GitHub
185.7K
Download
Models
Docs
Pricing
Sign in
Download
Models
Docs
Pricing
Sign in
frob
/
deepseek-v4.1-flash
:552b-a8b-1017-ud-iq1_M
205
Downloads
Updated
5 days ago
1017
1017
Cancel
vision
552b
deepseek-v4.1-flash:552b-a8b-1017-ud-iq1_M
...
/
model
8a406949fac9 ยท 50GB
Metadata
split.count
5
5
split.no
1
1
split.tensors.count
1043
1043
Tensor
Name
Type
Shape
token_embd.weight
Q8_0
Q8_0
[5120, 129280]
blk.0
blk.0.attn_kv.weight
Q6_K
Q6_K
[5120, 512]
blk.0.attn_kv_a_norm.weight
F32
F32
[512]
blk.0.attn_norm.weight
F32
F32
[5120]
blk.0.attn_output_a.weight
Q6_K
Q6_K
[4096, 8192]
blk.0.attn_output_b.weight
Q6_K
Q6_K
[8192, 5120]
blk.0.attn_q_a.weight
Q6_K
Q6_K
[5120, 1280]
blk.0.attn_q_a_norm.weight
F32
F32
[1280]
blk.0.attn_q_b.weight
Q6_K
Q6_K
[1280, 32768]
blk.0.attn_sinks.weight
F32
F32
[64]
blk.0.exp_probs_b.bias
F32
F32
[384]
blk.0.exp_probs_b_vl.bias
F32
F32
[384]
blk.0.ffn_down_exps.weight
IQ1_S
IQ1_S
[2304, 5120, 384]
blk.0.ffn_down_shexp.weight
Q8_0
Q8_0
[2304, 5120]
blk.0.ffn_gate_exps.weight
IQ1_S
IQ1_S
[5120, 2304, 384]
blk.0.ffn_gate_inp.weight
F32
F32
[5120, 384]
blk.0.ffn_gate_shexp.weight
Q8_0
Q8_0
[5120, 2304]
blk.0.ffn_norm.weight
F32
F32
[5120]
blk.0.ffn_up_exps.weight
IQ1_S
IQ1_S
[5120, 2304, 384]
blk.0.ffn_up_shexp.weight
Q8_0
Q8_0
[5120, 2304]
blk.0.hc_attn_base.weight
F32
F32
[24]
blk.0.hc_attn_fn.weight
F32
F32
[20480, 24]
blk.0.hc_attn_scale.weight
F32
F32
[3]
blk.0.hc_ffn_base.weight
F32
F32
[24]
blk.0.hc_ffn_fn.weight
F32
F32
[20480, 24]
blk.0.hc_ffn_scale.weight
F32
F32
[3]
blk.1
blk.1.attn_kv.weight
Q6_K
Q6_K
[5120, 512]
blk.1.attn_kv_a_norm.weight
F32
F32
[512]
blk.1.attn_norm.weight
F32
F32
[5120]
blk.1.attn_output_a.weight
Q6_K
Q6_K
[4096, 8192]
blk.1.attn_output_b.weight
Q6_K
Q6_K
[8192, 5120]
blk.1.attn_q_a.weight
Q6_K
Q6_K
[5120, 1280]
blk.1.attn_q_a_norm.weight
F32
F32
[1280]
blk.1.attn_q_b.weight
Q6_K
Q6_K
[1280, 32768]
blk.1.attn_sinks.weight
F32
F32
[64]
blk.1.engram_gate.weight
F32
F32
[5120, 4]
blk.1.engram_kv.weight
Q6_K
Q6_K
[6144, 25600]
blk.1.exp_probs_b.bias
F32
F32
[384]
blk.1.exp_probs_b_vl.bias
F32
F32
[384]
blk.1.ffn_down_exps.weight
IQ1_S
IQ1_S
[2304, 5120, 384]
blk.1.ffn_down_shexp.weight
Q8_0
Q8_0
[2304, 5120]
blk.1.ffn_gate_exps.weight
IQ1_S
IQ1_S
[5120, 2304, 384]
blk.1.ffn_gate_inp.weight
F32
F32
[5120, 384]
blk.1.ffn_gate_shexp.weight
Q8_0
Q8_0
[5120, 2304]
blk.1.ffn_norm.weight
F32
F32
[5120]
blk.1.ffn_up_exps.weight
IQ1_S
IQ1_S
[5120, 2304, 384]
blk.1.ffn_up_shexp.weight
Q8_0
Q8_0
[5120, 2304]
blk.1.hc_attn_base.weight
F32
F32
[24]
