llama3.2-vision

Llama 3.2 Vision is a collection of instruction-tuned image reasoning generative models in 11B and 90B sizes.

vision 11b 90b

126.1K Pulls Updated 2 weeks ago

9 Tags

llama3.2-vision ... /

projector

ece5e659647a · 1.9GB

Metadata

general.architecture

mllama
general.description

vision encoder for Mllama
general.file_type

1
general.name

Llama-3.2-11B-Vision-Instruct
general.type

projector
mllama.vision.attention.head_count

16
mllama.vision.attention.layer_norm_epsilon

1e-05
mllama.vision.block_count

32
mllama.vision.embedding_length

1280
mllama.vision.feed_forward_length

5120
mllama.vision.global.block_count

8
mllama.vision.image_size

560
mllama.vision.intermediate_layers_indices

[3, 7, 15, 23, 30]
mllama.vision.max_num_tiles

4
mllama.vision.num_channels

3
mllama.vision.patch_size

14
mllama.vision.projection_dim

4096

Tensor

Name

Type

Shape
mm.0.bias

F32

[4096]
mm.0.weight

F16

[7680, 4096]

v.blk.0

v.blk.0.attn_k.weight

F16

[1280, 1280]
v.blk.0.attn_out.weight

F16

[1280, 1280]
v.blk.0.attn_q.weight

F16

[1280, 1280]
v.blk.0.attn_v.weight

F16

[1280, 1280]
v.blk.0.ffn_down.bias

F32

[5120]
v.blk.0.ffn_down.weight

F16

[1280, 5120]
v.blk.0.ffn_up.bias

F32

[1280]
v.blk.0.ffn_up.weight

F16

[5120, 1280]
v.blk.0.ln1.bias

F32

[1280]
v.blk.0.ln1.weight

F32

[1280]
v.blk.0.ln2.bias

F32

[1280]
v.blk.0.ln2.weight

F32

[1280]

v.blk.1

v.blk.1.attn_k.weight

F16

[1280, 1280]
v.blk.1.attn_out.weight

F16

[1280, 1280]
v.blk.1.attn_q.weight

F16

[1280, 1280]
v.blk.1.attn_v.weight

F16

[1280, 1280]
v.blk.1.ffn_down.bias

F32

[5120]
v.blk.1.ffn_down.weight

F16

[1280, 5120]
v.blk.1.ffn_up.bias

F32

[1280]
v.blk.1.ffn_up.weight

F16

[5120, 1280]
v.blk.1.ln1.bias

F32

[1280]
v.blk.1.ln1.weight

F32

[1280]
v.blk.1.ln2.bias

F32

[1280]
v.blk.1.ln2.weight

F32

[1280]

v.blk.2

v.blk.2.attn_k.weight

F16

[1280, 1280]
v.blk.2.attn_out.weight

F16

[1280, 1280]
v.blk.2.attn_q.weight

F16

[1280, 1280]
v.blk.2.attn_v.weight

F16

[1280, 1280]
v.blk.2.ffn_down.bias

F32

[5120]
v.blk.2.ffn_down.weight

F16

[1280, 5120]
v.blk.2.ffn_up.bias

F32

[1280]
v.blk.2.ffn_up.weight

F16

[5120, 1280]
v.blk.2.ln1.bias

F32

[1280]
v.blk.2.ln1.weight

F32

[1280]
v.blk.2.ln2.bias

F32

[1280]
v.blk.2.ln2.weight

F32

[1280]

v.blk.3

v.blk.3.attn_k.weight

F16

[1280, 1280]
v.blk.3.attn_out.weight

F16

[1280, 1280]
v.blk.3.attn_q.weight

F16

[1280, 1280]
v.blk.3.attn_v.weight

F16

[1280, 1280]
v.blk.3.ffn_down.bias

F32

[5120]
v.blk.3.ffn_down.weight

F16

[1280, 5120]
v.blk.3.ffn_up.bias

F32

[1280]
v.blk.3.ffn_up.weight

F16

[5120, 1280]
v.blk.3.ln1.bias

F32

[1280]
v.blk.3.ln1.weight

F32

[1280]
v.blk.3.ln2.bias

F32

[1280]
v.blk.3.ln2.weight

F32

[1280]

