Gminecrafter/KAi1.0/model

Gminecrafter/ KAi1.0:latest

7 Downloads Updated 2 months ago

KAi 1.0 is Kikico's first AI model, fully trained on a RTX 3080 Ti. Qwen3 Architecture, Trained from scratch. 63M parameters.

KAi1.0:latest ... /

model

df40b3a26ca5 · 127MB

Metadata

general.architecture

qwen3

qwen3
general.file_type

F16

F16
qwen3.attention.head_count

10

10
qwen3.attention.head_count_kv

2

2
qwen3.attention.key_length

64

64
qwen3.attention.layer_norm_rms_epsilon

1e-06

1e-06
qwen3.attention.value_length

64

64
qwen3.block_count

10

10
qwen3.context_length

1024

1024
qwen3.embedding_length

640

640
qwen3.feed_forward_length

1664

1664
qwen3.rope.freq_base

10000

10000
tokenizer.ggml.bos_token_id

0

0
tokenizer.ggml.eos_token_id

3

3
tokenizer.ggml.merges

[Ġ t, Ġ a, h e, i n, r e, ...]

[Ġ t, Ġ a, h e, i n, r e, ...]
tokenizer.ggml.model

gpt2

gpt2
tokenizer.ggml.padding_token_id

4

4
tokenizer.ggml.pre

llama-bpe

llama-bpe
tokenizer.ggml.token_type

[3, 3, 3, 3, 3, ...]

[3, 3, 3, 3, 3, ...]
tokenizer.ggml.tokens

[<|endoftext|>, <|user|>, <|assistant|>, <|end|>, <|pad|>, ...]

[<|endoftext|>, <|user|>, <|assistant|>, <|end|>, <|pad|>, ...]

Tensor

Name

Type

Shape
token_embd.weight

F16

F16

[640, 32768]

blk.0

blk.0.attn_k.weight

F16

F16

[640, 128]
blk.0.attn_k_norm.weight

F32

F32

[64]
blk.0.attn_norm.weight

F32

F32

[640]
blk.0.attn_output.weight

F16

F16

[640, 640]
blk.0.attn_q.weight

F16

F16

[640, 640]
blk.0.attn_q_norm.weight

F32

F32

[64]
blk.0.attn_v.weight

F16

F16

[640, 128]
blk.0.ffn_down.weight

F16

F16

[1664, 640]
blk.0.ffn_gate.weight

F16

F16

[640, 1664]
blk.0.ffn_norm.weight

F32

F32

[640]
blk.0.ffn_up.weight

F16

F16

[640, 1664]

blk.1

blk.1.attn_k.weight

F16

F16

[640, 128]
blk.1.attn_k_norm.weight

F32

F32

[64]
blk.1.attn_norm.weight

F32

F32

[640]
blk.1.attn_output.weight

F16

F16

[640, 640]
blk.1.attn_q.weight

F16

F16

[640, 640]
blk.1.attn_q_norm.weight

F32

F32

[64]
blk.1.attn_v.weight

F16

F16

[640, 128]
blk.1.ffn_down.weight

F16

F16

[1664, 640]
blk.1.ffn_gate.weight

F16

F16

[640, 1664]
blk.1.ffn_norm.weight

F32

F32

[640]
blk.1.ffn_up.weight

F16

F16

[640, 1664]

blk.2

blk.2.attn_k.weight

F16

F16

[640, 128]
blk.2.attn_k_norm.weight

F32

F32

[64]
blk.2.attn_norm.weight

F32

F32

[640]
blk.2.attn_output.weight

F16

F16

[640, 640]
blk.2.attn_q.weight

F16

F16

[640, 640]
blk.2.attn_q_norm.weight

F32

F32

[64]
blk.2.attn_v.weight

F16

F16

[640, 128]
blk.2.ffn_down.weight

F16

F16

[1664, 640]
blk.2.ffn_gate.weight

F16

F16

[640, 1664]
blk.2.ffn_norm.weight

F32

F32

[640]
blk.2.ffn_up.weight

F16

F16

[640, 1664]

blk.3

blk.3.attn_k.weight

F16

F16

[640, 128]
blk.3.attn_k_norm.weight

F32

F32

[64]
blk.3.attn_norm.weight

F32

F32

[640]
blk.3.attn_output.weight

F16

F16

[640, 640]
blk.3.attn_q.weight

F16

F16

[640, 640]
blk.3.attn_q_norm.weight

F32

F32

[64]
blk.3.attn_v.weight

F16

F16

[640, 128]
blk.3.ffn_down.weight

F16

F16

[1664, 640]
blk.3.ffn_gate.weight

F16

F16

[640, 1664]
blk.3.ffn_norm.weight

F32

F32

[640]
blk.3.ffn_up.weight

F16

F16

[640, 1664]

