Models
GitHub
Discord
Turbo
Sign in
Download
Models
Download
GitHub
Discord
Sign in
Supa-AI
/
mixtral-8x7b-instruct-v0.1
:q5_k_m
296
Downloads
Updated
9 months ago
Quantized versions of the new mixtral 7b from Mistral! This is purely for research purposes and not for commercial distribution
Quantized versions of the new mixtral 7b from Mistral! This is purely for research purposes and not for commercial distribution
Cancel
mixtral-8x7b-instruct-v0.1:q5_k_m
...
/
model
8637455ad24c · 33GB
Metadata
general.architecture
llama
llama
general.file_type
Q5_K_M
Q5_K_M
llama.attention.head_count
32
32
llama.attention.head_count_kv
8
8
llama.attention.layer_norm_rms_epsilon
1e-05
1e-05
llama.block_count
32
32
llama.context_length
32768
32768
llama.embedding_length
4096
4096
llama.expert_count
8
8
llama.expert_used_count
2
2
llama.feed_forward_length
14336
14336
llama.rope.dimension_count
128
128
llama.rope.freq_base
1e+06
1e+06
llama.vocab_size
32000
32000
tokenizer.ggml.add_bos_token
true
true
tokenizer.ggml.add_eos_token
false
false
tokenizer.ggml.add_space_prefix
false
false
tokenizer.ggml.bos_token_id
1
1
tokenizer.ggml.eos_token_id
2
2
tokenizer.ggml.model
llama
llama
tokenizer.ggml.pre
default
default
tokenizer.ggml.scores
[-1000, -1000, -1000, 0, 0, ...]
[-1000, -1000, -1000, 0, 0, ...]
tokenizer.ggml.token_type
[3, 3, 3, 6, 6, ...]
[3, 3, 3, 6, 6, ...]
tokenizer.ggml.tokens
[<unk>, <s>, </s>, <0x00>, <0x01>, ...]
[<unk>, <s>, </s>, <0x00>, <0x01>, ...]
tokenizer.ggml.unknown_token_id
0
0
Tensor
Name
Type
Shape
token_embd.weight
Q5_K
Q5_K
[4096, 32000]
blk.0
blk.0.attn_k.weight
Q8_0
Q8_0
[4096, 1024]
blk.0.attn_norm.weight
F32
F32
[4096]
blk.0.attn_output.weight
Q5_K
Q5_K
[4096, 4096]
blk.0.attn_q.weight
Q5_K
Q5_K
[4096, 4096]
blk.0.attn_v.weight
Q8_0
Q8_0
[4096, 1024]
blk.0.ffn_down_exps.weight
Q6_K
Q6_K
[14336, 4096, 8]
blk.0.ffn_gate_exps.weight
Q5_K
Q5_K
[4096, 14336, 8]
blk.0.ffn_gate_inp.weight
F32
F32
[4096, 8]
blk.0.ffn_norm.weight
F32
F32
[4096]
blk.0.ffn_up_exps.weight
Q5_K
Q5_K
[4096, 14336, 8]
blk.1
blk.1.attn_k.weight
Q8_0
Q8_0
[4096, 1024]
blk.1.attn_norm.weight
F32
F32
[4096]
blk.1.attn_output.weight
Q5_K
Q5_K
[4096, 4096]
blk.1.attn_q.weight
Q5_K
Q5_K
[4096, 4096]
blk.1.attn_v.weight
Q8_0
Q8_0
[4096, 1024]
blk.1.ffn_down_exps.weight
Q6_K
Q6_K
[14336, 4096, 8]
blk.1.ffn_gate_exps.weight
Q5_K
Q5_K
[4096, 14336, 8]
blk.1.ffn_gate_inp.weight
