096]
blk.74.indexer.k_norm.bias
F32
F32
[128]
blk.74.indexer.k_norm.weight
F32
F32
[128]
blk.74.indexer.proj.weight
Q4_K
Q4_K
[6144, 32]
blk.75
blk.75.attn_k_b.weight
Q8_0
Q8_0
[192, 512, 64]
blk.75.attn_kv_a_mqa.weight
Q8_0
Q8_0
[6144, 576]
blk.75.attn_kv_a_norm.weight
F32
F32
[512]
blk.75.attn_norm.weight
F32
F32
[6144]
blk.75.attn_output.weight
Q5_K
Q5_K
[16384, 6144]
blk.75.attn_q_a.weight
Q5_K
Q5_K
[6144, 2048]
blk.75.attn_q_a_norm.weight
F32
F32
[2048]
blk.75.attn_q_b.weight
Q5_K
Q5_K
[2048, 16384]
blk.75.attn_v_b.weight
Q8_0
Q8_0
[512, 256, 64]
blk.75.exp_probs_b.bias
F32
F32
[256]
blk.75.ffn_down_exps.weight
Q5_K
Q5_K
[2048, 6144, 256]
blk.75.ffn_down_shexp.weight
Q8_0
Q8_0
[2048, 6144]
blk.75.ffn_gate_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.75.ffn_gate_inp.weight
F32
F32
[6144, 256]
blk.75.ffn_gate_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.75.ffn_norm.weight
F32
F32
[6144]
blk.75.ffn_up_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.75.ffn_up_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.75.indexer.attn_k.weight
Q4_K
Q4_K
[6144, 128]
blk.75.indexer.attn_q_b.weight
Q4_K
Q4_K
[2048, 4096]
blk.75.indexer.k_norm.bias
F32
F32
[128]
blk.75.indexer.k_norm.weight
F32
F32
[128]
blk.75.indexer.proj.weight
Q4_K
Q4_K
[6144, 32]
blk.76
blk.76.attn_k_b.weight
Q8_0
Q8_0
[192, 512, 64]
blk.76.attn_kv_a_mqa.weight
Q8_0
Q8_0
[6144, 576]
blk.76.attn_kv_a_norm.weight
F32
F32
[512]
blk.76.attn_norm.weight
F32
F32
[6144]
blk.76.attn_output.weight
Q4_K
Q4_K
[16384, 6144]
blk.76.attn_q_a.weight
Q4_K
Q4_K
[6144, 2048]
blk.76.attn_q_a_norm.weight
F32
F32
[2048]
blk.76.attn_q_b.weight
Q6_K
Q6_K
[2048, 16384]
blk.76.attn_v_b.weight
Q8_0
Q8_0
[512, 256, 64]
blk.76.exp_probs_b.bias
F32
F32
[256]
blk.76.ffn_down_exps.weight
Q5_K
Q5_K
[2048, 6144, 256]
blk.76.ffn_down_shexp.weight
Q6_K
Q6_K
[2048, 6144]
blk.76.ffn_gate_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.76.ffn_gate_inp.weight
F32
F32
[6144, 256]
blk.76.ffn_gate_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.76.ffn_norm.weight
F32
F32
[6144]
blk.76.ffn_up_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.76.ffn_up_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.76.indexer.attn_k.weight
Q4_K
Q4_K
[6144, 128]
blk.76.indexer.attn_q_b.weight
Q4_K
Q4_K
[2048, 4096]
blk.76.indexer.k_norm.bias
F32
F32
[128]
blk.76.indexer.k_norm.weight
F32
F32
[128]
blk.76.indexer.proj.weight
Q4_K
Q4_K
[6144, 32]
blk.77
blk.77.attn_k_b.weight
Q8_0
Q8_0
[192, 512, 64]
blk.77.attn_kv_a_mqa.weight
Q8_0
Q8_0
[6144, 576]
blk.77.attn_kv_a_norm.weight
F32
F32
[512]
blk.77.attn_norm.weight
F32
F32
[6144]
blk.77.attn_output.weight
Q5_K
Q5_K
[16384, 6144]
blk.77.attn_q_a.weight
Q5_K
Q5_K
[6144, 2048]
blk.77.attn_q_a_norm.weight
F32
F32
[2048]
blk.77.attn_q_b.weight
Q8_0
Q8_0
[2048, 16384]
blk.77.attn_v_b.weight
Q8_0
Q8_0
[512, 256, 64]
blk.77.exp_probs_b.bias
F32
F32
[256]
blk.77.ffn_down_exps.weight
Q5_K
Q5_K
[2048, 6144, 256]
blk.77.ffn_down_shexp.weight
Q8_0
Q8_0
[2048, 6144]
blk.77.ffn_gate_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.77.ffn_gate_inp.weight
F32
F32
[6144, 256]
blk.77.ffn_gate_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.77.ffn_norm.weight
F32
F32
[6144]
blk.77.ffn_up_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.77.ffn_up_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.77.indexer.attn_k.weight
Q4_K
Q4_K
[6144, 128]
blk.77.indexer.attn_q_b.weight
Q4_K
Q4_K
[2048, 4096]
blk.77.indexer.k_norm.bias
F32
F32
[128]
blk.77.indexer.k_norm.weight
F32
F32
[128]
blk.77.indexer.proj.weight
Q4_K
Q4_K
[6144, 32]
blk.78
blk.78.attn_k_b.weight
Q8_0
Q8_0
[192, 512, 64]
blk.78.attn_kv_a_mqa.weight
Q8_0
Q8_0
[6144, 576]
blk.78.attn_kv_a_norm.weight
F32
F32
[512]
blk.78.attn_norm.weight
F32
F32
[6144]
blk.78.attn_output.weight
Q4_K
Q4_K
[16384, 6144]
blk.78.attn_q_a.weight
Q4_K
Q4_K
[6144, 2048]
blk.78.attn_q_a_norm.weight
F32
F32
[2048]
blk.78.attn_q_b.weight
Q8_0
Q8_0
[2048, 16384]
blk.78.attn_v_b.weight
Q8_0
Q8_0
[512, 256, 64]
blk.78.exp_probs_b.bias
F32
F32
[256]
blk.78.ffn_down_exps.weight
Q4_K
Q4_K
[2048, 6144, 256]
blk.78.ffn_down_shexp.weight
Q6_K
Q6_K
[2048, 6144]
blk.78.ffn_gate_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.78.ffn_gate_inp.weight
F32
F32
[6144, 256]
blk.78.ffn_gate_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.78.ffn_norm.weight
F32
F32
[6144]
blk.78.ffn_up_exps.weight
Q4_K
Q4_K
[6144, 2048, 256]
blk.78.ffn_up_shexp.weight
Q4_K
Q4_K
[6144, 2048]
blk.78.indexer.attn_k.weight
Q4_K
Q4_K
[6144, 128]
blk.78.indexer.attn_q_b.weight
Q4_K
Q4_K
[2048, 4096]
blk.78.indexer.k_norm.bias
F32
F32
[128]
blk.78.indexer.k_norm.weight
F32
F32
[128]
blk.78.indexer.proj.weight
Q4_K
Q4_K
[6144, 32]
blk.78.nextn.eh_proj.weight
Q4_K
Q4_K
[12288, 6144]
blk.78.nextn.enorm.weight
F32
F32
[6144]
blk.78.nextn.hnorm.weight
F32
F32
[6144]
blk.78.nextn.shared_head_norm.weight
F32
F32
[6144]
output.weight
Q6_K
Q6_K
[6144, 154880]
output_norm.weight
F32
F32
[6144]