v.blk.4

v.blk.4.attn_k.weight

F16

[1280, 1280]
v.blk.4.attn_out.weight

F16

[1280, 1280]
v.blk.4.attn_q.weight

F16

[1280, 1280]
v.blk.4.attn_v.weight

F16

[1280, 1280]
v.blk.4.ffn_down.bias

F32

[5120]
v.blk.4.ffn_down.weight

F16

[1280, 5120]
v.blk.4.ffn_up.bias

F32

[1280]
v.blk.4.ffn_up.weight

F16

[5120, 1280]
v.blk.4.ln1.bias

F32

[1280]
v.blk.4.ln1.weight

F32

[1280]
v.blk.4.ln2.bias

F32

[1280]
v.blk.4.ln2.weight

F32

[1280]

v.blk.5

v.blk.5.attn_k.weight

F16

[1280, 1280]
v.blk.5.attn_out.weight

F16

[1280, 1280]
v.blk.5.attn_q.weight

F16

[1280, 1280]
v.blk.5.attn_v.weight

F16

[1280, 1280]
v.blk.5.ffn_down.bias

F32

[5120]
v.blk.5.ffn_down.weight

F16

[1280, 5120]
v.blk.5.ffn_up.bias

F32

[1280]
v.blk.5.ffn_up.weight

F16

[5120, 1280]
v.blk.5.ln1.bias

F32

[1280]
v.blk.5.ln1.weight

F32

[1280]
v.blk.5.ln2.bias

F32

[1280]
v.blk.5.ln2.weight

F32

[1280]

v.blk.6

v.blk.6.attn_k.weight

F16

[1280, 1280]
v.blk.6.attn_out.weight

F16

[1280, 1280]
v.blk.6.attn_q.weight

F16

[1280, 1280]
v.blk.6.attn_v.weight

F16

[1280, 1280]
v.blk.6.ffn_down.bias

F32

[5120]
v.blk.6.ffn_down.weight

F16

[1280, 5120]
v.blk.6.ffn_up.bias

F32

[1280]
v.blk.6.ffn_up.weight

F16

[5120, 1280]
v.blk.6.ln1.bias

F32

[1280]
v.blk.6.ln1.weight

F32

[1280]
v.blk.6.ln2.bias

F32

[1280]
v.blk.6.ln2.weight

F32

[1280]

v.blk.7

v.blk.7.attn_k.weight

F16

[1280, 1280]
v.blk.7.attn_out.weight

F16

[1280, 1280]
v.blk.7.attn_q.weight

F16

[1280, 1280]
v.blk.7.attn_v.weight

F16

[1280, 1280]
v.blk.7.ffn_down.bias

F32

[5120]
v.blk.7.ffn_down.weight

F16

[1280, 5120]
v.blk.7.ffn_up.bias

F32

[1280]
v.blk.7.ffn_up.weight

F16

[5120, 1280]
v.blk.7.ln1.bias

F32

[1280]
v.blk.7.ln1.weight

F32

[1280]
v.blk.7.ln2.bias

F32

[1280]
v.blk.7.ln2.weight

F32

[1280]

v.blk.8

v.blk.8.attn_k.weight

F16

[1280, 1280]
v.blk.8.attn_out.weight

F16

[1280, 1280]
v.blk.8.attn_q.weight

F16

[1280, 1280]
v.blk.8.attn_v.weight

F16

[1280, 1280]
v.blk.8.ffn_down.bias

F32

[5120]
v.blk.8.ffn_down.weight

F16

[1280, 5120]
v.blk.8.ffn_up.bias

F32

[1280]
v.blk.8.ffn_up.weight

F16

[5120, 1280]
v.blk.8.ln1.bias

F32

[1280]
v.blk.8.ln1.weight

F32

[1280]
v.blk.8.ln2.bias

F32

[1280]
v.blk.8.ln2.weight

F32

[1280]

v.blk.9

v.blk.9.attn_k.weight

F16

[1280, 1280]
v.blk.9.attn_out.weight

F16

[1280, 1280]
v.blk.9.attn_q.weight

F16

[1280, 1280]
v.blk.9.attn_v.weight

F16

[1280, 1280]
v.blk.9.ffn_down.bias

F32

[5120]
v.blk.9.ffn_down.weight

F16

[1280, 5120]
v.blk.9.ffn_up.bias

F32

[1280]
v.blk.9.ffn_up.weight

F16

[5120, 1280]
v.blk.9.ln1.bias

F32

[1280]
v.blk.9.ln1.weight

F32

[1280]
v.blk.9.ln2.bias

F32

[1280]
v.blk.9.ln2.weight

F32

[1280]

