Models
Docs
Pricing
Sign in
Download
Models
Download
Docs
Pricing
Sign in
maternion
/
NuMarkdown-Thinking
:8b
281
Downloads
Updated
1 week ago
It is the first reasoning OCR VLM. It is specifically trained to convert documents into clean Markdown files, well suited for RAG applications.
It is the first reasoning OCR VLM. It is specifically trained to convert documents into clean Markdown files, well suited for RAG applications.
Cancel
vision
thinking
8b
NuMarkdown-Thinking:8b
...
/
projector
ce075861da43 · 1.4GB
Metadata
general.architecture
clip
clip
general.file_type
BF16
BF16
clip.has_vision_encoder
true
true
clip.projector_type
qwen2.5vl_merger
qwen2.5vl_merger
clip.use_silu
true
true
clip.vision.attention.head_count
16
16
clip.vision.attention.layer_norm_epsilon
1e-06
1e-06
clip.vision.block_count
32
32
clip.vision.embedding_length
1280
1280
clip.vision.feed_forward_length
3420
3420
clip.vision.image_mean
[0.48145467, 0.4578275, 0.40821072]
[0.48145467, 0.4578275, 0.40821072]
clip.vision.image_size
560
560
clip.vision.image_std
[0.26862955, 0.2613026, 0.2757771]
[0.26862955, 0.2613026, 0.2757771]
clip.vision.n_wa_pattern
8
8
clip.vision.patch_size
14
14
clip.vision.projection_dim
3584
3584
Tensor
Name
Type
Shape
mm.0.bias
F32
F32
[5120]
mm.0.weight
BF16
BF16
[5120, 5120]
mm.2.bias
F32
F32
[3584]
mm.2.weight
BF16
BF16
[5120, 3584]
v.blk.0
v.blk.0.attn_k.bias
F32
F32
[1280]
v.blk.0.attn_k.weight
BF16
BF16
[1280, 1280]
v.blk.0.attn_out.bias
F32
F32
[1280]
v.blk.0.attn_out.weight
BF16
BF16
[1280, 1280]
v.blk.0.attn_q.bias
F32
F32
[1280]
v.blk.0.attn_q.weight
BF16
BF16
[1280, 1280]
v.blk.0.attn_v.bias
F32
F32
[1280]
v.blk.0.attn_v.weight
BF16
BF16
[1280, 1280]
v.blk.0.ffn_down.bias
F32
F32
[1280]
v.blk.0.ffn_down.weight
BF16
BF16
[3420, 1280]
v.blk.0.ffn_gate.bias
F32
F32
[3420]
v.blk.0.ffn_gate.weight
BF16
BF16
[1280, 3420]
v.blk.0.ffn_up.bias
F32
F32
[3420]
v.blk.0.ffn_up.weight
BF16
BF16
[1280, 3420]
v.blk.0.ln1.weight
F32
F32
[1280]
v.blk.0.ln2.weight
F32
F32
[1280]
v.blk.1
v.blk.1.attn_k.bias
F32
F32
[1280]
v.blk.1.attn_k.weight
BF16
BF16
[1280, 1280]
v.blk.1.attn_out.bias
F32
F32
[1280]
v.blk.1.attn_out.weight
BF16
BF16
[1280, 1280]
v.blk.1.attn_q.bias
F32
F32
[1280]
v.blk.1.attn_q.weight
BF16
BF16
[1280, 1280]
v.blk.1.attn_v.bias
