Models
Docs
Pricing
Sign in
Download
Models
Download
Docs
Pricing
Sign in
ndavat
/
ndavat-Nanbeige4.2-3B
:latest
178
Downloads
Updated
2 weeks ago
Cancel
ndavat-Nanbeige4.2-3B:latest
...
/
model
9ffd17d14472 · 2.6GB
Metadata
general.architecture
nanbeige
nanbeige
general.file_type
Q4_K_M
Q4_K_M
nanbeige.attention.head_count
48
48
nanbeige.attention.head_count_kv
8
8
nanbeige.attention.head_dim
128
128
nanbeige.attention.key_length
128
128
nanbeige.attention.layer_norm_rms_epsilon
1e-05
1e-05
nanbeige.attention.value_length
128
128
nanbeige.block_count
22
22
nanbeige.context_length
262144
262144
nanbeige.embedding_length
3072
3072
nanbeige.feed_forward_length
10752
10752
nanbeige.num_loops
2
2
nanbeige.rope.dimension_count
128
128
nanbeige.rope.freq_base
7e+07
7e+07
nanbeige.skip_loop_final_norm
false
false
nanbeige.vocab_size
166144
166144
tokenizer.ggml.add_bos_token
true
true
tokenizer.ggml.add_eos_token
false
false
tokenizer.ggml.add_sep_token
false
false
tokenizer.ggml.add_space_prefix
true
true
tokenizer.ggml.bos_token_id
166100
166100
tokenizer.ggml.eos_token_id
166101
166101
tokenizer.ggml.model
llama
llama
tokenizer.ggml.padding_token_id
0
0
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
Q4_K
Q4_K
[3072, 166144]
blk.0
blk.0.attn_k.weight
Q4_K
Q4_K
[3072, 1024]
blk.0.attn_norm.weight
F32
F32
[3072]
blk.0.attn_output.weight
Q4_K
Q4_K
[6144, 3072]
blk.0.attn_q.weight
Q4_K
Q4_K
[3072, 6144]
blk.0.attn_v.weight
Q6_K
Q6_K
[3072, 1024]
blk.0.ffn_down.weight
Q6_K
Q6_K
[10752, 3072]
blk.0.ffn_gate.weight
Q4_K
Q4_K
[3072, 10752]
blk.0.ffn_norm.weight
F32
F32
[3072]
blk.0.ffn_up.weight
Q4_K
Q4_K
[3072, 10752]
blk.1
blk.1.attn_k.weight
Q4_K
Q4_K
[3072, 1024]
blk.1.attn_norm.weight
F32
F32
[3072]
blk.1.attn_output.weight
Q4_K
Q4_K
[6144, 3072]
blk.1.attn_q.weight
Q4_K
Q4_K
[3072, 6144]
blk.1.attn_v.weight
Q6_K
Q6_K
[3072, 1024]
blk.1.ffn_down.weight
Q6_K
Q6_K
[10752, 3072]
blk.1.ffn_gate.weight
Q4_K
Q4_K
[3072, 10752]
blk.1.ffn_norm.weight
F32
F32
[3072]
blk.1.ffn_up.weight
Q4_K
Q4_K
[3072, 10752]
blk.2
blk.2.attn_k.weight
Q4_K
Q4_K
[3072, 1024]
blk.2.attn_norm.weight
F32
F32
[3072]
blk.2.attn_output.weight
Q4_K
Q4_K
[6144, 3072]
blk.2.attn_q.weight
Q4_K
Q4_K
[3072, 6144]
blk.2.attn_v.weight
Q4_K
Q4_K
[3072, 1024]
blk.2.ffn_down.weight
Q6_K
Q6_K
[10752, 3072]
blk.2.ffn_gate.weight
Q4_K
Q4_K
[3072, 10752]
blk.2.ffn_norm.weight
F32
F32
[3072]
blk.2.ffn_up.weight
Q4_K
Q4_K
[3072, 10752]
blk.3
blk.3.attn_k.weight
Q4_K
Q4_K
[3072, 1024]
blk.3.attn_norm.weight
F32
F32
[3072]