blk.1.hc_attn_fn.weight
F32
F32
[20480, 24]
blk.1.hc_attn_scale.weight
F32
F32
[3]
blk.1.hc_ffn_base.weight
F32
F32
[24]
blk.1.hc_ffn_fn.weight
F32
F32
[20480, 24]
blk.1.hc_ffn_scale.weight
F32
F32
[3]
blk.2
blk.2.attn_compressor_gate.weight
BF16
BF16
[5120, 512]
blk.2.attn_compressor_kv.weight
BF16
BF16
[5120, 512]
blk.2.attn_compressor_norm.weight
F32
F32
[512]
blk.2.attn_kv.weight
Q6_K
Q6_K
[5120, 512]
blk.2.attn_kv_a_norm.weight
F32
F32
[512]
blk.2.attn_norm.weight
F32
F32
[5120]
blk.2.attn_output_a.weight
Q6_K
Q6_K
[4096, 8192]
blk.2.attn_output_b.weight
Q6_K
Q6_K
[8192, 5120]
blk.2.attn_q_a.weight
Q6_K
Q6_K
[5120, 1280]
blk.2.attn_q_a_norm.weight
F32
F32
[1280]
blk.2.attn_q_b.weight
Q6_K
Q6_K
[1280, 32768]
blk.2.attn_sinks.weight
F32
F32
[64]
blk.2.exp_probs_b.bias
F32
F32
[384]
blk.2.exp_probs_b_vl.bias
F32
F32
[384]
blk.2.ffn_down_exps.weight
IQ1_S
IQ1_S
[2304, 5120, 384]
blk.2.ffn_down_shexp.weight
Q8_0
Q8_0
[2304, 5120]
blk.2.ffn_gate_exps.weight
IQ1_S
IQ1_S
[5120, 2304, 384]
blk.2.ffn_gate_inp.weight
F32
F32
[5120, 384]
blk.2.ffn_gate_shexp.weight
Q8_0
Q8_0
[5120, 2304]
blk.2.ffn_norm.weight
F32
F32
[5120]
blk.2.ffn_up_exps.weight
IQ1_S
IQ1_S
[5120, 2304, 384]
blk.2.ffn_up_shexp.weight
Q8_0
Q8_0
[5120, 2304]
blk.2.hc_attn_base.weight
F32
F32
[24]
blk.2.hc_attn_fn.weight
F32
F32
[20480, 24]
blk.2.hc_attn_scale.weight
F32
F32
[3]
blk.2.hc_ffn_base.weight
F32
F32
[24]
blk.2.hc_ffn_fn.weight
F32
F32
[20480, 24]
blk.2.hc_ffn_scale.weight
F32
F32
[3]
blk.2.indexer.attn_k.weight
BF16
BF16
[512, 128]
blk.2.indexer.attn_q_b.weight
BF16
BF16
[1280, 4096]
blk.2.indexer.k_norm.weight
F32
F32
[128]
blk.2.indexer.proj.weight
BF16
BF16
[5120, 32]
blk.3
blk.3.attn_kv.weight
Q6_K
Q6_K
[5120, 512]
blk.3.attn_kv_a_norm.weight
F32
F32
[512]
blk.3.attn_norm.weight
F32
F32
[5120]
blk.3.attn_output_a.weight
Q6_K
Q6_K
[4096, 8192]
blk.3.attn_output_b.weight
Q6_K
Q6_K
[8192, 5120]
blk.3.attn_q_a.weight
Q6_K
Q6_K
[5120, 1280]
blk.3.attn_q_a_norm.weight
F32
F32
[1280]
blk.3.attn_q_b.weight
Q6_K
Q6_K
[1280, 32768]
blk.3.attn_sinks.weight
F32
F32
[64]
blk.3.exp_probs_b.bias
F32
F32
[384]
blk.3.exp_probs_b_vl.bias
F32
F32
[384]
blk.3.ffn_down_exps.weight
IQ1_S
IQ1_S
[2304, 5120, 384]
blk.3.ffn_down_shexp.weight
Q8_0
Q8_0
[2304, 5120]
blk.3.ffn_gate_exps.weight
IQ1_S
IQ1_S
[5120, 2304, 384]
blk.3.ffn_gate_inp.weight
F32
F32
[5120, 384]
blk.3.ffn_gate_shexp.weight
Q8_0
Q8_0
[5120, 2304]
blk.3.ffn_norm.weight
F32
F32
[5120]
blk.3.ffn_up_exps.weight
IQ1_S
IQ1_S
[5120, 2304, 384]
blk.3.ffn_up_shexp.weight
Q8_0
Q8_0