F32
F32
[4096, 8]
blk.1.ffn_norm.weight
F32
F32
[4096]
blk.1.ffn_up_exps.weight
Q5_K
Q5_K
[4096, 14336, 8]
blk.2
blk.2.attn_k.weight
Q8_0
Q8_0
[4096, 1024]
blk.2.attn_norm.weight
F32
F32
[4096]
blk.2.attn_output.weight
Q5_K
Q5_K
[4096, 4096]
blk.2.attn_q.weight
Q5_K
Q5_K
[4096, 4096]
blk.2.attn_v.weight
Q8_0
Q8_0
[4096, 1024]
blk.2.ffn_down_exps.weight
Q6_K
Q6_K
[14336, 4096, 8]
blk.2.ffn_gate_exps.weight
Q5_K
Q5_K
[4096, 14336, 8]
blk.2.ffn_gate_inp.weight
F32
F32
[4096, 8]
blk.2.ffn_norm.weight
F32
F32
[4096]
blk.2.ffn_up_exps.weight
Q5_K
Q5_K
[4096, 14336, 8]
blk.3
blk.3.attn_k.weight
Q8_0
Q8_0
[4096, 1024]
blk.3.attn_norm.weight
F32
F32
[4096]
blk.3.attn_output.weight
Q5_K
Q5_K
[4096, 4096]
blk.3.attn_q.weight
Q5_K
Q5_K
[4096, 4096]
blk.3.attn_v.weight
Q8_0
Q8_0
[4096, 1024]
blk.3.ffn_down_exps.weight
Q6_K
Q6_K
[14336, 4096, 8]
blk.3.ffn_gate_exps.weight
Q5_K
Q5_K
[4096, 14336, 8]
blk.3.ffn_gate_inp.weight
F32
F32
[4096, 8]
blk.3.ffn_norm.weight
F32
F32
[4096]
blk.3.ffn_up_exps.weight
Q5_K
Q5_K
[4096, 14336, 8]
blk.4
blk.4.attn_k.weight
Q8_0
Q8_0
[4096, 1024]
blk.4.attn_norm.weight
F32
F32
[4096]
blk.4.attn_output.weight
Q5_K
Q5_K
[4096, 4096]
blk.4.attn_q.weight
Q5_K
Q5_K
[4096, 4096]
blk.4.attn_v.weight
Q8_0
Q8_0
[4096, 1024]
blk.4.ffn_down_exps.weight
Q5_K
Q5_K
[14336, 4096, 8]
blk.4.ffn_gate_exps.weight
Q5_K
Q5_K
[4096, 14336, 8]
blk.4.ffn_gate_inp.weight
F32
F32
[4096, 8]
blk.4.ffn_norm.weight
F32
F32
[4096]
blk.4.ffn_up_exps.weight
Q5_K
Q5_K
[4096, 14336, 8]
blk.5
blk.5.attn_k.weight
Q8_0
Q8_0
[4096, 1024]
blk.5.attn_norm.weight
F32
F32
[4096]
blk.5.attn_output.weight
Q5_K
Q5_K
[4096, 4096]
blk.5.attn_q.weight
Q5_K
Q5_K
[4096, 4096]
blk.5.attn_v.weight
Q8_0
Q8_0
[4096, 1024]
blk.5.ffn_down_exps.weight
Q5_K
Q5_K
[14336, 4096, 8]
blk.5.ffn_gate_exps.weight
Q5_K
Q5_K
[4096, 14336, 8]
blk.5.ffn_gate_inp.weight
F32
F32
[4096, 8]
blk.5.ffn_norm.weight
F32
F32
[4096]
blk.5.ffn_up_exps.weight
Q5_K
Q5_K
[4096, 14336, 8]
blk.6
blk.6.attn_k.weight
Q8_0
Q8_0
[4096, 1024]
blk.6.attn_norm.weight
F32
F32
[4096]
blk.6.attn_output.weight
Q5_K
Q5_K
[4096, 4096]
blk.6.attn_q.weight
Q5_K
Q5_K
[4096, 4096]
blk.6.attn_v.weight
Q8_0
Q8_0
[4096, 1024]
blk.6.ffn_down_exps.weight
Q6_K
Q6_K
[14336, 4096, 8]
blk.6.ffn_gate_exps.weight
Q5_K
Q5_K
[4096, 14336, 8]
blk.6.ffn_gate_inp.weight
F32
F32
[4096, 8]
blk.6.ffn_norm.weight
F32
F32
[4096]
