Models
Docs
Pricing
Sign in
Download
Models
Download
Docs
Pricing
Sign in
batiai
/
mistral-small-4
:q3
54
Downloads
Updated
1 week ago
Cancel
mistral-small-4:q3
...
/
model
af35c1f74332 · 48GB
Metadata
general.architecture
mistral4
mistral4
general.file_type
Q3_K_M
Q3_K_M
mistral4.attention.head_count
32
32
mistral4.attention.head_count_kv
1
1
mistral4.attention.key_length
320
320
mistral4.attention.key_length_mla
128
128
mistral4.attention.kv_lora_rank
256
256
mistral4.attention.layer_norm_rms_epsilon
1e-06
1e-06
mistral4.attention.q_lora_rank
1024
1024
mistral4.attention.value_length
256
256
mistral4.attention.value_length_mla
128
128
mistral4.block_count
36
36
mistral4.context_length
1048576
1048576
mistral4.embedding_length
4096
4096
mistral4.expert_count
128
128
mistral4.expert_feed_forward_length
2048
2048
mistral4.expert_group_count
1
1
mistral4.expert_group_used_count
1
1
mistral4.expert_shared_count
1
1
mistral4.expert_used_count
4
4
mistral4.expert_weights_norm
true
true
mistral4.expert_weights_scale
1
1
mistral4.feed_forward_length
12288
12288
mistral4.leading_dense_block_count
0
0
mistral4.rope.dimension_count
64
64
mistral4.rope.freq_base
10000
10000
mistral4.rope.scaling.factor
128
128
mistral4.rope.scaling.original_context_length
8192
8192
mistral4.rope.scaling.type
yarn
yarn
mistral4.rope.scaling.yarn_beta_fast
32
32
mistral4.rope.scaling.yarn_beta_slow
1
1
mistral4.rope.scaling.yarn_log_multiplier
0.1
0.1
mistral4.vocab_size
131072
131072
tokenizer.ggml.bos_token_id
1
1
tokenizer.ggml.eos_token_id
2
2
tokenizer.ggml.merges
[Ġ Ġ, Ġ t, e r, i n, Ġ ĠĠĠ, ...]
[Ġ Ġ, Ġ t, e r, i n, Ġ ĠĠĠ, ...]
tokenizer.ggml.model
gpt2
gpt2
tokenizer.ggml.padding_token_id
11
11
tokenizer.ggml.pre
pixtral
pixtral
tokenizer.ggml.token_type
[3, 3, 3, 3, 3, ...]
[3, 3, 3, 3, 3, ...]
tokenizer.ggml.tokens
[<unk>, <s>, </s>, [INST], [/INST], ...]
[<unk>, <s>, </s>, [INST], [/INST], ...]
tokenizer.ggml.unknown_token_id
0
0
split.count
2
2
split.no
0
0
split.tensors.count
579
579
Tensor
Name
Type
Shape
token_embd.weight
Q3_K
Q3_K
[4096, 131072]
blk.0
blk.0.attn_k_b.weight
Q4_0
Q4_0
[64, 256, 32]
blk.0.attn_kv_a_mqa.weight
Q3_K
Q3_K
[4096, 320]
blk.0.attn_kv_a_norm.weight
F32
F32
[256]
blk.0.attn_norm.weight
F32
F32
[4096]
blk.0.attn_output.weight
Q4_K
Q4_K
[4096, 4096]
blk.0.attn_q_a.weight
Q3_K
Q3_K
[4096, 1024]
blk.0.attn_q_a_norm.weight
F32
F32
[1024]
blk.0.attn_q_b.weight
Q3_K
Q3_K
[1024, 4096]
blk.0.attn_v_b.weight
Q3_K
Q3_K
[256, 128, 32]
blk.0.ffn_down_exps.weight
Q5_K
Q5_K
[2048, 4096, 128]
blk.0.ffn_down_shexp.weight
Q5_K
Q5_K
[2048, 4096]
blk.0.ffn_gate_inp.weight
F32
F32
[4096, 128]
blk.0.ffn_gate_shexp.weight
Q3_K
Q3_K
[4096, 2048]
blk.0.ffn_gate_up_exps.weight
Q3_K
Q3_K
[4096, 4096, 128]
blk.0.ffn_norm.weight
F32
F32
[4096]
blk.0.ffn_up_shexp.weight
Q3_K
Q3_K
[4096, 2048]
blk.1
blk.1.attn_k_b.weight
Q4_0
Q4_0
[64, 256, 32]
blk.1.attn_kv_a_mqa.weight
Q3_K
Q3_K
[4096, 320]
blk.1.attn_kv_a_norm.weight
F32
F32
[256]
blk.1.attn_norm.weight
F32
F32
[4096]
blk.1.attn_output.weight
Q4_K
Q4_K
[4096, 4096]
blk.1.attn_q_a.weight
Q3_K
Q3_K
[4096, 1024]
blk.1.attn_q_a_norm.weight
F32
F32
[1024]
blk.1.attn_q_b.weight
Q3_K
Q3_K
[1024, 4096]