blk.3.attn_output.weight
Q4_K
Q4_K
[6144, 3072]
blk.3.attn_q.weight
Q4_K
Q4_K
[3072, 6144]
blk.3.attn_v.weight
Q4_K
Q4_K
[3072, 1024]
blk.3.ffn_down.weight
Q6_K
Q6_K
[10752, 3072]
blk.3.ffn_gate.weight
Q4_K
Q4_K
[3072, 10752]
blk.3.ffn_norm.weight
F32
F32
[3072]
blk.3.ffn_up.weight
Q4_K
Q4_K
[3072, 10752]
blk.4
blk.4.attn_k.weight
Q4_K
Q4_K
[3072, 1024]
blk.4.attn_norm.weight
F32
F32
[3072]
blk.4.attn_output.weight
Q4_K
Q4_K
[6144, 3072]
blk.4.attn_q.weight
Q4_K
Q4_K
[3072, 6144]
blk.4.attn_v.weight
Q6_K
Q6_K
[3072, 1024]
blk.4.ffn_down.weight
Q6_K
Q6_K
[10752, 3072]
blk.4.ffn_gate.weight
Q4_K
Q4_K
[3072, 10752]
blk.4.ffn_norm.weight
F32
F32
[3072]
blk.4.ffn_up.weight
Q4_K
Q4_K
[3072, 10752]
blk.5
blk.5.attn_k.weight
Q4_K
Q4_K
[3072, 1024]
blk.5.attn_norm.weight
F32
F32
[3072]
blk.5.attn_output.weight
Q4_K
Q4_K
[6144, 3072]
blk.5.attn_q.weight
Q4_K
Q4_K
[3072, 6144]
blk.5.attn_v.weight
Q4_K
Q4_K
[3072, 1024]
blk.5.ffn_down.weight
Q4_K
Q4_K
[10752, 3072]
blk.5.ffn_gate.weight
Q4_K
Q4_K
[3072, 10752]
blk.5.ffn_norm.weight
F32
F32
[3072]
blk.5.ffn_up.weight
Q4_K
Q4_K
[3072, 10752]
blk.6
blk.6.attn_k.weight
Q4_K
Q4_K
[3072, 1024]
blk.6.attn_norm.weight
F32
F32
[3072]
blk.6.attn_output.weight
Q4_K
Q4_K
[6144, 3072]
blk.6.attn_q.weight
Q4_K
Q4_K
[3072, 6144]
blk.6.attn_v.weight
Q4_K
Q4_K
[3072, 1024]
blk.6.ffn_down.weight
Q4_K
Q4_K
[10752, 3072]
blk.6.ffn_gate.weight
Q4_K
Q4_K
[3072, 10752]
blk.6.ffn_norm.weight
F32
F32
[3072]
blk.6.ffn_up.weight
Q4_K
Q4_K
[3072, 10752]
blk.7
blk.7.attn_k.weight
Q4_K
Q4_K
[3072, 1024]
blk.7.attn_norm.weight
F32
F32
[3072]
blk.7.attn_output.weight
Q4_K
Q4_K
[6144, 3072]
blk.7.attn_q.weight
Q4_K
Q4_K
[3072, 6144]
blk.7.attn_v.weight
Q6_K
Q6_K
[3072, 1024]
blk.7.ffn_down.weight
Q6_K
Q6_K
[10752, 3072]
blk.7.ffn_gate.weight
Q4_K
Q4_K
[3072, 10752]
blk.7.ffn_norm.weight
F32
F32
[3072]
blk.7.ffn_up.weight
Q4_K
Q4_K
[3072, 10752]
blk.8
blk.8.attn_k.weight
Q4_K
Q4_K
[3072, 1024]
blk.8.attn_norm.weight
F32
F32
[3072]
blk.8.attn_output.weight
Q4_K
Q4_K
[6144, 3072]
blk.8.attn_q.weight
Q4_K
Q4_K
[3072, 6144]
blk.8.attn_v.weight
Q4_K
Q4_K
[3072, 1024]
blk.8.ffn_down.weight
Q4_K
Q4_K
[10752, 3072]
blk.8.ffn_gate.weight
Q4_K
Q4_K
[3072, 10752]
blk.8.ffn_norm.weight
F32
F32
[3072]
blk.8.ffn_up.weight
Q4_K
Q4_K
[3072, 10752]
blk.9
blk.9.attn_k.weight
Q4_K
Q4_K
[3072, 1024]
blk.9.attn_norm.weight
F32
F32
[3072]
blk.9.attn_output.weight
Q4_K
Q4_K
[6144, 3072]
blk.9.attn_q.weight
Q4_K
Q4_K
[3072, 6144]
blk.9.attn_v.weight
Q4_K
Q4_K
[3072, 1024]
blk.9.ffn_down.weight
Q4_K
Q4_K
[10752, 3072]