v.blk.10

v.blk.10.attn_k.weight

F16

[1280, 1280]
v.blk.10.attn_out.weight

F16

[1280, 1280]
v.blk.10.attn_q.weight

F16

[1280, 1280]
v.blk.10.attn_v.weight

F16

[1280, 1280]
v.blk.10.ffn_down.bias

F32

[5120]
v.blk.10.ffn_down.weight

F16

[1280, 5120]
v.blk.10.ffn_up.bias

F32

[1280]
v.blk.10.ffn_up.weight

F16

[5120, 1280]
v.blk.10.ln1.bias

F32

[1280]
v.blk.10.ln1.weight

F32

[1280]
v.blk.10.ln2.bias

F32

[1280]
v.blk.10.ln2.weight

F32

[1280]

v.blk.11

v.blk.11.attn_k.weight

F16

[1280, 1280]
v.blk.11.attn_out.weight

F16

[1280, 1280]
v.blk.11.attn_q.weight

F16

[1280, 1280]
v.blk.11.attn_v.weight

F16

[1280, 1280]
v.blk.11.ffn_down.bias

F32

[5120]
v.blk.11.ffn_down.weight

F16

[1280, 5120]
v.blk.11.ffn_up.bias

F32

[1280]
v.blk.11.ffn_up.weight

F16

[5120, 1280]
v.blk.11.ln1.bias

F32

[1280]
v.blk.11.ln1.weight

F32

[1280]
v.blk.11.ln2.bias

F32

[1280]
v.blk.11.ln2.weight

F32

[1280]

v.blk.12

v.blk.12.attn_k.weight

F16

[1280, 1280]
v.blk.12.attn_out.weight

F16

[1280, 1280]
v.blk.12.attn_q.weight

F16

[1280, 1280]
v.blk.12.attn_v.weight

F16

[1280, 1280]
v.blk.12.ffn_down.bias

F32

[5120]
v.blk.12.ffn_down.weight

F16

[1280, 5120]
v.blk.12.ffn_up.bias

F32

[1280]
v.blk.12.ffn_up.weight

F16

[5120, 1280]
v.blk.12.ln1.bias

F32

[1280]
v.blk.12.ln1.weight

F32

[1280]
v.blk.12.ln2.bias

F32

[1280]
v.blk.12.ln2.weight

F32

[1280]

v.blk.13

v.blk.13.attn_k.weight

F16

[1280, 1280]
v.blk.13.attn_out.weight

F16

[1280, 1280]
v.blk.13.attn_q.weight

F16

[1280, 1280]
v.blk.13.attn_v.weight

F16

[1280, 1280]
v.blk.13.ffn_down.bias

F32

[5120]
v.blk.13.ffn_down.weight

F16

[1280, 5120]
v.blk.13.ffn_up.bias

F32

[1280]
v.blk.13.ffn_up.weight

F16

[5120, 1280]
v.blk.13.ln1.bias

F32

[1280]
v.blk.13.ln1.weight

F32

[1280]
v.blk.13.ln2.bias

F32

[1280]
v.blk.13.ln2.weight

F32

[1280]

v.blk.14

v.blk.14.attn_k.weight

F16

[1280, 1280]
v.blk.14.attn_out.weight

F16

[1280, 1280]
v.blk.14.attn_q.weight

F16

[1280, 1280]
v.blk.14.attn_v.weight

F16

[1280, 1280]
v.blk.14.ffn_down.bias

F32

[5120]
v.blk.14.ffn_down.weight

F16

[1280, 5120]
v.blk.14.ffn_up.bias

F32

[1280]
v.blk.14.ffn_up.weight

F16

[5120, 1280]
v.blk.14.ln1.bias

F32

[1280]
v.blk.14.ln1.weight

F32

[1280]
v.blk.14.ln2.bias

F32

[1280]
v.blk.14.ln2.weight

F32

[1280]