[5120, 2304]
blk.3.hc_attn_base.weight
F32
F32
[24]
blk.3.hc_attn_fn.weight
F32
F32
[20480, 24]
blk.3.hc_attn_scale.weight
F32
F32
[3]
blk.3.hc_ffn_base.weight
F32
F32
[24]
blk.3.hc_ffn_fn.weight
F32
F32
[20480, 24]
blk.3.hc_ffn_scale.weight
F32
F32
[3]
blk.4
blk.4.attn_kv.weight
Q6_K
Q6_K
[5120, 512]
blk.4.attn_kv_a_norm.weight
F32
F32
[512]
blk.4.attn_norm.weight
F32
F32
[5120]
blk.4.attn_output_a.weight
Q6_K
Q6_K
[4096, 8192]
blk.4.attn_output_b.weight
Q6_K
Q6_K
[8192, 5120]
blk.4.attn_q_a.weight
Q6_K
Q6_K
[5120, 1280]
blk.4.attn_q_a_norm.weight
F32
F32
[1280]
blk.4.attn_q_b.weight
Q6_K
Q6_K
[1280, 32768]
blk.4.attn_sinks.weight
F32
F32
[64]
blk.4.exp_probs_b.bias
F32
F32
[384]
blk.4.exp_probs_b_vl.bias
F32
F32
[384]
blk.4.ffn_down_exps.weight
IQ1_S
IQ1_S
[2304, 5120, 384]
blk.4.ffn_down_shexp.weight
Q8_0
Q8_0
[2304, 5120]
blk.4.ffn_gate_exps.weight
IQ1_S
IQ1_S
[5120, 2304, 384]
blk.4.ffn_gate_inp.weight
F32
F32
[5120, 384]
blk.4.ffn_gate_shexp.weight
Q8_0
Q8_0
[5120, 2304]
blk.4.ffn_norm.weight
F32
F32
[5120]
blk.4.ffn_up_exps.weight
IQ1_S
IQ1_S
[5120, 2304, 384]
blk.4.ffn_up_shexp.weight
Q8_0
Q8_0
[5120, 2304]
blk.4.hc_attn_base.weight
F32
F32
[24]
blk.4.hc_attn_fn.weight
F32
F32
[20480, 24]
blk.4.hc_attn_scale.weight
F32
F32
[3]
blk.4.hc_ffn_base.weight
F32
F32
[24]
blk.4.hc_ffn_fn.weight
F32
F32
[20480, 24]
blk.4.hc_ffn_scale.weight
F32
F32
[3]
blk.5
blk.5.attn_kv.weight
Q6_K
Q6_K
[5120, 512]
blk.5.attn_kv_a_norm.weight
F32
F32
[512]
blk.5.attn_norm.weight
F32
F32
[5120]
blk.5.attn_output_a.weight
Q6_K
Q6_K
[4096, 8192]
blk.5.attn_output_b.weight
Q6_K
Q6_K
[8192, 5120]
blk.5.attn_q_a.weight
Q6_K
Q6_K
[5120, 1280]
blk.5.attn_q_a_norm.weight
F32
F32
[1280]
blk.5.attn_q_b.weight
Q6_K
Q6_K
[1280, 32768]
blk.5.attn_sinks.weight
F32
F32
[64]
blk.5.exp_probs_b.bias
F32
F32
[384]
blk.5.exp_probs_b_vl.bias
F32
F32
[384]
blk.5.ffn_down_exps.weight
IQ2_XXS
IQ2_XXS
[2304, 5120, 384]
blk.5.ffn_down_shexp.weight
Q8_0
Q8_0
[2304, 5120]
blk.5.ffn_gate_exps.weight
IQ1_S
IQ1_S
[5120, 2304, 384]
blk.5.ffn_gate_inp.weight
F32
F32
[5120, 384]
blk.5.ffn_gate_shexp.weight
Q8_0
Q8_0
[5120, 2304]
blk.5.ffn_norm.weight
F32
F32
[5120]
blk.5.ffn_up_exps.weight
IQ1_S
IQ1_S
[5120, 2304, 384]
blk.5.ffn_up_shexp.weight
Q8_0
Q8_0
[5120, 2304]
blk.5.hc_attn_base.weight
F32
F32
[24]
blk.5.hc_attn_fn.weight
F32
F32
[20480, 24]
blk.5.hc_attn_scale.weight
F32
F32
[3]
blk.5.hc_ffn_base.weight
F32
F32
[24]
blk.5.hc_ffn_fn.weight
F32
F32
[20480, 24]
blk.5.hc_ffn_scale.weight
F32
F32
[3]
blk.6
blk.6.attn_kv.weight
Q6_K
Q6_K
[5120, 512]