blk.1.attn_v_b.weight
Q3_K
Q3_K
[256, 128, 32]
blk.1.ffn_down_exps.weight
Q5_K
Q5_K
[2048, 4096, 128]
blk.1.ffn_down_shexp.weight
Q5_K
Q5_K
[2048, 4096]
blk.1.ffn_gate_inp.weight
F32
F32
[4096, 128]
blk.1.ffn_gate_shexp.weight
Q3_K
Q3_K
[4096, 2048]
blk.1.ffn_gate_up_exps.weight
Q3_K
Q3_K
[4096, 4096, 128]
blk.1.ffn_norm.weight
F32
F32
[4096]
blk.1.ffn_up_shexp.weight
Q3_K
Q3_K
[4096, 2048]
blk.2
blk.2.attn_k_b.weight
Q4_0
Q4_0
[64, 256, 32]
blk.2.attn_kv_a_mqa.weight
Q3_K
Q3_K
[4096, 320]
blk.2.attn_kv_a_norm.weight
F32
F32
[256]
blk.2.attn_norm.weight
F32
F32
[4096]
blk.2.attn_output.weight
Q4_K
Q4_K
[4096, 4096]
blk.2.attn_q_a.weight
Q3_K
Q3_K
[4096, 1024]
blk.2.attn_q_a_norm.weight
F32
F32
[1024]
blk.2.attn_q_b.weight
Q3_K
Q3_K
[1024, 4096]
blk.2.attn_v_b.weight
Q3_K
Q3_K
[256, 128, 32]
blk.2.ffn_down_exps.weight
Q4_K
Q4_K
[2048, 4096, 128]
blk.2.ffn_down_shexp.weight
Q4_K
Q4_K
[2048, 4096]
blk.2.ffn_gate_inp.weight
F32
F32
[4096, 128]
blk.2.ffn_gate_shexp.weight
Q3_K
Q3_K
[4096, 2048]
blk.2.ffn_gate_up_exps.weight
Q3_K
Q3_K
[4096, 4096, 128]
blk.2.ffn_norm.weight
F32
F32
[4096]
blk.2.ffn_up_shexp.weight
Q3_K
Q3_K
[4096, 2048]
blk.3
blk.3.attn_k_b.weight
Q4_0
Q4_0
[64, 256, 32]
blk.3.attn_kv_a_mqa.weight
Q3_K
Q3_K
[4096, 320]
blk.3.attn_kv_a_norm.weight
F32
F32
[256]
blk.3.attn_norm.weight
F32
F32
[4096]
blk.3.attn_output.weight
Q4_K
Q4_K
[4096, 4096]
blk.3.attn_q_a.weight
Q3_K
Q3_K
[4096, 1024]
blk.3.attn_q_a_norm.weight
F32
F32
[1024]
blk.3.attn_q_b.weight
Q3_K
Q3_K
[1024, 4096]
blk.3.attn_v_b.weight
Q3_K
Q3_K
[256, 128, 32]
blk.3.ffn_down_exps.weight
Q4_K
Q4_K
[2048, 4096, 128]
blk.3.ffn_down_shexp.weight
Q4_K
Q4_K
[2048, 4096]
blk.3.ffn_gate_inp.weight
F32
F32
[4096, 128]
blk.3.ffn_gate_shexp.weight
Q3_K
Q3_K
[4096, 2048]
blk.3.ffn_gate_up_exps.weight
Q3_K
Q3_K
[4096, 4096, 128]
blk.3.ffn_norm.weight
F32
F32
[4096]
blk.3.ffn_up_shexp.weight
Q3_K
Q3_K
[4096, 2048]
blk.4
blk.4.attn_k_b.weight
Q4_0
Q4_0
[64, 256, 32]
blk.4.attn_kv_a_mqa.weight
Q3_K
Q3_K
[4096, 320]
blk.4.attn_kv_a_norm.weight
F32
F32
[256]
blk.4.attn_norm.weight
F32
F32
[4096]
blk.4.attn_output.weight
Q4_K
Q4_K
[4096, 4096]
blk.4.attn_q_a.weight
Q3_K
Q3_K
[4096, 1024]
blk.4.attn_q_a_norm.weight
F32
F32
[1024]
blk.4.attn_q_b.weight
Q3_K
Q3_K
[1024, 4096]
blk.4.attn_v_b.weight
Q3_K
Q3_K
[256, 128, 32]
blk.4.ffn_down_exps.weight
Q4_K
Q4_K
[2048, 4096, 128]
blk.4.ffn_down_shexp.weight
Q4_K
Q4_K
[2048, 4096]
blk.4.ffn_gate_inp.weight
F32
F32
[4096, 128]
blk.4.ffn_gate_shexp.weight
Q3_K
Q3_K
[4096, 2048]
blk.4.ffn_gate_up_exps.weight
Q3_K
Q3_K
[4096, 4096, 128]
blk.4.ffn_norm.weight
F32
F32
[4096]
blk.4.ffn_up_shexp.weight
Q3_K
Q3_K
[4096, 2048]
blk.5
blk.5.attn_k_b.weight
Q4_0
Q4_0
[64, 256, 32]
blk.5.attn_kv_a_mqa.weight
Q3_K
Q3_K
[4096, 320]
blk.5.attn_kv_a_norm.weight
F32
F32
[256]
blk.5.attn_norm.weight
F32
F32
[4096]
blk.5.attn_output.weight
Q4_K
Q4_K
[4096, 4096]
blk.5.attn_q_a.weight
Q3_K
Q3_K
[4096, 1024]
blk.5.attn_q_a_norm.weight
F32
F32
[1024]
blk.5.attn_q_b.weight
Q3_K
Q3_K
[1024, 4096]
blk.5.attn_v_b.weight
Q3_K
Q3_K
[256, 128, 32]
blk.5.ffn_down_exps.weight
Q4_K
Q4_K
[2048, 4096, 128]
blk.5.ffn_down_shexp.weight
Q4_K
Q4_K
[2048, 4096]
blk.5.ffn_gate_inp.weight
F32