v.blk.15

v.blk.15.attn_k.weight

F16

[1280, 1280]
v.blk.15.attn_out.weight

F16

[1280, 1280]
v.blk.15.attn_q.weight

F16

[1280, 1280]
v.blk.15.attn_v.weight

F16

[1280, 1280]
v.blk.15.ffn_down.bias

F32

[5120]
v.blk.15.ffn_down.weight

F16

[1280, 5120]
v.blk.15.ffn_up.bias

F32

[1280]
v.blk.15.ffn_up.weight

F16

[5120, 1280]
v.blk.15.ln1.bias

F32

[1280]
v.blk.15.ln1.weight

F32

[1280]
v.blk.15.ln2.bias

F32

[1280]
v.blk.15.ln2.weight

F32

[1280]

v.blk.16

v.blk.16.attn_k.weight

F16

[1280, 1280]
v.blk.16.attn_out.weight

F16

[1280, 1280]
v.blk.16.attn_q.weight

F16

[1280, 1280]
v.blk.16.attn_v.weight

F16

[1280, 1280]
v.blk.16.ffn_down.bias

F32

[5120]
v.blk.16.ffn_down.weight

F16

[1280, 5120]
v.blk.16.ffn_up.bias

F32

[1280]
v.blk.16.ffn_up.weight

F16

[5120, 1280]
v.blk.16.ln1.bias

F32

[1280]
v.blk.16.ln1.weight

F32

[1280]
v.blk.16.ln2.bias

F32

[1280]
v.blk.16.ln2.weight

F32

[1280]

v.blk.17

v.blk.17.attn_k.weight

F16

[1280, 1280]
v.blk.17.attn_out.weight

F16

[1280, 1280]
v.blk.17.attn_q.weight

F16

[1280, 1280]
v.blk.17.attn_v.weight

F16

[1280, 1280]
v.blk.17.ffn_down.bias

F32

[5120]
v.blk.17.ffn_down.weight

F16

[1280, 5120]
v.blk.17.ffn_up.bias

F32

[1280]
v.blk.17.ffn_up.weight

F16

[5120, 1280]
v.blk.17.ln1.bias

F32

[1280]
v.blk.17.ln1.weight

F32

[1280]
v.blk.17.ln2.bias

F32

[1280]
v.blk.17.ln2.weight

F32

[1280]

v.blk.18

v.blk.18.attn_k.weight

F16

[1280, 1280]
v.blk.18.attn_out.weight

F16

[1280, 1280]
v.blk.18.attn_q.weight

F16

[1280, 1280]
v.blk.18.attn_v.weight

F16

[1280, 1280]
v.blk.18.ffn_down.bias

F32

[5120]
v.blk.18.ffn_down.weight

F16

[1280, 5120]
v.blk.18.ffn_up.bias

F32

[1280]
v.blk.18.ffn_up.weight

F16

[5120, 1280]
v.blk.18.ln1.bias

F32

[1280]
v.blk.18.ln1.weight

F32

[1280]
v.blk.18.ln2.bias

F32

[1280]
v.blk.18.ln2.weight

F32

[1280]

v.blk.19

v.blk.19.attn_k.weight

F16

[1280, 1280]
v.blk.19.attn_out.weight

F16

[1280, 1280]
v.blk.19.attn_q.weight

F16

[1280, 1280]
v.blk.19.attn_v.weight

F16

[1280, 1280]
v.blk.19.ffn_down.bias

F32

[5120]
v.blk.19.ffn_down.weight

F16

[1280, 5120]
v.blk.19.ffn_up.bias

F32

[1280]
v.blk.19.ffn_up.weight

F16

[5120, 1280]
v.blk.19.ln1.bias

F32

[1280]
v.blk.19.ln1.weight

F32

[1280]
v.blk.19.ln2.bias

F32

[1280]
v.blk.19.ln2.weight

F32

[1280]

v.blk.20

v.blk.20.attn_k.weight

F16

[1280, 1280]
v.blk.20.attn_out.weight

F16

[1280, 1280]
v.blk.20.attn_q.weight

F16

[1280, 1280]
v.blk.20.attn_v.weight

F16

[1280, 1280]
v.blk.20.ffn_down.bias

F32

[5120]
v.blk.20.ffn_down.weight

F16

[1280, 5120]
v.blk.20.ffn_up.bias

F32

[1280]
v.blk.20.ffn_up.weight

F16

[5120, 1280]
v.blk.20.ln1.bias

F32

[1280]
v.blk.20.ln1.weight

F32

[1280]
v.blk.20.ln2.bias

F32

[1280]
v.blk.20.ln2.weight

F32

[1280]

