Models
GitHub
Discord
Docs
Cloud
Sign in
Download
Models
Download
GitHub
Discord
Docs
Cloud
Sign in
heavylildude
/
magnus-x
:latest
8
Downloads
Updated
1 month ago
**reserved**
**reserved**
Cancel
tools
magnus-x:latest
...
/
model
6343744a3788 · 5.0GB
Metadata
general.architecture
lfm2moe
lfm2moe
general.file_type
Q4_K_M
Q4_K_M
lfm2moe.attention.head_count
32
32
lfm2moe.attention.head_count_kv
[0, 0, 8, 0, 0, ...]
[0, 0, 8, 0, 0, ...]
lfm2moe.attention.layer_norm_rms_epsilon
1e-05
1e-05
lfm2moe.block_count
24
24
lfm2moe.context_length
128000
128000
lfm2moe.embedding_length
2048
2048
lfm2moe.expert_count
32
32
lfm2moe.expert_feed_forward_length
1792
1792
lfm2moe.expert_gating_func
2
2
lfm2moe.expert_used_count
4
4
lfm2moe.feed_forward_length
7168
7168
lfm2moe.leading_dense_block_count
2
2
lfm2moe.rope.freq_base
1e+06
1e+06
lfm2moe.shortconv.l_cache
3
3
lfm2moe.vocab_size
65536
65536
tokenizer.ggml.add_bos_token
true
true
tokenizer.ggml.add_eos_token
false
false
tokenizer.ggml.add_sep_token
false
false
tokenizer.ggml.bos_token_id
1
1
tokenizer.ggml.eos_token_id
7
7
tokenizer.ggml.merges
[Ċ Ċ, Ċ ĊĊ, ĊĊ Ċ, Ċ ĊĊĊ, ĊĊ ĊĊ, ...]
[Ċ Ċ, Ċ ĊĊ, ĊĊ Ċ, Ċ ĊĊĊ, ĊĊ ĊĊ, ...]
tokenizer.ggml.model
gpt2
gpt2
tokenizer.ggml.padding_token_id
0
0
tokenizer.ggml.pre
lfm2
lfm2
tokenizer.ggml.token_type
[3, 3, 3, 3, 3, ...]
[3, 3, 3, 3, 3, ...]
tokenizer.ggml.tokens
[<|pad|>, <|startoftext|>, <|endoftext|>, <|fim_pre|>, <|fim_mid|>, ...]
[<|pad|>, <|startoftext|>, <|endoftext|>, <|fim_pre|>, <|fim_mid|>, ...]
dummy.padding
[0, 0, 0]
[0, 0, 0]
Tensor
Name
Type
Shape
token_embd.weight
Q6_K
Q6_K
[2048, 65536]
blk.0
blk.0.attn_norm.weight
F32
F32
[2048]
blk.0.ffn_down.weight
Q6_K
Q6_K
[7168, 2048]
blk.0.ffn_gate.weight
Q4_K
Q4_K
[2048, 7168]
blk.0.ffn_norm.weight
F32
F32
[2048]
blk.0.ffn_up.weight
Q4_K
Q4_K
[2048, 7168]
blk.0.shortconv.conv.weight
F32
F32
[3, 2048]
blk.0.shortconv.in_proj.weight
Q4_K
Q4_K
[2048, 6144]
blk.0.shortconv.out_proj.weight
Q4_K
Q4_K
[2048, 2048]
blk.1
blk.1.attn_norm.weight
F32
F32
[2048]
blk.1.ffn_down.weight
Q6_K
Q6_K
[7168, 2048]
blk.1.ffn_gate.weight
Q4_K
Q4_K
[2048, 7168]
blk.1.ffn_norm.weight
F32
F32
[2048]
blk.1.ffn_up.weight
Q4_K
Q4_K
[2048, 7168]
blk.1.shortconv.conv.weight
F32
F32
[3, 2048]
blk.1.shortconv.in_proj.weight
Q4_K