blk.6.attn_kv_a_norm.weight
F32
F32
[512]
blk.6.attn_norm.weight
F32
F32
[5120]
blk.6.attn_output_a.weight
Q6_K
Q6_K
[4096, 8192]
blk.6.attn_output_b.weight
Q6_K
Q6_K
[8192, 5120]
blk.6.attn_q_a.weight
Q6_K
Q6_K
[5120, 1280]
blk.6.attn_q_a_norm.weight
F32
F32
[1280]
blk.6.attn_q_b.weight
Q6_K
Q6_K
[1280, 32768]
blk.6.attn_sinks.weight
F32
F32
[64]
blk.6.exp_probs_b.bias
F32
F32
[384]
blk.6.exp_probs_b_vl.bias
F32
F32
[384]
blk.6.ffn_down_exps.weight
IQ2_XXS
IQ2_XXS
[2304, 5120, 384]
blk.6.ffn_down_shexp.weight
Q8_0
Q8_0
[2304, 5120]
blk.6.ffn_gate_exps.weight
IQ1_S
IQ1_S
[5120, 2304, 384]
blk.6.ffn_gate_inp.weight
F32
F32
[5120, 384]
blk.6.ffn_gate_shexp.weight
Q8_0
Q8_0
[5120, 2304]
blk.6.ffn_norm.weight
F32
F32
[5120]
blk.6.ffn_up_exps.weight
IQ1_S
IQ1_S
[5120, 2304, 384]
blk.6.ffn_up_shexp.weight
Q8_0
Q8_0
[5120, 2304]
blk.6.hc_attn_base.weight
F32
F32
[24]
blk.6.hc_attn_fn.weight
F32
F32
[20480, 24]
blk.6.hc_attn_scale.weight
F32
F32
[3]
blk.6.hc_ffn_base.weight
F32
F32
[24]
blk.6.hc_ffn_fn.weight
F32
F32
[20480, 24]
blk.6.hc_ffn_scale.weight
F32
F32
[3]
blk.7
blk.7.attn_kv.weight
Q6_K
Q6_K
[5120, 512]
blk.7.attn_kv_a_norm.weight
F32
F32
[512]
blk.7.attn_norm.weight
F32
F32
[5120]
blk.7.attn_output_a.weight
Q6_K
Q6_K
[4096, 8192]
blk.7.attn_output_b.weight
Q6_K
Q6_K
[8192, 5120]
blk.7.attn_q_a.weight
Q6_K
Q6_K
[5120, 1280]
blk.7.attn_q_a_norm.weight
F32
F32
[1280]
blk.7.attn_q_b.weight
Q6_K
Q6_K
[1280, 32768]
blk.7.attn_sinks.weight
F32
F32
[64]
blk.7.exp_probs_b.bias
F32
F32
[384]
blk.7.exp_probs_b_vl.bias
F32
F32
[384]
blk.7.ffn_down_exps.weight
IQ2_XXS
IQ2_XXS
[2304, 5120, 384]
blk.7.ffn_down_shexp.weight
Q8_0
Q8_0
[2304, 5120]
blk.7.ffn_gate_exps.weight
IQ1_S
IQ1_S
[5120, 2304, 384]
blk.7.ffn_gate_inp.weight
F32
F32
[5120, 384]
blk.7.ffn_gate_shexp.weight
Q8_0
Q8_0
[5120, 2304]
blk.7.ffn_norm.weight
F32
F32
[5120]
blk.7.ffn_up_exps.weight
IQ1_S
IQ1_S
[5120, 2304, 384]
blk.7.ffn_up_shexp.weight
Q8_0
Q8_0
[5120, 2304]
blk.7.hc_attn_base.weight
F32
F32
[24]
blk.7.hc_attn_fn.weight
F32
F32
[20480, 24]
blk.7.hc_attn_scale.weight
F32
F32
[3]
blk.7.hc_ffn_base.weight
F32
F32
[24]
blk.7.hc_ffn_fn.weight
F32
F32
[20480, 24]
blk.7.hc_ffn_scale.weight
F32
F32
[3]
blk.8
blk.8.attn_compressor_gate.weight
BF16
BF16
[5120, 512]
blk.8.attn_compressor_kv.weight
BF16
BF16
[5120, 512]
blk.8.attn_compressor_norm.weight
F32
F32
[512]
blk.8.attn_kv.weight
Q6_K
Q6_K
[5120, 512]
blk.8.attn_kv_a_norm.weight
F32
F32
[512]
blk.8.attn_norm.weight
F32
F32
[5120]
blk.8.attn_output_a.weight
Q6_K
Q6_K
[4096, 8192]