blk.6.ffn_up_exps.weight
Q5_K
Q5_K
[4096, 14336, 8]
blk.7
blk.7.attn_k.weight
Q8_0
Q8_0
[4096, 1024]
blk.7.attn_norm.weight
F32
F32
[4096]
blk.7.attn_output.weight
Q5_K
Q5_K
[4096, 4096]
blk.7.attn_q.weight
Q5_K
Q5_K
[4096, 4096]
blk.7.attn_v.weight
Q8_0
Q8_0
[4096, 1024]
blk.7.ffn_down_exps.weight
Q5_K
Q5_K
[14336, 4096, 8]
blk.7.ffn_gate_exps.weight
Q5_K
Q5_K
[4096, 14336, 8]
blk.7.ffn_gate_inp.weight
F32
F32
[4096, 8]
blk.7.ffn_norm.weight
F32
F32
[4096]
blk.7.ffn_up_exps.weight
Q5_K
Q5_K
[4096, 14336, 8]
blk.8
blk.8.attn_k.weight
Q8_0
Q8_0
[4096, 1024]
blk.8.attn_norm.weight
F32
F32
[4096]
blk.8.attn_output.weight
Q5_K
Q5_K
[4096, 4096]
blk.8.attn_q.weight
Q5_K
Q5_K
[4096, 4096]
blk.8.attn_v.weight
Q8_0
Q8_0
[4096, 1024]
blk.8.ffn_down_exps.weight
Q5_K
Q5_K
[14336, 4096, 8]
blk.8.ffn_gate_exps.weight
Q5_K
Q5_K
[4096, 14336, 8]
blk.8.ffn_gate_inp.weight
F32
F32
[4096, 8]
blk.8.ffn_norm.weight
F32
F32
[4096]
blk.8.ffn_up_exps.weight
Q5_K
Q5_K
[4096, 14336, 8]
blk.9
blk.9.attn_k.weight
Q8_0
Q8_0
[4096, 1024]
blk.9.attn_norm.weight
F32
F32
[4096]
blk.9.attn_output.weight
Q5_K
Q5_K
[4096, 4096]
blk.9.attn_q.weight
Q5_K
Q5_K
[4096, 4096]
blk.9.attn_v.weight
Q8_0
Q8_0
[4096, 1024]
blk.9.ffn_down_exps.weight
Q6_K
Q6_K
[14336, 4096, 8]
blk.9.ffn_gate_exps.weight
Q5_K
Q5_K
[4096, 14336, 8]
blk.9.ffn_gate_inp.weight
F32
F32
[4096, 8]
blk.9.ffn_norm.weight
F32
F32
[4096]
blk.9.ffn_up_exps.weight
Q5_K
Q5_K
[4096, 14336, 8]
blk.10
blk.10.attn_k.weight
Q8_0
Q8_0
[4096, 1024]
blk.10.attn_norm.weight
F32
F32
[4096]
blk.10.attn_output.weight
Q5_K
Q5_K
[4096, 4096]
blk.10.attn_q.weight
Q5_K
Q5_K
[4096, 4096]
blk.10.attn_v.weight
Q8_0
Q8_0
[4096, 1024]
blk.10.ffn_down_exps.weight
Q5_K
Q5_K
[14336, 4096, 8]
blk.10.ffn_gate_exps.weight
Q5_K
Q5_K
[4096, 14336, 8]
blk.10.ffn_gate_inp.weight
F32
F32
[4096, 8]
blk.10.ffn_norm.weight
F32
F32
[4096]
blk.10.ffn_up_exps.weight
Q5_K
Q5_K
[4096, 14336, 8]
blk.11
blk.11.attn_k.weight
Q8_0
Q8_0
[4096, 1024]
blk.11.attn_norm.weight
F32
F32
[4096]
blk.11.attn_output.weight
Q5_K
Q5_K
[4096, 4096]
blk.11.attn_q.weight
Q5_K
Q5_K
[4096, 4096]
blk.11.attn_v.weight
Q8_0
Q8_0
[4096, 1024]
blk.11.ffn_down_exps.weight
Q5_K
Q5_K
[14336, 4096, 8]
blk.11.ffn_gate_exps.weight
Q5_K
Q5_K
[4096, 14336, 8]
blk.11.ffn_gate_inp.weight
F32
F32
[4096, 8]
blk.11.ffn_norm.weight
F32
F32
[4096]
blk.11.ffn_up_exps.weight
Q5_K
Q5_K
[4096, 14336, 8]
blk.12