F32
F32
[1280]
v.blk.1.attn_v.weight
BF16
BF16
[1280, 1280]
v.blk.1.ffn_down.bias
F32
F32
[1280]
v.blk.1.ffn_down.weight
BF16
BF16
[3420, 1280]
v.blk.1.ffn_gate.bias
F32
F32
[3420]
v.blk.1.ffn_gate.weight
BF16
BF16
[1280, 3420]
v.blk.1.ffn_up.bias
F32
F32
[3420]
v.blk.1.ffn_up.weight
BF16
BF16
[1280, 3420]
v.blk.1.ln1.weight
F32
F32
[1280]
v.blk.1.ln2.weight
F32
F32
[1280]
v.blk.2
v.blk.2.attn_k.bias
F32
F32
[1280]
v.blk.2.attn_k.weight
BF16
BF16
[1280, 1280]
v.blk.2.attn_out.bias
F32
F32
[1280]
v.blk.2.attn_out.weight
BF16
BF16
[1280, 1280]
v.blk.2.attn_q.bias
F32
F32
[1280]
v.blk.2.attn_q.weight
BF16
BF16
[1280, 1280]
v.blk.2.attn_v.bias
F32
F32
[1280]
v.blk.2.attn_v.weight
BF16
BF16
[1280, 1280]
v.blk.2.ffn_down.bias
F32
F32
[1280]
v.blk.2.ffn_down.weight
BF16
BF16
[3420, 1280]
v.blk.2.ffn_gate.bias
F32
F32
[3420]
v.blk.2.ffn_gate.weight
BF16
BF16
[1280, 3420]
v.blk.2.ffn_up.bias
F32
F32
[3420]
v.blk.2.ffn_up.weight
BF16
BF16
[1280, 3420]
v.blk.2.ln1.weight
F32
F32
[1280]
v.blk.2.ln2.weight
F32
F32
[1280]
v.blk.3
v.blk.3.attn_k.bias
F32
F32
[1280]
v.blk.3.attn_k.weight
BF16
BF16
[1280, 1280]
v.blk.3.attn_out.bias
F32
F32
[1280]
v.blk.3.attn_out.weight
BF16
BF16
[1280, 1280]
v.blk.3.attn_q.bias
F32
F32
[1280]
v.blk.3.attn_q.weight
BF16
BF16
[1280, 1280]
v.blk.3.attn_v.bias
F32
F32
[1280]
v.blk.3.attn_v.weight
BF16
BF16
[1280, 1280]
v.blk.3.ffn_down.bias
F32
F32
[1280]
v.blk.3.ffn_down.weight
BF16
BF16
[3420, 1280]
v.blk.3.ffn_gate.bias
F32
F32
[3420]
v.blk.3.ffn_gate.weight
BF16
BF16
[1280, 3420]
v.blk.3.ffn_up.bias
F32
F32
[3420]
v.blk.3.ffn_up.weight
BF16
BF16
[1280, 3420]
v.blk.3.ln1.weight
F32
F32
[1280]
v.blk.3.ln2.weight
F32
F32
[1280]
v.blk.4
v.blk.4.attn_k.bias
F32
F32
[1280]
v.blk.4.attn_k.weight
BF16
BF16
[1280, 1280]
v.blk.4.attn_out.bias
F32
F32
[1280]
v.blk.4.attn_out.weight
BF16
BF16
[1280, 1280]
v.blk.4.attn_q.bias
F32
F32
[1280]
v.blk.4.attn_q.weight
BF16
BF16
[1280, 1280]
v.blk.4.attn_v.bias
F32
F32
[1280]
v.blk.4.attn_v.weight
BF16
BF16
[1280, 1280]
v.blk.4.ffn_down.bias
F32
F32
[1280]
v.blk.4.ffn_down.weight
BF16
BF16
[3420, 1280]
v.blk.4.ffn_gate.bias
F32
F32
[3420]
v.blk.4.ffn_gate.weight
BF16
BF16
[1280, 3420]
v.blk.4.ffn_up.bias
F32
F32
[3420]
v.blk.4.ffn_up.weight
BF16
BF16
[1280, 3420]
v.blk.4.ln1.weight
F32