blk.8.attn_output_b.weight
Q6_K
Q6_K
[8192, 5120]
blk.8.attn_q_a.weight
Q6_K
Q6_K
[5120, 1280]
blk.8.attn_q_a_norm.weight
F32
F32
[1280]
blk.8.attn_q_b.weight
Q6_K
Q6_K
[1280, 32768]
blk.8.attn_sinks.weight
F32
F32
[64]
blk.8.exp_probs_b.bias
F32
F32
[384]
blk.8.exp_probs_b_vl.bias
F32
F32
[384]
blk.8.ffn_down_exps.weight
IQ2_XXS
IQ2_XXS
[2304, 5120, 384]
blk.8.ffn_down_shexp.weight
Q8_0
Q8_0
[2304, 5120]
blk.8.ffn_gate_exps.weight
IQ1_S
IQ1_S
[5120, 2304, 384]
blk.8.ffn_gate_inp.weight
F32
F32
[5120, 384]
blk.8.ffn_gate_shexp.weight
Q8_0
Q8_0
[5120, 2304]
blk.8.ffn_norm.weight
F32
F32
[5120]
blk.8.ffn_up_exps.weight
IQ1_S
IQ1_S
[5120, 2304, 384]
blk.8.ffn_up_shexp.weight
Q8_0
Q8_0
[5120, 2304]
blk.8.hc_attn_base.weight
F32
F32
[24]
blk.8.hc_attn_fn.weight
F32
F32
[20480, 24]
blk.8.hc_attn_scale.weight
F32
F32
[3]
blk.8.hc_ffn_base.weight
F32
F32
[24]
blk.8.hc_ffn_fn.weight
F32
F32
[20480, 24]
blk.8.hc_ffn_scale.weight
F32
F32
[3]
blk.8.indexer.attn_k.weight
BF16
BF16
[512, 128]
blk.8.indexer.attn_q_b.weight
BF16
BF16
[1280, 4096]
blk.8.indexer.k_norm.weight
F32
F32
[128]
blk.8.indexer.proj.weight
BF16
BF16
[5120, 32]
blk.9
blk.9.attn_kv.weight
Q6_K
Q6_K
[5120, 512]
blk.9.attn_kv_a_norm.weight
F32
F32
[512]
blk.9.attn_norm.weight
F32
F32
[5120]
blk.9.attn_output_a.weight
Q6_K
Q6_K
[4096, 8192]
blk.9.attn_output_b.weight
Q6_K
Q6_K
[8192, 5120]
blk.9.attn_q_a.weight
Q6_K
Q6_K
[5120, 1280]
blk.9.attn_q_a_norm.weight
F32
F32
[1280]
blk.9.attn_q_b.weight
Q6_K
Q6_K
[1280, 32768]
blk.9.attn_sinks.weight
F32
F32
[64]
blk.9.exp_probs_b.bias
F32
F32
[384]
blk.9.exp_probs_b_vl.bias
F32
F32
[384]
blk.9.ffn_down_exps.weight
IQ2_XXS
IQ2_XXS
[2304, 5120, 384]
blk.9.ffn_down_shexp.weight
Q8_0
Q8_0
[2304, 5120]
blk.9.ffn_gate_exps.weight
IQ1_S
IQ1_S
[5120, 2304, 384]
blk.9.ffn_gate_inp.weight
F32
F32
[5120, 384]
blk.9.ffn_gate_shexp.weight
Q8_0
Q8_0
[5120, 2304]
blk.9.ffn_norm.weight
F32
F32
[5120]
blk.9.ffn_up_exps.weight
IQ1_S
IQ1_S
[5120, 2304, 384]
blk.9.ffn_up_shexp.weight
Q8_0
Q8_0
[5120, 2304]
blk.9.hc_attn_base.weight
F32
F32
[24]
blk.9.hc_attn_fn.weight
F32
F32
[20480, 24]
blk.9.hc_attn_scale.weight
F32
F32
[3]
blk.9.hc_ffn_base.weight
F32
F32
[24]
blk.9.hc_ffn_fn.weight
F32
F32
[20480, 24]
blk.9.hc_ffn_scale.weight
F32
F32
[3]
blk.10
blk.10.attn_kv.weight
Q6_K
Q6_K
[5120, 512]
blk.10.attn_kv_a_norm.weight
F32
F32
[512]
blk.10.attn_norm.weight
F32
F32
[5120]
blk.10.attn_output_a.weight
Q6_K
Q6_K
[4096, 8192]
blk.10.attn_output_b.weight
Q6_K
Q6_K
[8192, 5120]
blk.10.attn_q_a.weight
Q6_K
Q6_K
[5120, 1280]