F32
[4096, 128]
blk.5.ffn_gate_shexp.weight
Q3_K
Q3_K
[4096, 2048]
blk.5.ffn_gate_up_exps.weight
Q3_K
Q3_K
[4096, 4096, 128]
blk.5.ffn_norm.weight
F32
F32
[4096]
blk.5.ffn_up_shexp.weight
Q3_K
Q3_K
[4096, 2048]
blk.6
blk.6.attn_k_b.weight
Q4_0
Q4_0
[64, 256, 32]
blk.6.attn_kv_a_mqa.weight
Q3_K
Q3_K
[4096, 320]
blk.6.attn_kv_a_norm.weight
F32
F32
[256]
blk.6.attn_norm.weight
F32
F32
[4096]
blk.6.attn_output.weight
Q4_K
Q4_K
[4096, 4096]
blk.6.attn_q_a.weight
Q3_K
Q3_K
[4096, 1024]
blk.6.attn_q_a_norm.weight
F32
F32
[1024]
blk.6.attn_q_b.weight
Q3_K
Q3_K
[1024, 4096]
blk.6.attn_v_b.weight
Q3_K
Q3_K
[256, 128, 32]
blk.6.ffn_down_exps.weight
Q4_K
Q4_K
[2048, 4096, 128]
blk.6.ffn_down_shexp.weight
Q4_K
Q4_K
[2048, 4096]
blk.6.ffn_gate_inp.weight
F32
F32
[4096, 128]
blk.6.ffn_gate_shexp.weight
Q3_K
Q3_K
[4096, 2048]
blk.6.ffn_gate_up_exps.weight
Q3_K
Q3_K
[4096, 4096, 128]
blk.6.ffn_norm.weight
F32
F32
[4096]
blk.6.ffn_up_shexp.weight
Q3_K
Q3_K
[4096, 2048]
blk.7
blk.7.attn_k_b.weight
Q4_0
Q4_0
[64, 256, 32]
blk.7.attn_kv_a_mqa.weight
Q3_K
Q3_K
[4096, 320]
blk.7.attn_kv_a_norm.weight
F32
F32
[256]
blk.7.attn_norm.weight
F32
F32
[4096]
blk.7.attn_output.weight
Q4_K
Q4_K
[4096, 4096]
blk.7.attn_q_a.weight
Q3_K
Q3_K
[4096, 1024]
blk.7.attn_q_a_norm.weight
F32
F32
[1024]
blk.7.attn_q_b.weight
Q3_K
Q3_K
[1024, 4096]
blk.7.attn_v_b.weight
Q3_K
Q3_K
[256, 128, 32]
blk.7.ffn_down_exps.weight
Q4_K
Q4_K
[2048, 4096, 128]
blk.7.ffn_down_shexp.weight
Q4_K
Q4_K
[2048, 4096]
blk.7.ffn_gate_inp.weight
F32
F32
[4096, 128]
blk.7.ffn_gate_shexp.weight
Q3_K
Q3_K
[4096, 2048]
blk.7.ffn_gate_up_exps.weight
Q3_K
Q3_K
[4096, 4096, 128]
blk.7.ffn_norm.weight
F32
F32
[4096]
blk.7.ffn_up_shexp.weight
Q3_K
Q3_K
[4096, 2048]
blk.8
blk.8.attn_k_b.weight
Q4_0
Q4_0
[64, 256, 32]
blk.8.attn_kv_a_mqa.weight
Q3_K
Q3_K
[4096, 320]
blk.8.attn_kv_a_norm.weight
F32
F32
[256]
blk.8.attn_norm.weight
F32
F32
[4096]
blk.8.attn_output.weight
Q4_K
Q4_K
[4096, 4096]
blk.8.attn_q_a.weight
Q3_K
Q3_K
[4096, 1024]
blk.8.attn_q_a_norm.weight
F32
F32
[1024]
blk.8.attn_q_b.weight
Q3_K
Q3_K
[1024, 4096]
blk.8.attn_v_b.weight
Q3_K
Q3_K
[256, 128, 32]
blk.8.ffn_down_exps.weight
Q4_K
Q4_K
[2048, 4096, 128]
blk.8.ffn_down_shexp.weight
Q4_K
Q4_K
[2048, 4096]
blk.8.ffn_gate_inp.weight
F32
F32
[4096, 128]
blk.8.ffn_gate_shexp.weight
Q3_K
Q3_K
[4096, 2048]
blk.8.ffn_gate_up_exps.weight
Q3_K
Q3_K
[4096, 4096, 128]
blk.8.ffn_norm.weight
F32
F32
[4096]
blk.8.ffn_up_shexp.weight
Q3_K
Q3_K
[4096, 2048]
blk.9
blk.9.attn_k_b.weight
Q4_0
Q4_0
[64, 256, 32]
blk.9.attn_kv_a_mqa.weight
Q3_K
Q3_K
[4096, 320]
blk.9.attn_kv_a_norm.weight
F32
F32
[256]
blk.9.attn_norm.weight
F32
F32
[4096]
blk.9.attn_output.weight
Q4_K
Q4_K
[4096, 4096]
blk.9.attn_q_a.weight
Q3_K
Q3_K
[4096, 1024]
blk.9.attn_q_a_norm.weight
F32
F32
[1024]
blk.9.attn_q_b.weight
Q3_K
Q3_K
[1024, 4096]
blk.9.attn_v_b.weight
Q3_K
Q3_K
[256, 128, 32]
blk.9.ffn_down_exps.weight
Q4_K
Q4_K
[2048, 4096, 128]
blk.9.ffn_down_shexp.weight
Q4_K
Q4_K
[2048, 4096]
blk.9.ffn_gate_inp.weight
F32
F32
[4096, 128]
blk.9.ffn_gate_shexp.weight
Q3_K
Q3_K
[4096, 2048]
blk.9.ffn_gate_up_exps.weight
Q3_K
Q3_K
[4096, 4096, 128]
blk.9.ffn_norm.weight
F32
F32
[4096]