blk.9.ffn_gate.weight
Q4_K
Q4_K
[3072, 10752]
blk.9.ffn_norm.weight
F32
F32
[3072]
blk.9.ffn_up.weight
Q4_K
Q4_K
[3072, 10752]
blk.10
blk.10.attn_k.weight
Q4_K
Q4_K
[3072, 1024]
blk.10.attn_norm.weight
F32
F32
[3072]
blk.10.attn_output.weight
Q4_K
Q4_K
[6144, 3072]
blk.10.attn_q.weight
Q4_K
Q4_K
[3072, 6144]
blk.10.attn_v.weight
Q6_K
Q6_K
[3072, 1024]
blk.10.ffn_down.weight
Q6_K
Q6_K
[10752, 3072]
blk.10.ffn_gate.weight
Q4_K
Q4_K
[3072, 10752]
blk.10.ffn_norm.weight
F32
F32
[3072]
blk.10.ffn_up.weight
Q4_K
Q4_K
[3072, 10752]
blk.11
blk.11.attn_k.weight
Q4_K
Q4_K
[3072, 1024]
blk.11.attn_norm.weight
F32
F32
[3072]
blk.11.attn_output.weight
Q4_K
Q4_K
[6144, 3072]
blk.11.attn_q.weight
Q4_K
Q4_K
[3072, 6144]
blk.11.attn_v.weight
Q4_K
Q4_K
[3072, 1024]
blk.11.ffn_down.weight
Q4_K
Q4_K
[10752, 3072]
blk.11.ffn_gate.weight
Q4_K
Q4_K
[3072, 10752]
blk.11.ffn_norm.weight
F32
F32
[3072]
blk.11.ffn_up.weight
Q4_K
Q4_K
[3072, 10752]
blk.12
blk.12.attn_k.weight
Q4_K
Q4_K
[3072, 1024]
blk.12.attn_norm.weight
F32
F32
[3072]
blk.12.attn_output.weight
Q4_K
Q4_K
[6144, 3072]
blk.12.attn_q.weight
Q4_K
Q4_K
[3072, 6144]
blk.12.attn_v.weight
Q4_K
Q4_K
[3072, 1024]
blk.12.ffn_down.weight
Q4_K
Q4_K
[10752, 3072]
blk.12.ffn_gate.weight
Q4_K
Q4_K
[3072, 10752]
blk.12.ffn_norm.weight
F32
F32
[3072]
blk.12.ffn_up.weight
Q4_K
Q4_K
[3072, 10752]
blk.13
blk.13.attn_k.weight
Q4_K
Q4_K
[3072, 1024]
blk.13.attn_norm.weight
F32
F32
[3072]
blk.13.attn_output.weight
Q4_K
Q4_K
[6144, 3072]
blk.13.attn_q.weight
Q4_K
Q4_K
[3072, 6144]
blk.13.attn_v.weight
Q6_K
Q6_K
[3072, 1024]
blk.13.ffn_down.weight
Q6_K
Q6_K
[10752, 3072]
blk.13.ffn_gate.weight
Q4_K
Q4_K
[3072, 10752]
blk.13.ffn_norm.weight
F32
F32
[3072]
blk.13.ffn_up.weight
Q4_K
Q4_K
[3072, 10752]
blk.14
blk.14.attn_k.weight
Q4_K
Q4_K
[3072, 1024]
blk.14.attn_norm.weight
F32
F32
[3072]
blk.14.attn_output.weight
Q4_K
Q4_K
[6144, 3072]
blk.14.attn_q.weight
Q4_K
Q4_K
[3072, 6144]
blk.14.attn_v.weight
Q4_K
Q4_K
[3072, 1024]
blk.14.ffn_down.weight
Q4_K
Q4_K
[10752, 3072]
blk.14.ffn_gate.weight
Q4_K
Q4_K
[3072, 10752]
blk.14.ffn_norm.weight
F32
F32
[3072]
blk.14.ffn_up.weight
Q4_K
Q4_K
[3072, 10752]
blk.15
blk.15.attn_k.weight
Q4_K
Q4_K
[3072, 1024]
blk.15.attn_norm.weight
F32
F32
[3072]
blk.15.attn_output.weight
Q4_K
Q4_K
[6144, 3072]
blk.15.attn_q.weight
Q4_K
Q4_K
[3072, 6144]
blk.15.attn_v.weight
Q4_K
Q4_K
[3072, 1024]
blk.15.ffn_down.weight
Q4_K
Q4_K
[10752, 3072]
blk.15.ffn_gate.weight
Q4_K
Q4_K
[3072, 10752]
blk.15.ffn_norm.weight
F32
F32
[3072]
blk.15.ffn_up.weight