blk.10.attn_q_a_norm.weight
F32
F32
[1280]
blk.10.attn_q_b.weight
Q6_K
Q6_K
[1280, 32768]
blk.10.attn_sinks.weight
F32
F32
[64]
blk.10.exp_probs_b.bias
F32
F32
[384]
blk.10.exp_probs_b_vl.bias
F32
F32
[384]
blk.10.ffn_down_exps.weight
IQ2_XXS
IQ2_XXS
[2304, 5120, 384]
blk.10.ffn_down_shexp.weight
Q8_0
Q8_0
[2304, 5120]
blk.10.ffn_gate_exps.weight
IQ1_S
IQ1_S
[5120, 2304, 384]
blk.10.ffn_gate_inp.weight
F32
F32
[5120, 384]
blk.10.ffn_gate_shexp.weight
Q8_0
Q8_0
[5120, 2304]
blk.10.ffn_norm.weight
F32
F32
[5120]
blk.10.ffn_up_exps.weight
IQ1_S
IQ1_S
[5120, 2304, 384]
blk.10.ffn_up_shexp.weight
Q8_0
Q8_0
[5120, 2304]
blk.10.hc_attn_base.weight
F32
F32
[24]
blk.10.hc_attn_fn.weight
F32
F32
[20480, 24]
blk.10.hc_attn_scale.weight
F32
F32
[3]
blk.10.hc_ffn_base.weight
F32
F32
[24]
blk.10.hc_ffn_fn.weight
F32
F32
[20480, 24]
blk.10.hc_ffn_scale.weight
F32
F32
[3]
blk.11
blk.11.attn_kv.weight
Q6_K
Q6_K
[5120, 512]
blk.11.attn_kv_a_norm.weight
F32
F32
[512]
blk.11.attn_norm.weight
F32
F32
[5120]
blk.11.attn_output_a.weight
Q6_K
Q6_K
[4096, 8192]
blk.11.attn_output_b.weight
Q6_K
Q6_K
[8192, 5120]
blk.11.attn_q_a.weight
Q6_K
Q6_K
[5120, 1280]
blk.11.attn_q_a_norm.weight
F32
F32
[1280]
blk.11.attn_q_b.weight
Q6_K
Q6_K
[1280, 32768]
blk.11.attn_sinks.weight
F32
F32
[64]
blk.11.exp_probs_b.bias
F32
F32
[384]
blk.11.exp_probs_b_vl.bias
F32
F32
[384]
blk.11.ffn_down_exps.weight
IQ2_XXS
IQ2_XXS
[2304, 5120, 384]
blk.11.ffn_down_shexp.weight
Q8_0
Q8_0
[2304, 5120]
blk.11.ffn_gate_exps.weight
IQ1_S
IQ1_S
[5120, 2304, 384]
blk.11.ffn_gate_inp.weight
F32
F32
[5120, 384]
blk.11.ffn_gate_shexp.weight
Q8_0
Q8_0
[5120, 2304]
blk.11.ffn_norm.weight
F32
F32
[5120]
blk.11.ffn_up_exps.weight
IQ1_S
IQ1_S
[5120, 2304, 384]
blk.11.ffn_up_shexp.weight
Q8_0
Q8_0
[5120, 2304]
blk.11.hc_attn_base.weight
F32
F32
[24]
blk.11.hc_attn_fn.weight
F32
F32
[20480, 24]
blk.11.hc_attn_scale.weight
F32
F32
[3]
blk.11.hc_ffn_base.weight
F32
F32
[24]
blk.11.hc_ffn_fn.weight
F32
F32
[20480, 24]
blk.11.hc_ffn_scale.weight
F32
F32
[3]
blk.12
blk.12.attn_kv.weight
Q6_K
Q6_K
[5120, 512]
blk.12.attn_kv_a_norm.weight
F32
F32
[512]
blk.12.attn_norm.weight
F32
F32
[5120]
blk.12.attn_output_a.weight
Q6_K
Q6_K
[4096, 8192]
blk.12.attn_output_b.weight
Q6_K
Q6_K
[8192, 5120]
blk.12.attn_q_a.weight
Q6_K
Q6_K
[5120, 1280]
blk.12.attn_q_a_norm.weight
F32
F32
[1280]
blk.12.attn_q_b.weight
Q6_K
Q6_K
[1280, 32768]
blk.12.attn_sinks.weight
F32
F32
[64]
blk.12.exp_probs_b.bias
F32
F32
[384]
blk.12.exp_probs_b_vl.bias
F32
F32
[384]
blk.12.ffn_down_exps.weight
IQ2_XXS