Q4_K
[2048, 6144]
blk.1.shortconv.out_proj.weight
Q4_K
Q4_K
[2048, 2048]
blk.2
blk.2.attn_k.weight
Q4_K
Q4_K
[2048, 512]
blk.2.attn_k_norm.weight
F32
F32
[64]
blk.2.attn_norm.weight
F32
F32
[2048]
blk.2.attn_output.weight
Q4_K
Q4_K
[2048, 2048]
blk.2.attn_q.weight
Q4_K
Q4_K
[2048, 2048]
blk.2.attn_q_norm.weight
F32
F32
[64]
blk.2.attn_v.weight
Q4_K
Q4_K
[2048, 512]
blk.2.exp_probs_b.bias
F32
F32
[32]
blk.2.ffn_down_exps.weight
Q6_K
Q6_K
[1792, 2048, 32]
blk.2.ffn_gate_exps.weight
Q4_K
Q4_K
[2048, 1792, 32]
blk.2.ffn_gate_inp.weight
F32
F32
[2048, 32]
blk.2.ffn_norm.weight
F32
F32
[2048]
blk.2.ffn_up_exps.weight
Q4_K
Q4_K
[2048, 1792, 32]
blk.3
blk.3.attn_norm.weight
F32
F32
[2048]
blk.3.exp_probs_b.bias
F32
F32
[32]
blk.3.ffn_down_exps.weight
Q4_K
Q4_K
[1792, 2048, 32]
blk.3.ffn_gate_exps.weight
Q4_K
Q4_K
[2048, 1792, 32]
blk.3.ffn_gate_inp.weight
F32
F32
[2048, 32]
blk.3.ffn_norm.weight
F32
F32
[2048]
blk.3.ffn_up_exps.weight
Q4_K
Q4_K
[2048, 1792, 32]
blk.3.shortconv.conv.weight
F32
F32
[3, 2048]
blk.3.shortconv.in_proj.weight
Q4_K
Q4_K
[2048, 6144]
blk.3.shortconv.out_proj.weight
Q4_K
Q4_K
[2048, 2048]
blk.4
blk.4.attn_norm.weight
F32
F32
[2048]
blk.4.exp_probs_b.bias
F32
F32
[32]
blk.4.ffn_down_exps.weight
Q4_K
Q4_K
[1792, 2048, 32]
blk.4.ffn_gate_exps.weight
Q4_K
Q4_K
[2048, 1792, 32]
blk.4.ffn_gate_inp.weight
F32
F32
[2048, 32]
blk.4.ffn_norm.weight
F32
F32
[2048]
blk.4.ffn_up_exps.weight
Q4_K
Q4_K
[2048, 1792, 32]
blk.4.shortconv.conv.weight
F32
F32
[3, 2048]
blk.4.shortconv.in_proj.weight
Q4_K
Q4_K
[2048, 6144]
blk.4.shortconv.out_proj.weight
Q4_K
Q4_K
[2048, 2048]
blk.5
blk.5.attn_norm.weight
F32
F32
[2048]
blk.5.exp_probs_b.bias
F32
F32
[32]
blk.5.ffn_down_exps.weight
Q6_K
Q6_K
[1792, 2048, 32]
blk.5.ffn_gate_exps.weight
Q4_K
Q4_K
[2048, 1792, 32]
blk.5.ffn_gate_inp.weight
F32
F32
[2048, 32]
blk.5.ffn_norm.weight
F32
F32
[2048]
blk.5.ffn_up_exps.weight
Q4_K
Q4_K
[2048, 1792, 32]
blk.5.shortconv.conv.weight
F32
F32
[3, 2048]
blk.5.shortconv.in_proj.weight
Q4_K
Q4_K
[2048, 6144]
blk.5.shortconv.out_proj.weight
Q4_K
Q4_K
[2048, 2048]
blk.6
blk.6.attn_k.weight
Q4_K
Q4_K
[2048, 512]
blk.6.attn_k_norm.weight
F32
F32
[64]
blk.6.attn_norm.weight
F32
F32
[2048]
blk.6.attn_output.weight
Q4_K
Q4_K
[2048, 2048]
blk.6.attn_q.weight