v.blk.21

v.blk.21.attn_k.weight

F16

[1280, 1280]
v.blk.21.attn_out.weight

F16

[1280, 1280]
v.blk.21.attn_q.weight

F16

[1280, 1280]
v.blk.21.attn_v.weight

F16

[1280, 1280]
v.blk.21.ffn_down.bias

F32

[5120]
v.blk.21.ffn_down.weight

F16

[1280, 5120]
v.blk.21.ffn_up.bias

F32

[1280]
v.blk.21.ffn_up.weight

F16

[5120, 1280]
v.blk.21.ln1.bias

F32

[1280]
v.blk.21.ln1.weight

F32

[1280]
v.blk.21.ln2.bias

F32

[1280]
v.blk.21.ln2.weight

F32

[1280]

v.blk.22

v.blk.22.attn_k.weight

F16

[1280, 1280]
v.blk.22.attn_out.weight

F16

[1280, 1280]
v.blk.22.attn_q.weight

F16

[1280, 1280]
v.blk.22.attn_v.weight

F16

[1280, 1280]
v.blk.22.ffn_down.bias

F32

[5120]
v.blk.22.ffn_down.weight

F16

[1280, 5120]
v.blk.22.ffn_up.bias

F32

[1280]
v.blk.22.ffn_up.weight

F16

[5120, 1280]
v.blk.22.ln1.bias

F32

[1280]
v.blk.22.ln1.weight

F32

[1280]
v.blk.22.ln2.bias

F32

[1280]
v.blk.22.ln2.weight

F32

[1280]

v.blk.23

v.blk.23.attn_k.weight

F16

[1280, 1280]
v.blk.23.attn_out.weight

F16

[1280, 1280]
v.blk.23.attn_q.weight

F16

[1280, 1280]
v.blk.23.attn_v.weight

F16

[1280, 1280]
v.blk.23.ffn_down.bias

F32

[5120]
v.blk.23.ffn_down.weight

F16

[1280, 5120]
v.blk.23.ffn_up.bias

F32

[1280]
v.blk.23.ffn_up.weight

F16

[5120, 1280]
v.blk.23.ln1.bias

F32

[1280]
v.blk.23.ln1.weight

F32

[1280]
v.blk.23.ln2.bias

F32

[1280]
v.blk.23.ln2.weight

F32

[1280]

v.blk.24

v.blk.24.attn_k.weight

F16

[1280, 1280]
v.blk.24.attn_out.weight

F16

[1280, 1280]
v.blk.24.attn_q.weight

F16

[1280, 1280]
v.blk.24.attn_v.weight

F16

[1280, 1280]
v.blk.24.ffn_down.bias

F32

[5120]
v.blk.24.ffn_down.weight

F16

[1280, 5120]
v.blk.24.ffn_up.bias

F32

[1280]
v.blk.24.ffn_up.weight

F16

[5120, 1280]
v.blk.24.ln1.bias

F32

[1280]
v.blk.24.ln1.weight

F32

[1280]
v.blk.24.ln2.bias

F32

[1280]
v.blk.24.ln2.weight

F32

[1280]

v.blk.25

v.blk.25.attn_k.weight

F16

[1280, 1280]
v.blk.25.attn_out.weight

F16

[1280, 1280]
v.blk.25.attn_q.weight

F16

[1280, 1280]
v.blk.25.attn_v.weight

F16

[1280, 1280]
v.blk.25.ffn_down.bias

F32

[5120]
v.blk.25.ffn_down.weight

F16

[1280, 5120]
v.blk.25.ffn_up.bias

F32

[1280]
v.blk.25.ffn_up.weight

F16

[5120, 1280]
v.blk.25.ln1.bias

F32

[1280]
v.blk.25.ln1.weight

F32

[1280]
v.blk.25.ln2.bias

F32

[1280]
v.blk.25.ln2.weight

F32

[1280]