IQ2_XXS
[2304, 5120, 384]
blk.12.ffn_down_shexp.weight
Q8_0
Q8_0
[2304, 5120]
blk.12.ffn_gate_exps.weight
IQ1_S
IQ1_S
[5120, 2304, 384]
blk.12.ffn_gate_inp.weight
F32
F32
[5120, 384]
blk.12.ffn_gate_shexp.weight
Q8_0
Q8_0
[5120, 2304]
blk.12.ffn_norm.weight
F32
F32
[5120]
blk.12.ffn_up_exps.weight
IQ1_S
IQ1_S
[5120, 2304, 384]
blk.12.ffn_up_shexp.weight
Q8_0
Q8_0
[5120, 2304]
blk.12.hc_attn_base.weight
F32
F32
[24]
blk.12.hc_attn_fn.weight
F32
F32
[20480, 24]
blk.12.hc_attn_scale.weight
F32
F32
[3]
blk.12.hc_ffn_base.weight
F32
F32
[24]
blk.12.hc_ffn_fn.weight
F32
F32
[20480, 24]
blk.12.hc_ffn_scale.weight
F32
F32
[3]
blk.13
blk.13.attn_kv.weight
Q6_K
Q6_K
[5120, 512]
blk.13.attn_kv_a_norm.weight
F32
F32
[512]
blk.13.attn_norm.weight
F32
F32
[5120]
blk.13.attn_output_a.weight
Q6_K
Q6_K
[4096, 8192]
blk.13.attn_output_b.weight
Q6_K
Q6_K
[8192, 5120]
blk.13.attn_q_a.weight
Q6_K
Q6_K
[5120, 1280]
blk.13.attn_q_a_norm.weight
F32
F32
[1280]
blk.13.attn_q_b.weight
Q6_K
Q6_K
[1280, 32768]
blk.13.attn_sinks.weight
F32
F32
[64]
blk.13.exp_probs_b.bias
F32
F32
[384]
blk.13.exp_probs_b_vl.bias
F32
F32
[384]
blk.13.ffn_down_exps.weight
IQ2_XXS
IQ2_XXS
[2304, 5120, 384]
blk.13.ffn_down_shexp.weight
Q8_0
Q8_0
[2304, 5120]
blk.13.ffn_gate_exps.weight
IQ1_S
IQ1_S
[5120, 2304, 384]
blk.13.ffn_gate_inp.weight
F32
F32
[5120, 384]
blk.13.ffn_gate_shexp.weight
Q8_0
Q8_0
[5120, 2304]
blk.13.ffn_norm.weight
F32
F32
[5120]
blk.13.ffn_up_exps.weight
IQ1_S
IQ1_S
[5120, 2304, 384]
blk.13.ffn_up_shexp.weight
Q8_0
Q8_0
[5120, 2304]
blk.13.hc_attn_base.weight
F32
F32
[24]
blk.13.hc_attn_fn.weight
F32
F32
[20480, 24]
blk.13.hc_attn_scale.weight
F32
F32
[3]
blk.13.hc_ffn_base.weight
F32
F32
[24]
blk.13.hc_ffn_fn.weight
F32
F32
[20480, 24]
blk.13.hc_ffn_scale.weight
F32
F32
[3]
blk.14
blk.14.attn_compressor_gate.weight
BF16
BF16
[5120, 512]
blk.14.attn_compressor_kv.weight
BF16
BF16
[5120, 512]
blk.14.attn_compressor_norm.weight
F32
F32
[512]
blk.14.attn_kv.weight
Q6_K
Q6_K
[5120, 512]
blk.14.attn_kv_a_norm.weight
F32
F32
[512]
blk.14.attn_norm.weight
F32
F32
[5120]
blk.14.attn_output_a.weight
Q6_K
Q6_K
[4096, 8192]
blk.14.attn_output_b.weight
Q6_K
Q6_K
[8192, 5120]
blk.14.attn_q_a.weight
Q6_K
Q6_K
[5120, 1280]
blk.14.attn_q_a_norm.weight
F32
F32
[1280]
blk.14.attn_q_b.weight
Q6_K
Q6_K
[1280, 32768]
blk.14.attn_sinks.weight
F32
F32
[64]
blk.14.engram_gate.weight
F32
F32
[5120, 4]
blk.14.engram_kv.weight
Q6_K
Q6_K
[6144, 25600]
blk.14.exp_probs_b.bias
F32
F32
[384]
blk.14.exp_probs_b_vl.bias
F32
F32
[384]
blk.14.ffn_down_exps.weight