Q4_K
Q4_K
[2048, 2048]
blk.6.attn_q_norm.weight
F32
F32
[64]
blk.6.attn_v.weight
Q4_K
Q4_K
[2048, 512]
blk.6.exp_probs_b.bias
F32
F32
[32]
blk.6.ffn_down_exps.weight
Q4_K
Q4_K
[1792, 2048, 32]
blk.6.ffn_gate_exps.weight
Q4_K
Q4_K
[2048, 1792, 32]
blk.6.ffn_gate_inp.weight
F32
F32
[2048, 32]
blk.6.ffn_norm.weight
F32
F32
[2048]
blk.6.ffn_up_exps.weight
Q4_K
Q4_K
[2048, 1792, 32]
blk.7
blk.7.attn_norm.weight
F32
F32
[2048]
blk.7.exp_probs_b.bias
F32
F32
[32]
blk.7.ffn_down_exps.weight
Q4_K
Q4_K
[1792, 2048, 32]
blk.7.ffn_gate_exps.weight
Q4_K
Q4_K
[2048, 1792, 32]
blk.7.ffn_gate_inp.weight
F32
F32
[2048, 32]
blk.7.ffn_norm.weight
F32
F32
[2048]
blk.7.ffn_up_exps.weight
Q4_K
Q4_K
[2048, 1792, 32]
blk.7.shortconv.conv.weight
F32
F32
[3, 2048]
blk.7.shortconv.in_proj.weight
Q4_K
Q4_K
[2048, 6144]
blk.7.shortconv.out_proj.weight
Q4_K
Q4_K
[2048, 2048]
blk.8
blk.8.attn_norm.weight
F32
F32
[2048]
blk.8.exp_probs_b.bias
F32
F32
[32]
blk.8.ffn_down_exps.weight
Q6_K
Q6_K
[1792, 2048, 32]
blk.8.ffn_gate_exps.weight
Q4_K
Q4_K
[2048, 1792, 32]
blk.8.ffn_gate_inp.weight
F32
F32
[2048, 32]
blk.8.ffn_norm.weight
F32
F32
[2048]
blk.8.ffn_up_exps.weight
Q4_K
Q4_K
[2048, 1792, 32]
blk.8.shortconv.conv.weight
F32
F32
[3, 2048]
blk.8.shortconv.in_proj.weight
Q4_K
Q4_K
[2048, 6144]
blk.8.shortconv.out_proj.weight
Q4_K
Q4_K
[2048, 2048]
blk.9
blk.9.attn_norm.weight
F32
F32
[2048]
blk.9.exp_probs_b.bias
F32
F32
[32]
blk.9.ffn_down_exps.weight
Q4_K
Q4_K
[1792, 2048, 32]
blk.9.ffn_gate_exps.weight
Q4_K
Q4_K
[2048, 1792, 32]
blk.9.ffn_gate_inp.weight
F32
F32
[2048, 32]
blk.9.ffn_norm.weight
F32
F32
[2048]
blk.9.ffn_up_exps.weight
Q4_K
Q4_K
[2048, 1792, 32]
blk.9.shortconv.conv.weight
F32
F32
[3, 2048]
blk.9.shortconv.in_proj.weight
Q4_K
Q4_K
[2048, 6144]
blk.9.shortconv.out_proj.weight
Q4_K
Q4_K
[2048, 2048]
blk.10
blk.10.attn_k.weight
Q4_K
Q4_K
[2048, 512]
blk.10.attn_k_norm.weight
F32
F32
[64]
blk.10.attn_norm.weight
F32
F32
[2048]
blk.10.attn_output.weight
Q4_K
Q4_K
[2048, 2048]
blk.10.attn_q.weight
Q4_K
Q4_K
[2048, 2048]
blk.10.attn_q_norm.weight
F32
F32
[64]
blk.10.attn_v.weight
Q6_K
Q6_K
[2048, 512]
blk.10.exp_probs_b.bias
F32
F32
[32]
blk.10.ffn_down_exps.weight
Q4_K
Q4_K
[1792, 2048, 32]
blk.10.ffn_gate_exps.weight
Q4_K
Q4_K
[2048, 1792, 32]