v.blk.26

v.blk.26.attn_k.weight

F16

[1280, 1280]
v.blk.26.attn_out.weight

F16

[1280, 1280]
v.blk.26.attn_q.weight

F16

[1280, 1280]
v.blk.26.attn_v.weight

F16

[1280, 1280]
v.blk.26.ffn_down.bias

F32

[5120]
v.blk.26.ffn_down.weight

F16

[1280, 5120]
v.blk.26.ffn_up.bias

F32

[1280]
v.blk.26.ffn_up.weight

F16

[5120, 1280]
v.blk.26.ln1.bias

F32

[1280]
v.blk.26.ln1.weight

F32

[1280]
v.blk.26.ln2.bias

F32

[1280]
v.blk.26.ln2.weight

F32

[1280]

v.blk.27

v.blk.27.attn_k.weight

F16

[1280, 1280]
v.blk.27.attn_out.weight

F16

[1280, 1280]
v.blk.27.attn_q.weight

F16

[1280, 1280]
v.blk.27.attn_v.weight

F16

[1280, 1280]
v.blk.27.ffn_down.bias

F32

[5120]
v.blk.27.ffn_down.weight

F16

[1280, 5120]
v.blk.27.ffn_up.bias

F32

[1280]
v.blk.27.ffn_up.weight

F16

[5120, 1280]
v.blk.27.ln1.bias

F32

[1280]
v.blk.27.ln1.weight

F32

[1280]
v.blk.27.ln2.bias

F32

[1280]
v.blk.27.ln2.weight

F32

[1280]

v.blk.28

v.blk.28.attn_k.weight

F16

[1280, 1280]
v.blk.28.attn_out.weight

F16

[1280, 1280]
v.blk.28.attn_q.weight

F16

[1280, 1280]
v.blk.28.attn_v.weight

F16

[1280, 1280]
v.blk.28.ffn_down.bias

F32

[5120]
v.blk.28.ffn_down.weight

F16

[1280, 5120]
v.blk.28.ffn_up.bias

F32

[1280]
v.blk.28.ffn_up.weight

F16

[5120, 1280]
v.blk.28.ln1.bias

F32

[1280]
v.blk.28.ln1.weight

F32

[1280]
v.blk.28.ln2.bias

F32

[1280]
v.blk.28.ln2.weight

F32

[1280]

v.blk.29

v.blk.29.attn_k.weight

F16

[1280, 1280]
v.blk.29.attn_out.weight

F16

[1280, 1280]
v.blk.29.attn_q.weight

F16

[1280, 1280]
v.blk.29.attn_v.weight

F16

[1280, 1280]
v.blk.29.ffn_down.bias

F32

[5120]
v.blk.29.ffn_down.weight

F16

[1280, 5120]
v.blk.29.ffn_up.bias

F32

[1280]
v.blk.29.ffn_up.weight

F16

[5120, 1280]
v.blk.29.ln1.bias

F32

[1280]
v.blk.29.ln1.weight

F32

[1280]
v.blk.29.ln2.bias

F32

[1280]
v.blk.29.ln2.weight

F32

[1280]

v.blk.30

v.blk.30.attn_k.weight

F16

[1280, 1280]
v.blk.30.attn_out.weight

F16

[1280, 1280]
v.blk.30.attn_q.weight

F16

[1280, 1280]
v.blk.30.attn_v.weight

F16

[1280, 1280]
v.blk.30.ffn_down.bias

F32

[5120]
v.blk.30.ffn_down.weight

F16

[1280, 5120]
v.blk.30.ffn_up.bias

F32

[1280]
v.blk.30.ffn_up.weight

F16

[5120, 1280]
v.blk.30.ln1.bias

F32

[1280]
v.blk.30.ln1.weight

F32

[1280]
v.blk.30.ln2.bias

F32

[1280]
v.blk.30.ln2.weight

F32

[1280]

v.blk.31

v.blk.31.attn_k.weight

F16

[1280, 1280]
v.blk.31.attn_out.weight

F16

[1280, 1280]
v.blk.31.attn_q.weight

F16

[1280, 1280]
v.blk.31.attn_v.weight

F16

[1280, 1280]
v.blk.31.ffn_down.bias

F32

[5120]
v.blk.31.ffn_down.weight

F16

[1280, 5120]
v.blk.31.ffn_up.bias

F32

[1280]
v.blk.31.ffn_up.weight

F16

[5120, 1280]
v.blk.31.ln1.bias

F32

[1280]
v.blk.31.ln1.weight

F32

[1280]
v.blk.31.ln2.bias

F32

[1280]
v.blk.31.ln2.weight

F32

[1280]