Q4_K
Q4_K
[3072, 10752]
blk.16
blk.16.attn_k.weight
Q4_K
Q4_K
[3072, 1024]
blk.16.attn_norm.weight
F32
F32
[3072]
blk.16.attn_output.weight
Q4_K
Q4_K
[6144, 3072]
blk.16.attn_q.weight
Q4_K
Q4_K
[3072, 6144]
blk.16.attn_v.weight
Q6_K
Q6_K
[3072, 1024]
blk.16.ffn_down.weight
Q6_K
Q6_K
[10752, 3072]
blk.16.ffn_gate.weight
Q4_K
Q4_K
[3072, 10752]
blk.16.ffn_norm.weight
F32
F32
[3072]
blk.16.ffn_up.weight
Q4_K
Q4_K
[3072, 10752]
blk.17
blk.17.attn_k.weight
Q4_K
Q4_K
[3072, 1024]
blk.17.attn_norm.weight
F32
F32
[3072]
blk.17.attn_output.weight
Q4_K
Q4_K
[6144, 3072]
blk.17.attn_q.weight
Q4_K
Q4_K
[3072, 6144]
blk.17.attn_v.weight
Q4_K
Q4_K
[3072, 1024]
blk.17.ffn_down.weight
Q4_K
Q4_K
[10752, 3072]
blk.17.ffn_gate.weight
Q4_K
Q4_K
[3072, 10752]
blk.17.ffn_norm.weight
F32
F32
[3072]
blk.17.ffn_up.weight
Q4_K
Q4_K
[3072, 10752]
blk.18
blk.18.attn_k.weight
Q4_K
Q4_K
[3072, 1024]
blk.18.attn_norm.weight
F32
F32
[3072]
blk.18.attn_output.weight
Q4_K
Q4_K
[6144, 3072]
blk.18.attn_q.weight
Q4_K
Q4_K
[3072, 6144]
blk.18.attn_v.weight
Q4_K
Q4_K
[3072, 1024]
blk.18.ffn_down.weight
Q4_K
Q4_K
[10752, 3072]
blk.18.ffn_gate.weight
Q4_K
Q4_K
[3072, 10752]
blk.18.ffn_norm.weight
F32
F32
[3072]
blk.18.ffn_up.weight
Q4_K
Q4_K
[3072, 10752]
blk.19
blk.19.attn_k.weight
Q4_K
Q4_K
[3072, 1024]
blk.19.attn_norm.weight
F32
F32
[3072]
blk.19.attn_output.weight
Q4_K
Q4_K
[6144, 3072]
blk.19.attn_q.weight
Q4_K
Q4_K
[3072, 6144]
blk.19.attn_v.weight
Q6_K
Q6_K
[3072, 1024]
blk.19.ffn_down.weight
Q6_K
Q6_K
[10752, 3072]
blk.19.ffn_gate.weight
Q4_K
Q4_K
[3072, 10752]
blk.19.ffn_norm.weight
F32
F32
[3072]
blk.19.ffn_up.weight
Q4_K
Q4_K
[3072, 10752]
blk.20
blk.20.attn_k.weight
Q4_K
Q4_K
[3072, 1024]
blk.20.attn_norm.weight
F32
F32
[3072]
blk.20.attn_output.weight
Q4_K
Q4_K
[6144, 3072]
blk.20.attn_q.weight
Q4_K
Q4_K
[3072, 6144]
blk.20.attn_v.weight
Q6_K
Q6_K
[3072, 1024]
blk.20.ffn_down.weight
Q4_K
Q4_K
[10752, 3072]
blk.20.ffn_gate.weight
Q4_K
Q4_K
[3072, 10752]
blk.20.ffn_norm.weight
F32
F32
[3072]
blk.20.ffn_up.weight
Q4_K
Q4_K
[3072, 10752]
blk.21
blk.21.attn_k.weight
Q4_K
Q4_K
[3072, 1024]
blk.21.attn_norm.weight
F32
F32
[3072]
blk.21.attn_output.weight
Q4_K
Q4_K
[6144, 3072]
blk.21.attn_q.weight
Q4_K
Q4_K
[3072, 6144]
blk.21.attn_v.weight
Q6_K
Q6_K
[3072, 1024]
blk.21.ffn_down.weight
Q4_K
Q4_K
[10752, 3072]
blk.21.ffn_gate.weight
Q4_K
Q4_K
[3072, 10752]
blk.21.ffn_norm.weight
F32
F32
[3072]
blk.21.ffn_up.weight
Q4_K
Q4_K
[3072, 10752]
output.weight
Q6_K
Q6_K
[3072, 166144]
output_norm.weight
F32
F32
[3072]