blk.12.attn_k.weight
Q8_0
Q8_0
[4096, 1024]
blk.12.attn_norm.weight
F32
F32
[4096]
blk.12.attn_output.weight
Q5_K
Q5_K
[4096, 4096]
blk.12.attn_q.weight
Q5_K
Q5_K
[4096, 4096]
blk.12.attn_v.weight
Q8_0
Q8_0
[4096, 1024]
blk.12.ffn_down_exps.weight
Q6_K
Q6_K
[14336, 4096, 8]
blk.12.ffn_gate_exps.weight
Q5_K
Q5_K
[4096, 14336, 8]
blk.12.ffn_gate_inp.weight
F32
F32
[4096, 8]
blk.12.ffn_norm.weight
F32
F32
[4096]
blk.12.ffn_up_exps.weight
Q5_K
Q5_K
[4096, 14336, 8]
blk.13
blk.13.attn_k.weight
Q8_0
Q8_0
[4096, 1024]
blk.13.attn_norm.weight
F32
F32
[4096]
blk.13.attn_output.weight
Q5_K
Q5_K
[4096, 4096]
blk.13.attn_q.weight
Q5_K
Q5_K
[4096, 4096]
blk.13.attn_v.weight
Q8_0
Q8_0
[4096, 1024]
blk.13.ffn_down_exps.weight
Q5_K
Q5_K
[14336, 4096, 8]
blk.13.ffn_gate_exps.weight
Q5_K
Q5_K
[4096, 14336, 8]
blk.13.ffn_gate_inp.weight
F32
F32
[4096, 8]
blk.13.ffn_norm.weight
F32
F32
[4096]
blk.13.ffn_up_exps.weight
Q5_K
Q5_K
[4096, 14336, 8]
blk.14
blk.14.attn_k.weight
Q8_0
Q8_0
[4096, 1024]
blk.14.attn_norm.weight
F32
F32
[4096]
blk.14.attn_output.weight
Q5_K
Q5_K
[4096, 4096]
blk.14.attn_q.weight
Q5_K
Q5_K
[4096, 4096]
blk.14.attn_v.weight
Q8_0
Q8_0
[4096, 1024]
blk.14.ffn_down_exps.weight
Q5_K
Q5_K
[14336, 4096, 8]
blk.14.ffn_gate_exps.weight
Q5_K
Q5_K
[4096, 14336, 8]
blk.14.ffn_gate_inp.weight
F32
F32
[4096, 8]
blk.14.ffn_norm.weight
F32
F32
[4096]
blk.14.ffn_up_exps.weight
Q5_K
Q5_K
[4096, 14336, 8]
blk.15
blk.15.attn_k.weight
Q8_0
Q8_0
[4096, 1024]
blk.15.attn_norm.weight
F32
F32
[4096]
blk.15.attn_output.weight
Q5_K
Q5_K
[4096, 4096]
blk.15.attn_q.weight
Q5_K
Q5_K
[4096, 4096]
blk.15.attn_v.weight
Q8_0
Q8_0
[4096, 1024]
blk.15.ffn_down_exps.weight
Q6_K
Q6_K
[14336, 4096, 8]
blk.15.ffn_gate_exps.weight
Q5_K
Q5_K
[4096, 14336, 8]
blk.15.ffn_gate_inp.weight
F32
F32
[4096, 8]
blk.15.ffn_norm.weight
F32
F32
[4096]
blk.15.ffn_up_exps.weight
Q5_K
Q5_K
[4096, 14336, 8]
blk.16
blk.16.attn_k.weight
Q8_0
Q8_0
[4096, 1024]
blk.16.attn_norm.weight
F32
F32
[4096]
blk.16.attn_output.weight
Q5_K
Q5_K
[4096, 4096]
blk.16.attn_q.weight
Q5_K
Q5_K
[4096, 4096]
blk.16.attn_v.weight
Q8_0
Q8_0
[4096, 1024]
blk.16.ffn_down_exps.weight
Q5_K
Q5_K
[14336, 4096, 8]
blk.16.ffn_gate_exps.weight
Q5_K
Q5_K
[4096, 14336, 8]
blk.16.ffn_gate_inp.weight
F32
F32
[4096, 8]
blk.16.ffn_norm.weight
F32
F32
[4096]
blk.16.ffn_up_exps.weight
Q5_K
Q5_K
[4096, 14336, 8]
blk.17
blk.17.attn_k.weight
