navthings/tale/model

navthings/ tale:latest

19 Downloads Updated 1 month ago

a small language model trained on tinystories.

tale:latest ... /

model

673ddb629d3f · 97MB

Metadata

general.architecture

llama

llama
general.file_type

F16

F16
llama.attention.head_count

6

6
llama.attention.head_count_kv

2

2
llama.attention.key_length

64

64
llama.attention.layer_norm_rms_epsilon

1e-05

1e-05
llama.attention.value_length

64

64
llama.block_count

13

13
llama.context_length

384

384
llama.embedding_length

384

384
llama.feed_forward_length

1536

1536
llama.rope.dimension_count

64

64
llama.rope.freq_base

10000

10000
llama.vocab_size

50257

50257
tokenizer.ggml.add_space_prefix

false

false
tokenizer.ggml.bos_token_id

50256

50256
tokenizer.ggml.eos_token_id

50256

50256
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

50256

50256
tokenizer.ggml.pre

gpt-2

gpt-2
tokenizer.ggml.token_type

[1, 1, 1, 1, 1, ...]

[1, 1, 1, 1, 1, ...]
tokenizer.ggml.tokens

[!, ", #, $, %, ...]

[!, ", #, $, %, ...]
tokenizer.ggml.unknown_token_id

50256

50256

Tensor

Name

Type

Shape
token_embd.weight

F16

F16

[384, 50257]

blk.0

blk.0.attn_k.weight

F16

F16

[384, 128]
blk.0.attn_norm.weight

F32

F32

[384]
blk.0.attn_output.weight

F16

F16

[384, 384]
blk.0.attn_q.weight

F16

F16

[384, 384]
blk.0.attn_v.weight

F16

F16

[384, 128]
blk.0.ffn_down.weight

F16

F16

[1536, 384]
blk.0.ffn_gate.weight

F16

F16

[384, 1536]
blk.0.ffn_norm.weight

F32

F32

[384]
blk.0.ffn_up.weight

F16

F16

[384, 1536]

blk.1

blk.1.attn_k.weight

F16

F16

[384, 128]
blk.1.attn_norm.weight

F32

F32

[384]
blk.1.attn_output.weight

F16

F16

[384, 384]
blk.1.attn_q.weight

F16

F16

[384, 384]
blk.1.attn_v.weight

F16

F16

[384, 128]
blk.1.ffn_down.weight

F16

F16

[1536, 384]
blk.1.ffn_gate.weight

F16

F16

[384, 1536]
blk.1.ffn_norm.weight

F32

F32

[384]
blk.1.ffn_up.weight

F16

F16

[384, 1536]

blk.2

blk.2.attn_k.weight

F16

F16

[384, 128]
blk.2.attn_norm.weight

F32

F32

[384]
blk.2.attn_output.weight

F16

F16

[384, 384]
blk.2.attn_q.weight

F16

F16

[384, 384]
blk.2.attn_v.weight

F16

F16

[384, 128]
blk.2.ffn_down.weight

F16

F16

[1536, 384]
blk.2.ffn_gate.weight

F16

F16

[384, 1536]
blk.2.ffn_norm.weight

F32

F32

[384]
blk.2.ffn_up.weight

F16

F16

[384, 1536]

blk.3

blk.3.attn_k.weight

F16

F16

[384, 128]
blk.3.attn_norm.weight

F32

F32

[384]
blk.3.attn_output.weight

F16

F16

[384, 384]
blk.3.attn_q.weight

F16

F16

[384, 384]
blk.3.attn_v.weight

F16

F16

[384, 128]
blk.3.ffn_down.weight

F16

F16

[1536, 384]
blk.3.ffn_gate.weight

F16

F16

[384, 1536]
blk.3.ffn_norm.weight

F32

F32

[384]
blk.3.ffn_up.weight

F16

F16

[384, 1536]

blk.4

blk.4.attn_k.weight

F16

F16

[384, 128]
blk.4.attn_norm.weight

F32

F32

[384]
blk.4.attn_output.weight

F16

F16

[384, 384]
blk.4.attn_q.weight

F16

F16

[384, 384]
blk.4.attn_v.weight

F16

F16

[384, 128]
blk.4.ffn_down.weight

F16

F16

[1536, 384]
blk.4.ffn_gate.weight

F16

F16

[384, 1536]
blk.4.ffn_norm.weight

F32

F32

[384]
blk.4.ffn_up.weight

F16

F16

[384, 1536]