blk.9.ffn_up_shexp.weight
Q3_K
Q3_K
[4096, 2048]
blk.10
blk.10.attn_k_b.weight
Q4_0
Q4_0
[64, 256, 32]
blk.10.attn_kv_a_mqa.weight
Q3_K
Q3_K
[4096, 320]
blk.10.attn_kv_a_norm.weight
F32
F32
[256]
blk.10.attn_norm.weight
F32
F32
[4096]
blk.10.attn_output.weight
Q4_K
Q4_K
[4096, 4096]
blk.10.attn_q_a.weight
Q3_K
Q3_K
[4096, 1024]
blk.10.attn_q_a_norm.weight
F32
F32
[1024]
blk.10.attn_q_b.weight
Q3_K
Q3_K
[1024, 4096]
blk.10.attn_v_b.weight
Q3_K
Q3_K
[256, 128, 32]
blk.10.ffn_down_exps.weight
Q4_K
Q4_K
[2048, 4096, 128]
blk.10.ffn_down_shexp.weight
Q4_K
Q4_K
[2048, 4096]
blk.10.ffn_gate_inp.weight
F32
F32
[4096, 128]
blk.10.ffn_gate_shexp.weight
Q3_K
Q3_K
[4096, 2048]
blk.10.ffn_gate_up_exps.weight
Q3_K
Q3_K
[4096, 4096, 128]
blk.10.ffn_norm.weight
F32
F32
[4096]
blk.10.ffn_up_shexp.weight
Q3_K
Q3_K
[4096, 2048]
blk.11
blk.11.attn_k_b.weight
Q4_0
Q4_0
[64, 256, 32]
blk.11.attn_kv_a_mqa.weight
Q3_K
Q3_K
[4096, 320]
blk.11.attn_kv_a_norm.weight
F32
F32
[256]
blk.11.attn_norm.weight
F32
F32
[4096]
blk.11.attn_output.weight
Q4_K
Q4_K
[4096, 4096]
blk.11.attn_q_a.weight
Q3_K
Q3_K
[4096, 1024]
blk.11.attn_q_a_norm.weight
F32
F32
[1024]
blk.11.attn_q_b.weight
Q3_K
Q3_K
[1024, 4096]
blk.11.attn_v_b.weight
Q3_K
Q3_K
[256, 128, 32]
blk.11.ffn_down_exps.weight
Q4_K
Q4_K
[2048, 4096, 128]
blk.11.ffn_down_shexp.weight
Q4_K
Q4_K
[2048, 4096]
blk.11.ffn_gate_inp.weight
F32
F32
[4096, 128]
blk.11.ffn_gate_shexp.weight
Q3_K
Q3_K
[4096, 2048]
blk.11.ffn_gate_up_exps.weight
Q3_K
Q3_K
[4096, 4096, 128]
blk.11.ffn_norm.weight
F32
F32
[4096]
blk.11.ffn_up_shexp.weight
Q3_K
Q3_K
[4096, 2048]
blk.12
blk.12.attn_k_b.weight
Q4_0
Q4_0
[64, 256, 32]
blk.12.attn_kv_a_mqa.weight
Q3_K
Q3_K
[4096, 320]
blk.12.attn_kv_a_norm.weight
F32
F32
[256]
blk.12.attn_norm.weight
F32
F32
[4096]
blk.12.attn_output.weight
Q4_K
Q4_K
[4096, 4096]
blk.12.attn_q_a.weight
Q3_K
Q3_K
[4096, 1024]
blk.12.attn_q_a_norm.weight
F32
F32
[1024]
blk.12.attn_q_b.weight
Q3_K
Q3_K
[1024, 4096]
blk.12.attn_v_b.weight
Q3_K
Q3_K
[256, 128, 32]
blk.12.ffn_down_exps.weight
Q4_K
Q4_K
[2048, 4096, 128]
blk.12.ffn_down_shexp.weight
Q4_K
Q4_K
[2048, 4096]
blk.12.ffn_gate_inp.weight
F32
F32
[4096, 128]
blk.12.ffn_gate_shexp.weight
Q3_K
Q3_K
[4096, 2048]
blk.12.ffn_gate_up_exps.weight
Q3_K
Q3_K
[4096, 4096, 128]
blk.12.ffn_norm.weight
F32
F32
[4096]
blk.12.ffn_up_shexp.weight
Q3_K
Q3_K
[4096, 2048]
blk.13
blk.13.attn_k_b.weight
Q4_0
Q4_0
[64, 256, 32]
blk.13.attn_kv_a_mqa.weight
Q3_K
Q3_K
[4096, 320]
blk.13.attn_kv_a_norm.weight
F32
F32
[256]
blk.13.attn_norm.weight
F32
F32
[4096]
blk.13.attn_output.weight
Q4_K
Q4_K
[4096, 4096]
blk.13.attn_q_a.weight
Q3_K
Q3_K
[4096, 1024]
blk.13.attn_q_a_norm.weight
F32
F32
[1024]
blk.13.attn_q_b.weight
Q3_K
Q3_K
[1024, 4096]
blk.13.attn_v_b.weight
Q3_K
Q3_K
[256, 128, 32]
blk.13.ffn_down_exps.weight
Q4_K
Q4_K
[2048, 4096, 128]
blk.13.ffn_down_shexp.weight
Q4_K
Q4_K
[2048, 4096]
blk.13.ffn_gate_inp.weight
F32
F32
[4096, 128]
blk.13.ffn_gate_shexp.weight
Q3_K
Q3_K
[4096, 2048]
blk.13.ffn_gate_up_exps.weight
Q3_K
Q3_K
[4096, 4096, 128]
blk.13.ffn_norm.weight
F32
F32
[4096]
blk.13.ffn_up_shexp.weight
Q3_K
Q3_K
[4096, 2048]
blk.14
blk.14.attn_k_b.weight
Q4_0
Q4_0
[64, 256, 32]
blk.14.attn_kv_a_mqa.weight