Q8_0
Q8_0
[4096, 1024]
blk.17.attn_norm.weight
F32
F32
[4096]
blk.17.attn_output.weight
Q5_K
Q5_K
[4096, 4096]
blk.17.attn_q.weight
Q5_K
Q5_K
[4096, 4096]
blk.17.attn_v.weight
Q8_0
Q8_0
[4096, 1024]
blk.17.ffn_down_exps.weight
Q5_K
Q5_K
[14336, 4096, 8]
blk.17.ffn_gate_exps.weight
Q5_K
Q5_K
[4096, 14336, 8]
blk.17.ffn_gate_inp.weight
F32
F32
[4096, 8]
blk.17.ffn_norm.weight
F32
F32
[4096]
blk.17.ffn_up_exps.weight
Q5_K
Q5_K
[4096, 14336, 8]
blk.18
blk.18.attn_k.weight
Q8_0
Q8_0
[4096, 1024]
blk.18.attn_norm.weight
F32
F32
[4096]
blk.18.attn_output.weight
Q5_K
Q5_K
[4096, 4096]
blk.18.attn_q.weight
Q5_K
Q5_K
[4096, 4096]
blk.18.attn_v.weight
Q8_0
Q8_0
[4096, 1024]
blk.18.ffn_down_exps.weight
Q6_K
Q6_K
[14336, 4096, 8]
blk.18.ffn_gate_exps.weight
Q5_K
Q5_K
[4096, 14336, 8]
blk.18.ffn_gate_inp.weight
F32
F32
[4096, 8]
blk.18.ffn_norm.weight
F32
F32
[4096]
blk.18.ffn_up_exps.weight
Q5_K
Q5_K
[4096, 14336, 8]
blk.19
blk.19.attn_k.weight
Q8_0
Q8_0
[4096, 1024]
blk.19.attn_norm.weight
F32
F32
[4096]
blk.19.attn_output.weight
Q5_K
Q5_K
[4096, 4096]
blk.19.attn_q.weight
Q5_K
Q5_K
[4096, 4096]
blk.19.attn_v.weight
Q8_0
Q8_0
[4096, 1024]
blk.19.ffn_down_exps.weight
Q5_K
Q5_K
[14336, 4096, 8]
blk.19.ffn_gate_exps.weight
Q5_K
Q5_K
[4096, 14336, 8]
blk.19.ffn_gate_inp.weight
F32
F32
[4096, 8]
blk.19.ffn_norm.weight
F32
F32
[4096]
blk.19.ffn_up_exps.weight
Q5_K
Q5_K
[4096, 14336, 8]
blk.20
blk.20.attn_k.weight
Q8_0
Q8_0
[4096, 1024]
blk.20.attn_norm.weight
F32
F32
[4096]
blk.20.attn_output.weight
Q5_K
Q5_K
[4096, 4096]
blk.20.attn_q.weight
Q5_K
Q5_K
[4096, 4096]
blk.20.attn_v.weight
Q8_0
Q8_0
[4096, 1024]
blk.20.ffn_down_exps.weight
Q5_K
Q5_K
[14336, 4096, 8]
blk.20.ffn_gate_exps.weight
Q5_K
Q5_K
[4096, 14336, 8]
blk.20.ffn_gate_inp.weight
F32
F32
[4096, 8]
blk.20.ffn_norm.weight
F32
F32
[4096]
blk.20.ffn_up_exps.weight
Q5_K
Q5_K
[4096, 14336, 8]
blk.21
blk.21.attn_k.weight
Q8_0
Q8_0
[4096, 1024]
blk.21.attn_norm.weight
F32
F32
[4096]
blk.21.attn_output.weight
Q5_K
Q5_K
[4096, 4096]
blk.21.attn_q.weight
Q5_K
Q5_K
[4096, 4096]
blk.21.attn_v.weight
Q8_0
Q8_0
[4096, 1024]
blk.21.ffn_down_exps.weight
Q6_K
Q6_K
[14336, 4096, 8]
blk.21.ffn_gate_exps.weight
Q5_K
Q5_K
[4096, 14336, 8]
blk.21.ffn_gate_inp.weight
F32
F32
[4096, 8]
blk.21.ffn_norm.weight
F32
F32
[4096]
blk.21.ffn_up_exps.weight
Q5_K
Q5_K
[4096, 14336, 8]
blk.22
blk.22.attn_k.weight
Q8_0
Q8_0