IQ2_XXS
IQ2_XXS
[2304, 5120, 384]
blk.14.ffn_down_shexp.weight
Q8_0
Q8_0
[2304, 5120]
blk.14.ffn_gate_exps.weight
IQ1_S
IQ1_S
[5120, 2304, 384]
blk.14.ffn_gate_inp.weight
F32
F32
[5120, 384]
blk.14.ffn_gate_shexp.weight
Q8_0
Q8_0
[5120, 2304]
blk.14.ffn_norm.weight
F32
F32
[5120]
blk.14.ffn_up_exps.weight
IQ1_S
IQ1_S
[5120, 2304, 384]
blk.14.ffn_up_shexp.weight
Q8_0
Q8_0
[5120, 2304]
blk.14.hc_attn_base.weight
F32
F32
[24]
blk.14.hc_attn_fn.weight
F32
F32
[20480, 24]
blk.14.hc_attn_scale.weight
F32
F32
[3]
blk.14.hc_ffn_base.weight
F32
F32
[24]
blk.14.hc_ffn_fn.weight
F32
F32
[20480, 24]
blk.14.hc_ffn_scale.weight
F32
F32
[3]
blk.14.indexer.attn_k.weight
BF16
BF16
[512, 128]
blk.14.indexer.attn_q_b.weight
BF16
BF16
[1280, 4096]
blk.14.indexer.k_norm.weight
F32
F32
[128]
blk.14.indexer.proj.weight
BF16
BF16
[5120, 32]
blk.15
blk.15.attn_kv.weight
Q6_K
Q6_K
[5120, 512]
blk.15.attn_kv_a_norm.weight
F32
F32
[512]
blk.15.attn_norm.weight
F32
F32
[5120]
blk.15.attn_output_a.weight
Q6_K
Q6_K
[4096, 8192]
blk.15.attn_output_b.weight
Q6_K
Q6_K
[8192, 5120]
blk.15.attn_q_a.weight
Q6_K
Q6_K
[5120, 1280]
blk.15.attn_q_a_norm.weight
F32
F32
[1280]
blk.15.attn_q_b.weight
Q6_K
Q6_K
[1280, 32768]
blk.15.attn_sinks.weight
F32
F32
[64]
blk.15.exp_probs_b.bias
F32
F32
[384]
blk.15.exp_probs_b_vl.bias
F32
F32
[384]
blk.15.ffn_down_exps.weight
IQ2_XXS
IQ2_XXS
[2304, 5120, 384]
blk.15.ffn_down_shexp.weight
Q8_0
Q8_0
[2304, 5120]
blk.15.ffn_gate_exps.weight
IQ1_S
IQ1_S
[5120, 2304, 384]
blk.15.ffn_gate_inp.weight
F32
F32
[5120, 384]
blk.15.ffn_gate_shexp.weight
Q8_0
Q8_0
[5120, 2304]
blk.15.ffn_norm.weight
F32
F32
[5120]
blk.15.ffn_up_exps.weight
IQ1_S
IQ1_S
[5120, 2304, 384]
blk.15.ffn_up_shexp.weight
Q8_0
Q8_0
[5120, 2304]
blk.15.hc_attn_base.weight
F32
F32
[24]
blk.15.hc_attn_fn.weight
F32
F32
[20480, 24]
blk.15.hc_attn_scale.weight
F32
F32
[3]
blk.15.hc_ffn_base.weight
F32
F32
[24]
blk.15.hc_ffn_fn.weight
F32
F32
[20480, 24]
blk.15.hc_ffn_scale.weight
F32
F32
[3]
blk.16
blk.16.attn_kv.weight
Q6_K
Q6_K
[5120, 512]
blk.16.attn_kv_a_norm.weight
F32
F32
[512]
blk.16.attn_norm.weight
F32
F32
[5120]
blk.16.attn_output_a.weight
Q6_K
Q6_K
[4096, 8192]
blk.16.attn_output_b.weight
Q6_K
Q6_K
[8192, 5120]
blk.16.attn_q_a.weight
Q6_K
Q6_K
[5120, 1280]
blk.16.attn_q_a_norm.weight
F32
F32
[1280]
blk.16.attn_q_b.weight
Q6_K
Q6_K
[1280, 32768]
blk.16.attn_sinks.weight
F32
F32
[64]
blk.16.exp_probs_b.bias
F32
F32
[384]
blk.16.exp_probs_b_vl.bias
F32
F32
[384]
output.weight
Q6_K
Q6_K
[5120, 129280]
output_norm.weight
F32
F32
[5120]