Q3_K
Q3_K
[4096, 320]
blk.14.attn_kv_a_norm.weight
F32
F32
[256]
blk.14.attn_norm.weight
F32
F32
[4096]
blk.14.attn_output.weight
Q4_K
Q4_K
[4096, 4096]
blk.14.attn_q_a.weight
Q3_K
Q3_K
[4096, 1024]
blk.14.attn_q_a_norm.weight
F32
F32
[1024]
blk.14.attn_q_b.weight
Q3_K
Q3_K
[1024, 4096]
blk.14.attn_v_b.weight
Q3_K
Q3_K
[256, 128, 32]
blk.14.ffn_down_exps.weight
Q4_K
Q4_K
[2048, 4096, 128]
blk.14.ffn_down_shexp.weight
Q4_K
Q4_K
[2048, 4096]
blk.14.ffn_gate_inp.weight
F32
F32
[4096, 128]
blk.14.ffn_gate_shexp.weight
Q3_K
Q3_K
[4096, 2048]
blk.14.ffn_gate_up_exps.weight
Q3_K
Q3_K
[4096, 4096, 128]
blk.14.ffn_norm.weight
F32
F32
[4096]
blk.14.ffn_up_shexp.weight
Q3_K
Q3_K
[4096, 2048]
blk.15
blk.15.attn_k_b.weight
Q4_0
Q4_0
[64, 256, 32]
blk.15.attn_kv_a_mqa.weight
Q3_K
Q3_K
[4096, 320]
blk.15.attn_kv_a_norm.weight
F32
F32
[256]
blk.15.attn_norm.weight
F32
F32
[4096]
blk.15.attn_output.weight
Q4_K
Q4_K
[4096, 4096]
blk.15.attn_q_a.weight
Q3_K
Q3_K
[4096, 1024]
blk.15.attn_q_a_norm.weight
F32
F32
[1024]
blk.15.attn_q_b.weight
Q3_K
Q3_K
[1024, 4096]
blk.15.attn_v_b.weight
Q3_K
Q3_K
[256, 128, 32]
blk.15.ffn_down_exps.weight
Q4_K
Q4_K
[2048, 4096, 128]
blk.15.ffn_down_shexp.weight
Q4_K
Q4_K
[2048, 4096]
blk.15.ffn_gate_inp.weight
F32
F32
[4096, 128]
blk.15.ffn_gate_shexp.weight
Q3_K
Q3_K
[4096, 2048]
blk.15.ffn_gate_up_exps.weight
Q3_K
Q3_K
[4096, 4096, 128]
blk.15.ffn_norm.weight
F32
F32
[4096]
blk.15.ffn_up_shexp.weight
Q3_K
Q3_K
[4096, 2048]
blk.16
blk.16.attn_k_b.weight
Q4_0
Q4_0
[64, 256, 32]
blk.16.attn_kv_a_mqa.weight
Q3_K
Q3_K
[4096, 320]
blk.16.attn_kv_a_norm.weight
F32
F32
[256]
blk.16.attn_norm.weight
F32
F32
[4096]
blk.16.attn_output.weight
Q4_K
Q4_K
[4096, 4096]
blk.16.attn_q_a.weight
Q3_K
Q3_K
[4096, 1024]
blk.16.attn_q_a_norm.weight
F32
F32
[1024]
blk.16.attn_q_b.weight
Q3_K
Q3_K
[1024, 4096]
blk.16.attn_v_b.weight
Q3_K
Q3_K
[256, 128, 32]
blk.16.ffn_down_exps.weight
Q4_K
Q4_K
[2048, 4096, 128]
blk.16.ffn_down_shexp.weight
Q4_K
Q4_K
[2048, 4096]
blk.16.ffn_gate_inp.weight
F32
F32
[4096, 128]
blk.16.ffn_gate_shexp.weight
Q3_K
Q3_K
[4096, 2048]
blk.16.ffn_gate_up_exps.weight
Q3_K
Q3_K
[4096, 4096, 128]
blk.16.ffn_norm.weight
F32
F32
[4096]
blk.16.ffn_up_shexp.weight
Q3_K
Q3_K
[4096, 2048]
blk.17
blk.17.attn_k_b.weight
Q4_0
Q4_0
[64, 256, 32]
blk.17.attn_kv_a_mqa.weight
Q3_K
Q3_K
[4096, 320]
blk.17.attn_kv_a_norm.weight
F32
F32
[256]
blk.17.attn_norm.weight
F32
F32
[4096]
blk.17.attn_output.weight
Q4_K
Q4_K
[4096, 4096]
blk.17.attn_q_a.weight
Q3_K
Q3_K
[4096, 1024]
blk.17.attn_q_a_norm.weight
F32
F32
[1024]
blk.17.attn_q_b.weight
Q3_K
Q3_K
[1024, 4096]
blk.17.attn_v_b.weight
Q3_K
Q3_K
[256, 128, 32]
blk.17.ffn_down_exps.weight
Q4_K
Q4_K
[2048, 4096, 128]
blk.17.ffn_down_shexp.weight
Q4_K
Q4_K
[2048, 4096]
blk.17.ffn_gate_inp.weight
F32
F32
[4096, 128]
blk.17.ffn_gate_shexp.weight
Q3_K
Q3_K
[4096, 2048]
blk.17.ffn_gate_up_exps.weight
Q3_K
Q3_K
[4096, 4096, 128]
blk.17.ffn_norm.weight
F32
F32
[4096]
blk.17.ffn_up_shexp.weight
Q3_K
Q3_K
[4096, 2048]
blk.18
blk.18.attn_k_b.weight
Q4_0
Q4_0
[64, 256, 32]
blk.18.attn_kv_a_mqa.weight
Q3_K
Q3_K
[4096, 320]
blk.18.attn_kv_a_norm.weight
F32
F32
[256]
blk.18.attn_norm.weight
F32