blk.10.ffn_gate_inp.weight
F32
F32
[2048, 32]
blk.10.ffn_norm.weight
F32
F32
[2048]
blk.10.ffn_up_exps.weight
Q4_K
Q4_K
[2048, 1792, 32]
blk.11
blk.11.attn_norm.weight
F32
F32
[2048]
blk.11.exp_probs_b.bias
F32
F32
[32]
blk.11.ffn_down_exps.weight
Q6_K
Q6_K
[1792, 2048, 32]
blk.11.ffn_gate_exps.weight
Q4_K
Q4_K
[2048, 1792, 32]
blk.11.ffn_gate_inp.weight
F32
F32
[2048, 32]
blk.11.ffn_norm.weight
F32
F32
[2048]
blk.11.ffn_up_exps.weight
Q4_K
Q4_K
[2048, 1792, 32]
blk.11.shortconv.conv.weight
F32
F32
[3, 2048]
blk.11.shortconv.in_proj.weight
Q4_K
Q4_K
[2048, 6144]
blk.11.shortconv.out_proj.weight
Q4_K
Q4_K
[2048, 2048]
blk.12
blk.12.attn_norm.weight
F32
F32
[2048]
blk.12.exp_probs_b.bias
F32
F32
[32]
blk.12.ffn_down_exps.weight
Q4_K
Q4_K
[1792, 2048, 32]
blk.12.ffn_gate_exps.weight
Q4_K
Q4_K
[2048, 1792, 32]
blk.12.ffn_gate_inp.weight
F32
F32
[2048, 32]
blk.12.ffn_norm.weight
F32
F32
[2048]
blk.12.ffn_up_exps.weight
Q4_K
Q4_K
[2048, 1792, 32]
blk.12.shortconv.conv.weight
F32
F32
[3, 2048]
blk.12.shortconv.in_proj.weight
Q4_K
Q4_K
[2048, 6144]
blk.12.shortconv.out_proj.weight
Q4_K
Q4_K
[2048, 2048]
blk.13
blk.13.attn_norm.weight
F32
F32
[2048]
blk.13.exp_probs_b.bias
F32
F32
[32]
blk.13.ffn_down_exps.weight
Q4_K
Q4_K
[1792, 2048, 32]
blk.13.ffn_gate_exps.weight
Q4_K
Q4_K
[2048, 1792, 32]
blk.13.ffn_gate_inp.weight
F32
F32
[2048, 32]
blk.13.ffn_norm.weight
F32
F32
[2048]
blk.13.ffn_up_exps.weight
Q4_K
Q4_K
[2048, 1792, 32]
blk.13.shortconv.conv.weight
F32
F32
[3, 2048]
blk.13.shortconv.in_proj.weight
Q4_K
Q4_K
[2048, 6144]
blk.13.shortconv.out_proj.weight
Q4_K
Q4_K
[2048, 2048]
blk.14
blk.14.attn_k.weight
Q4_K
Q4_K
[2048, 512]
blk.14.attn_k_norm.weight
F32
F32
[64]
blk.14.attn_norm.weight
F32
F32
[2048]
blk.14.attn_output.weight
Q4_K
Q4_K
[2048, 2048]
blk.14.attn_q.weight
Q4_K
Q4_K
[2048, 2048]
blk.14.attn_q_norm.weight
F32
F32
[64]
blk.14.attn_v.weight
Q4_K
Q4_K
[2048, 512]
blk.14.exp_probs_b.bias
F32
F32
[32]
blk.14.ffn_down_exps.weight
Q6_K
Q6_K
[1792, 2048, 32]
blk.14.ffn_gate_exps.weight
Q4_K
Q4_K
[2048, 1792, 32]
blk.14.ffn_gate_inp.weight
F32
F32
[2048, 32]
blk.14.ffn_norm.weight
F32
F32
[2048]
blk.14.ffn_up_exps.weight
Q4_K
Q4_K
[2048, 1792, 32]
blk.15
blk.15.attn_norm.weight
F32
F32
[2048]
blk.15.exp_probs_b.bias
F32
F32
[32]