blk.4

blk.4.attn_k.weight

F16

F16

[640, 128]
blk.4.attn_k_norm.weight

F32

F32

[64]
blk.4.attn_norm.weight

F32

F32

[640]
blk.4.attn_output.weight

F16

F16

[640, 640]
blk.4.attn_q.weight

F16

F16

[640, 640]
blk.4.attn_q_norm.weight

F32

F32

[64]
blk.4.attn_v.weight

F16

F16

[640, 128]
blk.4.ffn_down.weight

F16

F16

[1664, 640]
blk.4.ffn_gate.weight

F16

F16

[640, 1664]
blk.4.ffn_norm.weight

F32

F32

[640]
blk.4.ffn_up.weight

F16

F16

[640, 1664]

blk.5

blk.5.attn_k.weight

F16

F16

[640, 128]
blk.5.attn_k_norm.weight

F32

F32

[64]
blk.5.attn_norm.weight

F32

F32

[640]
blk.5.attn_output.weight

F16

F16

[640, 640]
blk.5.attn_q.weight

F16

F16

[640, 640]
blk.5.attn_q_norm.weight

F32

F32

[64]
blk.5.attn_v.weight

F16

F16

[640, 128]
blk.5.ffn_down.weight

F16

F16

[1664, 640]
blk.5.ffn_gate.weight

F16

F16

[640, 1664]
blk.5.ffn_norm.weight

F32

F32

[640]
blk.5.ffn_up.weight

F16

F16

[640, 1664]

blk.6

blk.6.attn_k.weight

F16

F16

[640, 128]
blk.6.attn_k_norm.weight

F32

F32

[64]
blk.6.attn_norm.weight

F32

F32

[640]
blk.6.attn_output.weight

F16

F16

[640, 640]
blk.6.attn_q.weight

F16

F16

[640, 640]
blk.6.attn_q_norm.weight

F32

F32

[64]
blk.6.attn_v.weight

F16

F16

[640, 128]
blk.6.ffn_down.weight

F16

F16

[1664, 640]
blk.6.ffn_gate.weight

F16

F16

[640, 1664]
blk.6.ffn_norm.weight

F32

F32

[640]
blk.6.ffn_up.weight

F16

F16

[640, 1664]

blk.7

blk.7.attn_k.weight

F16

F16

[640, 128]
blk.7.attn_k_norm.weight

F32

F32

[64]
blk.7.attn_norm.weight

F32

F32

[640]
blk.7.attn_output.weight

F16

F16

[640, 640]
blk.7.attn_q.weight

F16

F16

[640, 640]
blk.7.attn_q_norm.weight

F32

F32

[64]
blk.7.attn_v.weight

F16

F16

[640, 128]
blk.7.ffn_down.weight

F16

F16

[1664, 640]
blk.7.ffn_gate.weight

F16

F16

[640, 1664]
blk.7.ffn_norm.weight

F32

F32

[640]
blk.7.ffn_up.weight

F16

F16

[640, 1664]

blk.8

blk.8.attn_k.weight

F16

F16

[640, 128]
blk.8.attn_k_norm.weight

F32

F32

[64]
blk.8.attn_norm.weight

F32

F32

[640]
blk.8.attn_output.weight

F16

F16

[640, 640]
blk.8.attn_q.weight

F16

F16

[640, 640]
blk.8.attn_q_norm.weight

F32

F32

[64]
blk.8.attn_v.weight

F16

F16

[640, 128]
blk.8.ffn_down.weight

F16

F16

[1664, 640]
blk.8.ffn_gate.weight

F16

F16

[640, 1664]
blk.8.ffn_norm.weight

F32

F32

[640]
blk.8.ffn_up.weight

F16

F16

[640, 1664]

blk.9

blk.9.attn_k.weight

F16

F16

[640, 128]
blk.9.attn_k_norm.weight

F32

F32

[64]
blk.9.attn_norm.weight

F32

F32

[640]
blk.9.attn_output.weight

F16

F16

[640, 640]
blk.9.attn_q.weight

F16

F16

[640, 640]
blk.9.attn_q_norm.weight

F32

F32

[64]
blk.9.attn_v.weight

F16

F16

[640, 128]
blk.9.ffn_down.weight

F16

F16

[1664, 640]
blk.9.ffn_gate.weight

F16

F16

[640, 1664]
blk.9.ffn_norm.weight

F32

F32

[640]
blk.9.ffn_up.weight

F16

F16

[640, 1664]

output_norm.weight

F32

F32

[640]