F32
[4096]
blk.18.attn_output.weight
Q4_K
Q4_K
[4096, 4096]
blk.18.attn_q_a.weight
Q3_K
Q3_K
[4096, 1024]
blk.18.attn_q_a_norm.weight
F32
F32
[1024]
blk.18.attn_q_b.weight
Q3_K
Q3_K
[1024, 4096]
blk.18.attn_v_b.weight
Q3_K
Q3_K
[256, 128, 32]
blk.18.ffn_down_exps.weight
Q4_K
Q4_K
[2048, 4096, 128]
blk.18.ffn_down_shexp.weight
Q4_K
Q4_K
[2048, 4096]
blk.18.ffn_gate_inp.weight
F32
F32
[4096, 128]
blk.18.ffn_gate_shexp.weight
Q3_K
Q3_K
[4096, 2048]
blk.18.ffn_gate_up_exps.weight
Q3_K
Q3_K
[4096, 4096, 128]
blk.18.ffn_norm.weight
F32
F32
[4096]
blk.18.ffn_up_shexp.weight
Q3_K
Q3_K
[4096, 2048]
blk.19
blk.19.attn_k_b.weight
Q4_0
Q4_0
[64, 256, 32]
blk.19.attn_kv_a_mqa.weight
Q3_K
Q3_K
[4096, 320]
blk.19.attn_kv_a_norm.weight
F32
F32
[256]
blk.19.attn_norm.weight
F32
F32
[4096]
blk.19.attn_output.weight
Q4_K
Q4_K
[4096, 4096]
blk.19.attn_q_a.weight
Q3_K
Q3_K
[4096, 1024]
blk.19.attn_q_a_norm.weight
F32
F32
[1024]
blk.19.attn_q_b.weight
Q3_K
Q3_K
[1024, 4096]
blk.19.attn_v_b.weight
Q3_K
Q3_K
[256, 128, 32]
blk.19.ffn_down_exps.weight
Q4_K
Q4_K
[2048, 4096, 128]
blk.19.ffn_down_shexp.weight
Q4_K
Q4_K
[2048, 4096]
blk.19.ffn_gate_inp.weight
F32
F32
[4096, 128]
blk.19.ffn_gate_shexp.weight
Q3_K
Q3_K
[4096, 2048]
blk.19.ffn_gate_up_exps.weight
Q3_K
Q3_K
[4096, 4096, 128]
blk.19.ffn_norm.weight
F32
F32
[4096]
blk.19.ffn_up_shexp.weight
Q3_K
Q3_K
[4096, 2048]
blk.20
blk.20.attn_k_b.weight
Q4_0
Q4_0
[64, 256, 32]
blk.20.attn_kv_a_mqa.weight
Q3_K
Q3_K
[4096, 320]
blk.20.attn_kv_a_norm.weight
F32
F32
[256]
blk.20.attn_norm.weight
F32
F32
[4096]
blk.20.attn_output.weight
Q4_K
Q4_K
[4096, 4096]
blk.20.attn_q_a.weight
Q3_K
Q3_K
[4096, 1024]
blk.20.attn_q_a_norm.weight
F32
F32
[1024]
blk.20.attn_q_b.weight
Q3_K
Q3_K
[1024, 4096]
blk.20.attn_v_b.weight
Q3_K
Q3_K
[256, 128, 32]
blk.20.ffn_down_exps.weight
Q4_K
Q4_K
[2048, 4096, 128]
blk.20.ffn_down_shexp.weight
Q4_K
Q4_K
[2048, 4096]
blk.20.ffn_gate_inp.weight
F32
F32
[4096, 128]
blk.20.ffn_gate_shexp.weight
Q3_K
Q3_K
[4096, 2048]
blk.20.ffn_gate_up_exps.weight
Q3_K
Q3_K
[4096, 4096, 128]
blk.20.ffn_norm.weight
F32
F32
[4096]
blk.20.ffn_up_shexp.weight
Q3_K
Q3_K
[4096, 2048]
blk.21
blk.21.attn_k_b.weight
Q4_0
Q4_0
[64, 256, 32]
blk.21.attn_kv_a_mqa.weight
Q3_K
Q3_K
[4096, 320]
blk.21.attn_kv_a_norm.weight
F32
F32
[256]
blk.21.attn_norm.weight
F32
F32
[4096]
blk.21.attn_output.weight
Q4_K
Q4_K
[4096, 4096]
blk.21.attn_q_a.weight
Q3_K
Q3_K
[4096, 1024]
blk.21.attn_q_a_norm.weight
F32
F32
[1024]
blk.21.attn_q_b.weight
Q3_K
Q3_K
[1024, 4096]
blk.21.attn_v_b.weight
Q3_K
Q3_K
[256, 128, 32]
blk.21.ffn_down_exps.weight
Q4_K
Q4_K
[2048, 4096, 128]
blk.21.ffn_down_shexp.weight
Q4_K
Q4_K
[2048, 4096]
blk.21.ffn_gate_inp.weight
F32
F32
[4096, 128]
blk.21.ffn_gate_shexp.weight
Q3_K
Q3_K
[4096, 2048]
blk.21.ffn_gate_up_exps.weight
Q3_K
Q3_K
[4096, 4096, 128]
blk.21.ffn_norm.weight
F32
F32
[4096]
blk.21.ffn_up_shexp.weight
Q3_K
Q3_K
[4096, 2048]
blk.22
blk.22.attn_k_b.weight
Q4_0
Q4_0
[64, 256, 32]
blk.22.attn_kv_a_mqa.weight
Q3_K
Q3_K
[4096, 320]
blk.22.attn_kv_a_norm.weight
F32
F32
[256]
blk.22.attn_norm.weight
F32
F32
[4096]
blk.22.attn_output.weight
Q4_K
Q4_K
[4096, 4096]
blk.22.attn_q_a.weight
Q3_K
Q3_K
[4096, 1024]