blk.15.ffn_down_exps.weight
Q4_K
Q4_K
[1792, 2048, 32]
blk.15.ffn_gate_exps.weight
Q4_K
Q4_K
[2048, 1792, 32]
blk.15.ffn_gate_inp.weight
F32
F32
[2048, 32]
blk.15.ffn_norm.weight
F32
F32
[2048]
blk.15.ffn_up_exps.weight
Q4_K
Q4_K
[2048, 1792, 32]
blk.15.shortconv.conv.weight
F32
F32
[3, 2048]
blk.15.shortconv.in_proj.weight
Q4_K
Q4_K
[2048, 6144]
blk.15.shortconv.out_proj.weight
Q4_K
Q4_K
[2048, 2048]
blk.16
blk.16.attn_norm.weight
F32
F32
[2048]
blk.16.exp_probs_b.bias
F32
F32
[32]
blk.16.ffn_down_exps.weight
Q4_K
Q4_K
[1792, 2048, 32]
blk.16.ffn_gate_exps.weight
Q4_K
Q4_K
[2048, 1792, 32]
blk.16.ffn_gate_inp.weight
F32
F32
[2048, 32]
blk.16.ffn_norm.weight
F32
F32
[2048]
blk.16.ffn_up_exps.weight
Q4_K
Q4_K
[2048, 1792, 32]
blk.16.shortconv.conv.weight
F32
F32
[3, 2048]
blk.16.shortconv.in_proj.weight
Q4_K
Q4_K
[2048, 6144]
blk.16.shortconv.out_proj.weight
Q4_K
Q4_K
[2048, 2048]
blk.17
blk.17.attn_norm.weight
F32
F32
[2048]
blk.17.exp_probs_b.bias
F32
F32
[32]
blk.17.ffn_down_exps.weight
Q6_K
Q6_K
[1792, 2048, 32]
blk.17.ffn_gate_exps.weight
Q4_K
Q4_K
[2048, 1792, 32]
blk.17.ffn_gate_inp.weight
F32
F32
[2048, 32]
blk.17.ffn_norm.weight
F32
F32
[2048]
blk.17.ffn_up_exps.weight
Q4_K
Q4_K
[2048, 1792, 32]
blk.17.shortconv.conv.weight
F32
F32
[3, 2048]
blk.17.shortconv.in_proj.weight
Q4_K
Q4_K
[2048, 6144]
blk.17.shortconv.out_proj.weight
Q4_K
Q4_K
[2048, 2048]
blk.18
blk.18.attn_k.weight
Q4_K
Q4_K
[2048, 512]
blk.18.attn_k_norm.weight
F32
F32
[64]
blk.18.attn_norm.weight
F32
F32
[2048]
blk.18.attn_output.weight
Q4_K
Q4_K
[2048, 2048]
blk.18.attn_q.weight
Q4_K
Q4_K
[2048, 2048]
blk.18.attn_q_norm.weight
F32
F32
[64]
blk.18.attn_v.weight
Q4_K
Q4_K
[2048, 512]
blk.18.exp_probs_b.bias
F32
F32
[32]
blk.18.ffn_down_exps.weight
Q4_K
Q4_K
[1792, 2048, 32]
blk.18.ffn_gate_exps.weight
Q4_K
Q4_K
[2048, 1792, 32]
blk.18.ffn_gate_inp.weight
F32
F32
[2048, 32]
blk.18.ffn_norm.weight
F32
F32
[2048]
blk.18.ffn_up_exps.weight
Q4_K
Q4_K
[2048, 1792, 32]
blk.19
blk.19.attn_norm.weight
F32
F32
[2048]
blk.19.exp_probs_b.bias
F32
F32
[32]
blk.19.ffn_down_exps.weight
Q4_K
Q4_K
[1792, 2048, 32]
blk.19.ffn_gate_exps.weight
Q4_K
Q4_K
[2048, 1792, 32]
blk.19.ffn_gate_inp.weight
F32
F32
[2048, 32]
blk.19.ffn_norm.weight
F32
F32
[2048]
blk.19.ffn_up_exps.weight
Q4_K