blk.5

blk.5.attn_k.weight

F16

F16

[384, 128]
blk.5.attn_norm.weight

F32

F32

[384]
blk.5.attn_output.weight

F16

F16

[384, 384]
blk.5.attn_q.weight

F16

F16

[384, 384]
blk.5.attn_v.weight

F16

F16

[384, 128]
blk.5.ffn_down.weight

F16

F16

[1536, 384]
blk.5.ffn_gate.weight

F16

F16

[384, 1536]
blk.5.ffn_norm.weight

F32

F32

[384]
blk.5.ffn_up.weight

F16

F16

[384, 1536]

blk.6

blk.6.attn_k.weight

F16

F16

[384, 128]
blk.6.attn_norm.weight

F32

F32

[384]
blk.6.attn_output.weight

F16

F16

[384, 384]
blk.6.attn_q.weight

F16

F16

[384, 384]
blk.6.attn_v.weight

F16

F16

[384, 128]
blk.6.ffn_down.weight

F16

F16

[1536, 384]
blk.6.ffn_gate.weight

F16

F16

[384, 1536]
blk.6.ffn_norm.weight

F32

F32

[384]
blk.6.ffn_up.weight

F16

F16

[384, 1536]

blk.7

blk.7.attn_k.weight

F16

F16

[384, 128]
blk.7.attn_norm.weight

F32

F32

[384]
blk.7.attn_output.weight

F16

F16

[384, 384]
blk.7.attn_q.weight

F16

F16

[384, 384]
blk.7.attn_v.weight

F16

F16

[384, 128]
blk.7.ffn_down.weight

F16

F16

[1536, 384]
blk.7.ffn_gate.weight

F16

F16

[384, 1536]
blk.7.ffn_norm.weight

F32

F32

[384]
blk.7.ffn_up.weight

F16

F16

[384, 1536]

blk.8

blk.8.attn_k.weight

F16

F16

[384, 128]
blk.8.attn_norm.weight

F32

F32

[384]
blk.8.attn_output.weight

F16

F16

[384, 384]
blk.8.attn_q.weight

F16

F16

[384, 384]
blk.8.attn_v.weight

F16

F16

[384, 128]
blk.8.ffn_down.weight

F16

F16

[1536, 384]
blk.8.ffn_gate.weight

F16

F16

[384, 1536]
blk.8.ffn_norm.weight

F32

F32

[384]
blk.8.ffn_up.weight

F16

F16

[384, 1536]

blk.9

blk.9.attn_k.weight

F16

F16

[384, 128]
blk.9.attn_norm.weight

F32

F32

[384]
blk.9.attn_output.weight

F16

F16

[384, 384]
blk.9.attn_q.weight

F16

F16

[384, 384]
blk.9.attn_v.weight

F16

F16

[384, 128]
blk.9.ffn_down.weight

F16

F16

[1536, 384]
blk.9.ffn_gate.weight

F16

F16

[384, 1536]
blk.9.ffn_norm.weight

F32

F32

[384]
blk.9.ffn_up.weight

F16

F16

[384, 1536]

blk.10

blk.10.attn_k.weight

F16

F16

[384, 128]
blk.10.attn_norm.weight

F32

F32

[384]
blk.10.attn_output.weight

F16

F16

[384, 384]
blk.10.attn_q.weight

F16

F16

[384, 384]
blk.10.attn_v.weight

F16

F16

[384, 128]
blk.10.ffn_down.weight

F16

F16

[1536, 384]
blk.10.ffn_gate.weight

F16

F16

[384, 1536]
blk.10.ffn_norm.weight

F32

F32

[384]
blk.10.ffn_up.weight

F16

F16

[384, 1536]

blk.11

blk.11.attn_k.weight

F16

F16

[384, 128]
blk.11.attn_norm.weight

F32

F32

[384]
blk.11.attn_output.weight

F16

F16

[384, 384]
blk.11.attn_q.weight

F16

F16

[384, 384]
blk.11.attn_v.weight

F16

F16

[384, 128]
blk.11.ffn_down.weight

F16

F16

[1536, 384]
blk.11.ffn_gate.weight

F16

F16

[384, 1536]
blk.11.ffn_norm.weight

F32

F32

[384]
blk.11.ffn_up.weight

F16

F16

[384, 1536]

blk.12

blk.12.attn_k.weight

F16

F16

[384, 128]
blk.12.attn_norm.weight

F32

F32

[384]
blk.12.attn_output.weight

F16

F16

[384, 384]
blk.12.attn_q.weight

F16

F16

[384, 384]
blk.12.attn_v.weight

F16

F16

[384, 128]
blk.12.ffn_down.weight

F16

F16

[1536, 384]
blk.12.ffn_gate.weight

F16

F16

[384, 1536]
blk.12.ffn_norm.weight

F32

F32

[384]
blk.12.ffn_up.weight

F16

F16

[384, 1536]

output_norm.weight

F32

F32

[384]