v.class_embd

F32

[1280]

v.global.blk.0

v.global.blk.0.attn_gate

F32

[1]
v.global.blk.0.attn_k.weight

F16

[1280, 1280]
v.global.blk.0.attn_out.weight

F16

[1280, 1280]
v.global.blk.0.attn_q.weight

F16

[1280, 1280]
v.global.blk.0.attn_v.weight

F16

[1280, 1280]
v.global.blk.0.ffn_down.bias

F32

[5120]
v.global.blk.0.ffn_down.weight

F16

[1280, 5120]
v.global.blk.0.ffn_gate

F32

[1]
v.global.blk.0.ffn_up.bias

F32

[1280]
v.global.blk.0.ffn_up.weight

F16

[5120, 1280]
v.global.blk.0.ln1.bias

F32

[1280]
v.global.blk.0.ln1.weight

F32

[1280]
v.global.blk.0.ln2.bias

F32

[1280]
v.global.blk.0.ln2.weight

F32

[1280]

v.global.blk.1

v.global.blk.1.attn_gate

F32

[1]
v.global.blk.1.attn_k.weight

F16

[1280, 1280]
v.global.blk.1.attn_out.weight

F16

[1280, 1280]
v.global.blk.1.attn_q.weight

F16

[1280, 1280]
v.global.blk.1.attn_v.weight

F16

[1280, 1280]
v.global.blk.1.ffn_down.bias

F32

[5120]
v.global.blk.1.ffn_down.weight

F16

[1280, 5120]
v.global.blk.1.ffn_gate

F32

[1]
v.global.blk.1.ffn_up.bias

F32

[1280]
v.global.blk.1.ffn_up.weight

F16

[5120, 1280]
v.global.blk.1.ln1.bias

F32

[1280]
v.global.blk.1.ln1.weight

F32

[1280]
v.global.blk.1.ln2.bias

F32

[1280]
v.global.blk.1.ln2.weight

F32

[1280]

v.global.blk.2

v.global.blk.2.attn_gate

F32

[1]
v.global.blk.2.attn_k.weight

F16

[1280, 1280]
v.global.blk.2.attn_out.weight

F16

[1280, 1280]
v.global.blk.2.attn_q.weight

F16

[1280, 1280]
v.global.blk.2.attn_v.weight

F16

[1280, 1280]
v.global.blk.2.ffn_down.bias

F32

[5120]
v.global.blk.2.ffn_down.weight

F16

[1280, 5120]
v.global.blk.2.ffn_gate

F32

[1]
v.global.blk.2.ffn_up.bias

F32

[1280]
v.global.blk.2.ffn_up.weight

F16

[5120, 1280]
v.global.blk.2.ln1.bias

F32

[1280]
v.global.blk.2.ln1.weight

F32

[1280]
v.global.blk.2.ln2.bias

F32

[1280]
v.global.blk.2.ln2.weight

F32

[1280]

v.global.blk.3

v.global.blk.3.attn_gate

F32

[1]
v.global.blk.3.attn_k.weight

F16

[1280, 1280]
v.global.blk.3.attn_out.weight

F16

[1280, 1280]
v.global.blk.3.attn_q.weight

F16

[1280, 1280]
v.global.blk.3.attn_v.weight

F16

[1280, 1280]
v.global.blk.3.ffn_down.bias

F32

[5120]
v.global.blk.3.ffn_down.weight

F16

[1280, 5120]
v.global.blk.3.ffn_gate

F32

[1]
v.global.blk.3.ffn_up.bias

F32

[1280]
v.global.blk.3.ffn_up.weight

F16

[5120, 1280]
v.global.blk.3.ln1.bias

F32

[1280]
v.global.blk.3.ln1.weight

F32

[1280]
v.global.blk.3.ln2.bias

F32

[1280]
v.global.blk.3.ln2.weight

F32

[1280]