[4096, 1024]
blk.22.attn_norm.weight
F32
F32
[4096]
blk.22.attn_output.weight
Q5_K
Q5_K
[4096, 4096]
blk.22.attn_q.weight
Q5_K
Q5_K
[4096, 4096]
blk.22.attn_v.weight
Q8_0
Q8_0
[4096, 1024]
blk.22.ffn_down_exps.weight
Q5_K
Q5_K
[14336, 4096, 8]
blk.22.ffn_gate_exps.weight
Q5_K
Q5_K
[4096, 14336, 8]
blk.22.ffn_gate_inp.weight
F32
F32
[4096, 8]
blk.22.ffn_norm.weight
F32
F32
[4096]
blk.22.ffn_up_exps.weight
Q5_K
Q5_K
[4096, 14336, 8]
blk.23
blk.23.attn_k.weight
Q8_0
Q8_0
[4096, 1024]
blk.23.attn_norm.weight
F32
F32
[4096]
blk.23.attn_output.weight
Q5_K
Q5_K
[4096, 4096]
blk.23.attn_q.weight
Q5_K
Q5_K
[4096, 4096]
blk.23.attn_v.weight
Q8_0
Q8_0
[4096, 1024]
blk.23.ffn_down_exps.weight
Q5_K
Q5_K
[14336, 4096, 8]
blk.23.ffn_gate_exps.weight
Q5_K
Q5_K
[4096, 14336, 8]
blk.23.ffn_gate_inp.weight
F32
F32
[4096, 8]
blk.23.ffn_norm.weight
F32
F32
[4096]
blk.23.ffn_up_exps.weight
Q5_K
Q5_K
[4096, 14336, 8]
blk.24
blk.24.attn_k.weight
Q8_0
Q8_0
[4096, 1024]
blk.24.attn_norm.weight
F32
F32
[4096]
blk.24.attn_output.weight
Q5_K
Q5_K
[4096, 4096]
blk.24.attn_q.weight
Q5_K
Q5_K
[4096, 4096]
blk.24.attn_v.weight
Q8_0
Q8_0
[4096, 1024]
blk.24.ffn_down_exps.weight
Q6_K
Q6_K
[14336, 4096, 8]
blk.24.ffn_gate_exps.weight
Q5_K
Q5_K
[4096, 14336, 8]
blk.24.ffn_gate_inp.weight
F32
F32
[4096, 8]
blk.24.ffn_norm.weight
F32
F32
[4096]
blk.24.ffn_up_exps.weight
Q5_K
Q5_K
[4096, 14336, 8]
blk.25
blk.25.attn_k.weight
Q8_0
Q8_0
[4096, 1024]
blk.25.attn_norm.weight
F32
F32
[4096]
blk.25.attn_output.weight
Q5_K
Q5_K
[4096, 4096]
blk.25.attn_q.weight
Q5_K
Q5_K
[4096, 4096]
blk.25.attn_v.weight
Q8_0
Q8_0
[4096, 1024]
blk.25.ffn_down_exps.weight
Q5_K
Q5_K
[14336, 4096, 8]
blk.25.ffn_gate_exps.weight
Q5_K
Q5_K
[4096, 14336, 8]
blk.25.ffn_gate_inp.weight
F32
F32
[4096, 8]
blk.25.ffn_norm.weight
F32
F32
[4096]
blk.25.ffn_up_exps.weight
Q5_K
Q5_K
[4096, 14336, 8]
blk.26
blk.26.attn_k.weight
Q8_0
Q8_0
[4096, 1024]
blk.26.attn_norm.weight
F32
F32
[4096]
blk.26.attn_output.weight
Q5_K
Q5_K
[4096, 4096]
blk.26.attn_q.weight
Q5_K
Q5_K
[4096, 4096]
blk.26.attn_v.weight
Q8_0
Q8_0
[4096, 1024]
blk.26.ffn_down_exps.weight
Q5_K
Q5_K
[14336, 4096, 8]
blk.26.ffn_gate_exps.weight
Q5_K
Q5_K
[4096, 14336, 8]
blk.26.ffn_gate_inp.weight
F32
F32
[4096, 8]
blk.26.ffn_norm.weight
F32
F32
[4096]
blk.26.ffn_up_exps.weight
Q5_K
Q5_K
[4096, 14336, 8]
blk.27
blk.27.attn_k.weight
Q8_0
Q8_0
[4096, 1024]