Q4_K
[2048, 1792, 32]
blk.19.shortconv.conv.weight
F32
F32
[3, 2048]
blk.19.shortconv.in_proj.weight
Q4_K
Q4_K
[2048, 6144]
blk.19.shortconv.out_proj.weight
Q4_K
Q4_K
[2048, 2048]
blk.20
blk.20.attn_norm.weight
F32
F32
[2048]
blk.20.exp_probs_b.bias
F32
F32
[32]
blk.20.ffn_down_exps.weight
Q6_K
Q6_K
[1792, 2048, 32]
blk.20.ffn_gate_exps.weight
Q4_K
Q4_K
[2048, 1792, 32]
blk.20.ffn_gate_inp.weight
F32
F32
[2048, 32]
blk.20.ffn_norm.weight
F32
F32
[2048]
blk.20.ffn_up_exps.weight
Q4_K
Q4_K
[2048, 1792, 32]
blk.20.shortconv.conv.weight
F32
F32
[3, 2048]
blk.20.shortconv.in_proj.weight
Q4_K
Q4_K
[2048, 6144]
blk.20.shortconv.out_proj.weight
Q4_K
Q4_K
[2048, 2048]
blk.21
blk.21.attn_k.weight
Q4_K
Q4_K
[2048, 512]
blk.21.attn_k_norm.weight
F32
F32
[64]
blk.21.attn_norm.weight
F32
F32
[2048]
blk.21.attn_output.weight
Q4_K
Q4_K
[2048, 2048]
blk.21.attn_q.weight
Q4_K
Q4_K
[2048, 2048]
blk.21.attn_q_norm.weight
F32
F32
[64]
blk.21.attn_v.weight
Q6_K
Q6_K
[2048, 512]
blk.21.exp_probs_b.bias
F32
F32
[32]
blk.21.ffn_down_exps.weight
Q6_K
Q6_K
[1792, 2048, 32]
blk.21.ffn_gate_exps.weight
Q4_K
Q4_K
[2048, 1792, 32]
blk.21.ffn_gate_inp.weight
F32
F32
[2048, 32]
blk.21.ffn_norm.weight
F32
F32
[2048]
blk.21.ffn_up_exps.weight
Q4_K
Q4_K
[2048, 1792, 32]
blk.22
blk.22.attn_norm.weight
F32
F32
[2048]
blk.22.exp_probs_b.bias
F32
F32
[32]
blk.22.ffn_down_exps.weight
Q6_K
Q6_K
[1792, 2048, 32]
blk.22.ffn_gate_exps.weight
Q4_K
Q4_K
[2048, 1792, 32]
blk.22.ffn_gate_inp.weight
F32
F32
[2048, 32]
blk.22.ffn_norm.weight
F32
F32
[2048]
blk.22.ffn_up_exps.weight
Q4_K
Q4_K
[2048, 1792, 32]
blk.22.shortconv.conv.weight
F32
F32
[3, 2048]
blk.22.shortconv.in_proj.weight
Q4_K
Q4_K
[2048, 6144]
blk.22.shortconv.out_proj.weight
Q4_K
Q4_K
[2048, 2048]
blk.23
blk.23.attn_norm.weight
F32
F32
[2048]
blk.23.exp_probs_b.bias
F32
F32
[32]
blk.23.ffn_down_exps.weight
Q6_K
Q6_K
[1792, 2048, 32]
blk.23.ffn_gate_exps.weight
Q4_K
Q4_K
[2048, 1792, 32]
blk.23.ffn_gate_inp.weight
F32
F32
[2048, 32]
blk.23.ffn_norm.weight
F32
F32
[2048]
blk.23.ffn_up_exps.weight
Q4_K
Q4_K
[2048, 1792, 32]
blk.23.shortconv.conv.weight
F32
F32
[3, 2048]
blk.23.shortconv.in_proj.weight
Q4_K
Q4_K
[2048, 6144]
blk.23.shortconv.out_proj.weight
Q4_K
Q4_K
[2048, 2048]
token_embd_norm.weight
F32
F32
[2048]