v.global.blk.4

v.global.blk.4.attn_gate

F32

[1]
v.global.blk.4.attn_k.weight

F16

[1280, 1280]
v.global.blk.4.attn_out.weight

F16

[1280, 1280]
v.global.blk.4.attn_q.weight

F16

[1280, 1280]
v.global.blk.4.attn_v.weight

F16

[1280, 1280]
v.global.blk.4.ffn_down.bias

F32

[5120]
v.global.blk.4.ffn_down.weight

F16

[1280, 5120]
v.global.blk.4.ffn_gate

F32

[1]
v.global.blk.4.ffn_up.bias

F32

[1280]
v.global.blk.4.ffn_up.weight

F16

[5120, 1280]
v.global.blk.4.ln1.bias

F32

[1280]
v.global.blk.4.ln1.weight

F32

[1280]
v.global.blk.4.ln2.bias

F32

[1280]
v.global.blk.4.ln2.weight

F32

[1280]

v.global.blk.5

v.global.blk.5.attn_gate

F32

[1]
v.global.blk.5.attn_k.weight

F16

[1280, 1280]
v.global.blk.5.attn_out.weight

F16

[1280, 1280]
v.global.blk.5.attn_q.weight

F16

[1280, 1280]
v.global.blk.5.attn_v.weight

F16

[1280, 1280]
v.global.blk.5.ffn_down.bias

F32

[5120]
v.global.blk.5.ffn_down.weight

F16

[1280, 5120]
v.global.blk.5.ffn_gate

F32

[1]
v.global.blk.5.ffn_up.bias

F32

[1280]
v.global.blk.5.ffn_up.weight

F16

[5120, 1280]
v.global.blk.5.ln1.bias

F32

[1280]
v.global.blk.5.ln1.weight

F32

[1280]
v.global.blk.5.ln2.bias

F32

[1280]
v.global.blk.5.ln2.weight

F32

[1280]

v.global.blk.6

v.global.blk.6.attn_gate

F32

[1]
v.global.blk.6.attn_k.weight

F16

[1280, 1280]
v.global.blk.6.attn_out.weight

F16

[1280, 1280]
v.global.blk.6.attn_q.weight

F16

[1280, 1280]
v.global.blk.6.attn_v.weight

F16

[1280, 1280]
v.global.blk.6.ffn_down.bias

F32

[5120]
v.global.blk.6.ffn_down.weight

F16

[1280, 5120]
v.global.blk.6.ffn_gate

F32

[1]
v.global.blk.6.ffn_up.bias

F32

[1280]
v.global.blk.6.ffn_up.weight

F16

[5120, 1280]
v.global.blk.6.ln1.bias

F32

[1280]
v.global.blk.6.ln1.weight

F32

[1280]
v.global.blk.6.ln2.bias

F32

[1280]
v.global.blk.6.ln2.weight

F32

[1280]

v.global.blk.7

v.global.blk.7.attn_gate

F32

[1]
v.global.blk.7.attn_k.weight

F16

[1280, 1280]
v.global.blk.7.attn_out.weight

F16

[1280, 1280]
v.global.blk.7.attn_q.weight

F16

[1280, 1280]
v.global.blk.7.attn_v.weight

F16

[1280, 1280]
v.global.blk.7.ffn_down.bias

F32

[5120]
v.global.blk.7.ffn_down.weight

F16

[1280, 5120]
v.global.blk.7.ffn_gate

F32

[1]
v.global.blk.7.ffn_up.bias

F32

[1280]
v.global.blk.7.ffn_up.weight

F16

[5120, 1280]
v.global.blk.7.ln1.bias

F32

[1280]
v.global.blk.7.ln1.weight

F32

[1280]
v.global.blk.7.ln2.bias

F32

[1280]
v.global.blk.7.ln2.weight

F32

[1280]

v.patch_embd.weight

F16

[14, 14, 3, 1280]
v.position_embd.gate

F32

[1]
v.position_embd.weight

F16

[1280, 1601]
v.post_ln.bias

F32

[1280]
v.post_ln.weight

F32

[1280]
v.post_tile_position_embd.gate

F32

[1]
v.post_tile_position_embd.weight

F32

[5120, 9]
v.pre_ln.bias

F32

[1280]
v.pre_ln.weight

F32

[1280]
v.pre_tile_position_embd.gate

F32

[1]
v.pre_tile_position_embd.weight

F32

[5120, 9]
v.tile_position_embd.gate

F32

[1]
v.tile_position_embd.weight

F32

[8197120, 9]