blk.27.attn_norm.weight
F32
F32
[4096]
blk.27.attn_output.weight
Q5_K
Q5_K
[4096, 4096]
blk.27.attn_q.weight
Q5_K
Q5_K
[4096, 4096]
blk.27.attn_v.weight
Q8_0
Q8_0
[4096, 1024]
blk.27.ffn_down_exps.weight
Q6_K
Q6_K
[14336, 4096, 8]
blk.27.ffn_gate_exps.weight
Q5_K
Q5_K
[4096, 14336, 8]
blk.27.ffn_gate_inp.weight
F32
F32
[4096, 8]
blk.27.ffn_norm.weight
F32
F32
[4096]
blk.27.ffn_up_exps.weight
Q5_K
Q5_K
[4096, 14336, 8]
blk.28
blk.28.attn_k.weight
Q8_0
Q8_0
[4096, 1024]
blk.28.attn_norm.weight
F32
F32
[4096]
blk.28.attn_output.weight
Q5_K
Q5_K
[4096, 4096]
blk.28.attn_q.weight
Q5_K
Q5_K
[4096, 4096]
blk.28.attn_v.weight
Q8_0
Q8_0
[4096, 1024]
blk.28.ffn_down_exps.weight
Q6_K
Q6_K
[14336, 4096, 8]
blk.28.ffn_gate_exps.weight
Q5_K
Q5_K
[4096, 14336, 8]
blk.28.ffn_gate_inp.weight
F32
F32
[4096, 8]
blk.28.ffn_norm.weight
F32
F32
[4096]
blk.28.ffn_up_exps.weight
Q5_K
Q5_K
[4096, 14336, 8]
blk.29
blk.29.attn_k.weight
Q8_0
Q8_0
[4096, 1024]
blk.29.attn_norm.weight
F32
F32
[4096]
blk.29.attn_output.weight
Q5_K
Q5_K
[4096, 4096]
blk.29.attn_q.weight
Q5_K
Q5_K
[4096, 4096]
blk.29.attn_v.weight
Q8_0
Q8_0
[4096, 1024]
blk.29.ffn_down_exps.weight
Q6_K
Q6_K
[14336, 4096, 8]
blk.29.ffn_gate_exps.weight
Q5_K
Q5_K
[4096, 14336, 8]
blk.29.ffn_gate_inp.weight
F32
F32
[4096, 8]
blk.29.ffn_norm.weight
F32
F32
[4096]
blk.29.ffn_up_exps.weight
Q5_K
Q5_K
[4096, 14336, 8]
blk.30
blk.30.attn_k.weight
Q8_0
Q8_0
[4096, 1024]
blk.30.attn_norm.weight
F32
F32
[4096]
blk.30.attn_output.weight
Q5_K
Q5_K
[4096, 4096]
blk.30.attn_q.weight
Q5_K
Q5_K
[4096, 4096]
blk.30.attn_v.weight
Q8_0
Q8_0
[4096, 1024]
blk.30.ffn_down_exps.weight
Q6_K
Q6_K
[14336, 4096, 8]
blk.30.ffn_gate_exps.weight
Q5_K
Q5_K
[4096, 14336, 8]
blk.30.ffn_gate_inp.weight
F32
F32
[4096, 8]
blk.30.ffn_norm.weight
F32
F32
[4096]
blk.30.ffn_up_exps.weight
Q5_K
Q5_K
[4096, 14336, 8]
blk.31
blk.31.attn_k.weight
Q8_0
Q8_0
[4096, 1024]
blk.31.attn_norm.weight
F32
F32
[4096]
blk.31.attn_output.weight
Q5_K
Q5_K
[4096, 4096]
blk.31.attn_q.weight
Q5_K
Q5_K
[4096, 4096]
blk.31.attn_v.weight
Q8_0
Q8_0
[4096, 1024]
blk.31.ffn_down_exps.weight
Q6_K
Q6_K
[14336, 4096, 8]
blk.31.ffn_gate_exps.weight
Q5_K
Q5_K
[4096, 14336, 8]
blk.31.ffn_gate_inp.weight
F32
F32
[4096, 8]
blk.31.ffn_norm.weight
F32
F32
[4096]
blk.31.ffn_up_exps.weight
Q5_K
Q5_K
[4096, 14336, 8]
output.weight
Q6_K
Q6_K
[4096, 32000]
output_norm.weight
F32
F32
[4096]