blk.22.attn_q_a_norm.weight
F32
F32
[1024]
blk.22.attn_q_b.weight
Q3_K
Q3_K
[1024, 4096]
blk.22.attn_v_b.weight
Q3_K
Q3_K
[256, 128, 32]
blk.22.ffn_down_exps.weight
Q4_K
Q4_K
[2048, 4096, 128]
blk.22.ffn_down_shexp.weight
Q4_K
Q4_K
[2048, 4096]
blk.22.ffn_gate_inp.weight
F32
F32
[4096, 128]
blk.22.ffn_gate_shexp.weight
Q3_K
Q3_K
[4096, 2048]
blk.22.ffn_gate_up_exps.weight
Q3_K
Q3_K
[4096, 4096, 128]
blk.22.ffn_norm.weight
F32
F32
[4096]
blk.22.ffn_up_shexp.weight
Q3_K
Q3_K
[4096, 2048]
blk.23
blk.23.attn_k_b.weight
Q4_0
Q4_0
[64, 256, 32]
blk.23.attn_kv_a_mqa.weight
Q3_K
Q3_K
[4096, 320]
blk.23.attn_kv_a_norm.weight
F32
F32
[256]
blk.23.attn_norm.weight
F32
F32
[4096]
blk.23.attn_output.weight
Q4_K
Q4_K
[4096, 4096]
blk.23.attn_q_a.weight
Q3_K
Q3_K
[4096, 1024]
blk.23.attn_q_a_norm.weight
F32
F32
[1024]
blk.23.attn_q_b.weight
Q3_K
Q3_K
[1024, 4096]
blk.23.attn_v_b.weight
Q3_K
Q3_K
[256, 128, 32]
blk.23.ffn_down_exps.weight
Q4_K
Q4_K
[2048, 4096, 128]
blk.23.ffn_down_shexp.weight
Q4_K
Q4_K
[2048, 4096]
blk.23.ffn_gate_inp.weight
F32
F32
[4096, 128]
blk.23.ffn_gate_shexp.weight
Q3_K
Q3_K
[4096, 2048]
blk.23.ffn_gate_up_exps.weight
Q3_K
Q3_K
[4096, 4096, 128]
blk.23.ffn_norm.weight
F32
F32
[4096]
blk.23.ffn_up_shexp.weight
Q3_K
Q3_K
[4096, 2048]
blk.24
blk.24.attn_k_b.weight
Q4_0
Q4_0
[64, 256, 32]
blk.24.attn_kv_a_mqa.weight
Q3_K
Q3_K
[4096, 320]
blk.24.attn_kv_a_norm.weight
F32
F32
[256]
blk.24.attn_norm.weight
F32
F32
[4096]
blk.24.attn_output.weight
Q4_K
Q4_K
[4096, 4096]
blk.24.attn_q_a.weight
Q3_K
Q3_K
[4096, 1024]
blk.24.attn_q_a_norm.weight
F32
F32
[1024]
blk.24.attn_q_b.weight
Q3_K
Q3_K
[1024, 4096]
blk.24.attn_v_b.weight
Q3_K
Q3_K
[256, 128, 32]
blk.24.ffn_down_exps.weight
Q4_K
Q4_K
[2048, 4096, 128]
blk.24.ffn_down_shexp.weight
Q4_K
Q4_K
[2048, 4096]
blk.24.ffn_gate_inp.weight
F32
F32
[4096, 128]
blk.24.ffn_gate_shexp.weight
Q3_K
Q3_K
[4096, 2048]
blk.24.ffn_gate_up_exps.weight
Q3_K
Q3_K
[4096, 4096, 128]
blk.24.ffn_norm.weight
F32
F32
[4096]
blk.24.ffn_up_shexp.weight
Q3_K
Q3_K
[4096, 2048]
blk.25
blk.25.attn_k_b.weight
Q4_0
Q4_0
[64, 256, 32]
blk.25.attn_kv_a_mqa.weight
Q3_K
Q3_K
[4096, 320]
blk.25.attn_kv_a_norm.weight
F32
F32
[256]
blk.25.attn_norm.weight
F32
F32
[4096]
blk.25.attn_output.weight
Q4_K
Q4_K
[4096, 4096]
blk.25.attn_q_a.weight
Q3_K
Q3_K
[4096, 1024]
blk.25.attn_q_a_norm.weight
F32
F32
[1024]
blk.25.attn_q_b.weight
Q3_K
Q3_K
[1024, 4096]
blk.25.attn_v_b.weight
Q3_K
Q3_K
[256, 128, 32]
blk.25.ffn_down_exps.weight
Q4_K
Q4_K
[2048, 4096, 128]
blk.25.ffn_down_shexp.weight
Q4_K
Q4_K
[2048, 4096]
blk.25.ffn_gate_inp.weight
F32
F32
[4096, 128]
blk.25.ffn_gate_shexp.weight
Q3_K
Q3_K
[4096, 2048]
blk.25.ffn_gate_up_exps.weight
Q3_K
Q3_K
[4096, 4096, 128]
blk.25.ffn_norm.weight
F32
F32
[4096]
blk.25.ffn_up_shexp.weight
Q3_K
Q3_K
[4096, 2048]
blk.26
blk.26.attn_k_b.weight
Q4_0
Q4_0
[64, 256, 32]
blk.26.attn_kv_a_mqa.weight
Q3_K
Q3_K
[4096, 320]
blk.26.attn_kv_a_norm.weight
F32
F32
[256]
blk.26.attn_norm.weight
F32
F32
[4096]
blk.26.attn_output.weight
Q4_K
Q4_K
[4096, 4096]
blk.26.attn_q_a.weight
Q3_K
Q3_K
[4096, 1024]
blk.26.attn_q_a_norm.weight
F32
F32
[1024]
blk.26.attn_q_b.weight
Q3_K
Q3_K
[1024, 4096]
blk.26.attn_v_b.weight
Q3_K
Q3_K
[256, 128, 32]
blk.26.ffn_down_exps.weight
Q4_K
Q4_K
[2048, 4096, 128]
blk.26.ffn_down_shexp.weight
Q4_K
Q4_K
[2048, 4096]
blk.26.ffn_gate_inp.weight
F32
F32
[4096, 128]
blk.26.ffn_gate_shexp.weight
Q3_K
Q3_K
[4096, 2048]
blk.26.ffn_gate_up_exps.weight
Q3_K
Q3_K
[4096, 4096, 128]
blk.26.ffn_norm.weight
F32
F32
[4096]
blk.26.ffn_up_shexp.weight
Q3_K
Q3_K
[4096, 2048]
blk.27
blk.27.attn_k_b.weight
Q4_0
Q4_0
[64, 256, 32]
blk.27.attn_kv_a_mqa.weight
Q3_K
Q3_K
[4096, 320]
blk.27.attn_kv_a_norm.weight
F32
F32
[256]
blk.27.attn_norm.weight
F32
F32
[4096]
blk.27.attn_output.weight
Q4_K
Q4_K
[4096, 4096]
blk.27.attn_q_a.weight
Q3_K
Q3_K
[4096, 1024]
blk.27.attn_q_a_norm.weight
F32
F32
[1024]
blk.27.attn_q_b.weight
Q3_K
Q3_K
[1024, 4096]
blk.27.attn_v_b.weight
Q3_K
Q3_K
[256, 128, 32]
blk.27.ffn_down_exps.weight
Q4_K
Q4_K
[2048, 4096, 128]
blk.27.ffn_down_shexp.weight
Q4_K
Q4_K
[2048, 4096]
blk.27.ffn_gate_inp.weight
F32
F32
[4096, 128]
blk.27.ffn_gate_shexp.weight
Q3_K
Q3_K
[4096, 2048]
blk.27.ffn_gate_up_exps.weight
Q3_K
Q3_K
[4096, 4096, 128]
blk.27.ffn_norm.weight
F32
F32
[4096]
blk.27.ffn_up_shexp.weight
Q3_K
Q3_K
[4096, 2048]
blk.28
blk.28.attn_k_b.weight
Q4_0
Q4_0
[64, 256, 32]
blk.28.attn_kv_a_mqa.weight
Q3_K
Q3_K
[4096, 320]
blk.28.attn_kv_a_norm.weight
F32
F32
[256]
blk.28.attn_norm.weight
F32
F32
[4096]
blk.28.attn_output.weight
Q4_K
Q4_K
[4096, 4096]
blk.28.attn_q_a.weight
Q3_K
Q3_K
[4096, 1024]
blk.28.attn_q_a_norm.weight
F32
F32
[1024]
blk.28.attn_q_b.weight
Q3_K
Q3_K
[1024, 4096]
blk.28.attn_v_b.weight
Q3_K
Q3_K
[256, 128, 32]
blk.28.ffn_down_exps.weight
Q4_K
Q4_K
[2048, 4096, 128]
blk.28.ffn_down_shexp.weight
Q4_K
Q4_K
[2048, 4096]
blk.28.ffn_gate_inp.weight
F32
F32
[4096, 128]
blk.28.ffn_gate_shexp.weight
Q3_K
Q3_K
[4096, 2048]
blk.28.ffn_gate_up_exps.weight
Q3_K
Q3_K
[4096, 4096, 128]
blk.28.ffn_norm.weight
F32
F32
[4096]
blk.28.ffn_up_shexp.weight
Q3_K
Q3_K
[4096, 2048]
blk.29
blk.29.attn_k_b.weight
Q4_0
Q4_0
[64, 256, 32]
blk.29.attn_kv_a_mqa.weight
Q3_K
Q3_K
[4096, 320]
blk.29.attn_kv_a_norm.weight
F32
F32
[256]
blk.29.attn_norm.weight
F32
F32
[4096]
blk.29.attn_output.weight
Q4_K
Q4_K
[4096, 4096]
blk.29.attn_q_a.weight
Q3_K
Q3_K
[4096, 1024]
blk.29.attn_q_a_norm.weight
F32
F32
[1024]
blk.29.attn_q_b.weight
Q3_K
Q3_K
[1024, 4096]
blk.29.attn_v_b.weight
Q3_K
Q3_K
[256, 128, 32]
blk.29.ffn_down_exps.weight
Q4_K
Q4_K
[2048, 4096, 128]
blk.29.ffn_down_shexp.weight
Q4_K
Q4_K
[2048, 4096]
blk.29.ffn_gate_inp.weight
F32
F32
[4096, 128]
blk.29.ffn_gate_shexp.weight
Q3_K
Q3_K
[4096, 2048]
blk.29.ffn_gate_up_exps.weight
Q3_K
Q3_K
[4096, 4096, 128]
blk.29.ffn_norm.weight
F32
F32
[4096]
blk.29.ffn_up_shexp.weight
Q3_K
Q3_K
[4096, 2048]
blk.30
blk.30.attn_k_b.weight
Q4_0
Q4_0
[64, 256, 32]
blk.30.attn_kv_a_mqa.weight
Q3_K
Q3_K
[4096, 320]
blk.30.attn_kv_a_norm.weight
F32
F32
[256]
blk.30.attn_norm.weight
F32
F32
[4096]
blk.30.attn_output.weight
Q4_K
Q4_K
[4096, 4096]
blk.30.attn_q_a.weight
Q3_K
Q3_K
[4096, 1024]
blk.30.attn_q_a_norm.weight
F32
F32
[1024]
blk.30.attn_q_b.weight
Q3_K
Q3_K
[1024, 4096]
blk.30.attn_v_b.weight
Q3_K
Q3_K
[256, 128, 32]
output.weight
Q6_K
Q6_K
[4096, 131072]
output_norm.weight
F32
F32
[4096]