F32
[1280]
v.blk.4.ln2.weight
F32
F32
[1280]
v.blk.5
v.blk.5.attn_k.bias
F32
F32
[1280]
v.blk.5.attn_k.weight
BF16
BF16
[1280, 1280]
v.blk.5.attn_out.bias
F32
F32
[1280]
v.blk.5.attn_out.weight
BF16
BF16
[1280, 1280]
v.blk.5.attn_q.bias
F32
F32
[1280]
v.blk.5.attn_q.weight
BF16
BF16
[1280, 1280]
v.blk.5.attn_v.bias
F32
F32
[1280]
v.blk.5.attn_v.weight
BF16
BF16
[1280, 1280]
v.blk.5.ffn_down.bias
F32
F32
[1280]
v.blk.5.ffn_down.weight
BF16
BF16
[3420, 1280]
v.blk.5.ffn_gate.bias
F32
F32
[3420]
v.blk.5.ffn_gate.weight
BF16
BF16
[1280, 3420]
v.blk.5.ffn_up.bias
F32
F32
[3420]
v.blk.5.ffn_up.weight
BF16
BF16
[1280, 3420]
v.blk.5.ln1.weight
F32
F32
[1280]
v.blk.5.ln2.weight
F32
F32
[1280]
v.blk.6
v.blk.6.attn_k.bias
F32
F32
[1280]
v.blk.6.attn_k.weight
BF16
BF16
[1280, 1280]
v.blk.6.attn_out.bias
F32
F32
[1280]
v.blk.6.attn_out.weight
BF16
BF16
[1280, 1280]
v.blk.6.attn_q.bias
F32
F32
[1280]
v.blk.6.attn_q.weight
BF16
BF16
[1280, 1280]
v.blk.6.attn_v.bias
F32
F32
[1280]
v.blk.6.attn_v.weight
BF16
BF16
[1280, 1280]
v.blk.6.ffn_down.bias
F32
F32
[1280]
v.blk.6.ffn_down.weight
BF16
BF16
[3420, 1280]
v.blk.6.ffn_gate.bias
F32
F32
[3420]
v.blk.6.ffn_gate.weight
BF16
BF16
[1280, 3420]
v.blk.6.ffn_up.bias
F32
F32
[3420]
v.blk.6.ffn_up.weight
BF16
BF16
[1280, 3420]
v.blk.6.ln1.weight
F32
F32
[1280]
v.blk.6.ln2.weight
F32
F32
[1280]
v.blk.7
v.blk.7.attn_k.bias
F32
F32
[1280]
v.blk.7.attn_k.weight
BF16
BF16
[1280, 1280]
v.blk.7.attn_out.bias
F32
F32
[1280]
v.blk.7.attn_out.weight
BF16
BF16
[1280, 1280]
v.blk.7.attn_q.bias
F32
F32
[1280]
v.blk.7.attn_q.weight
BF16
BF16
[1280, 1280]
v.blk.7.attn_v.bias
F32
F32
[1280]
v.blk.7.attn_v.weight
BF16
BF16
[1280, 1280]
v.blk.7.ffn_down.bias
F32
F32
[1280]
v.blk.7.ffn_down.weight
BF16
BF16
[3420, 1280]
v.blk.7.ffn_gate.bias
F32
F32
[3420]
v.blk.7.ffn_gate.weight
BF16
BF16
[1280, 3420]
v.blk.7.ffn_up.bias
F32
F32
[3420]
v.blk.7.ffn_up.weight
BF16
BF16
[1280, 3420]
v.blk.7.ln1.weight
F32
F32
[1280]
v.blk.7.ln2.weight
F32
F32
[1280]
v.blk.8
v.blk.8.attn_k.bias
F32
F32
[1280]
v.blk.8.attn_k.weight
BF16
BF16
[1280, 1280]
v.blk.8.attn_out.bias
F32
F32
[1280]
v.blk.8.attn_out.weight
BF16
BF16
[1280, 1280]
v.blk.8.attn_q.bias
F32
F32
[1280]
v.blk.8.attn_q.weight
BF16
BF16
[1280, 1280]
v.blk.8.attn_v.bias
F32
F32
[1280]
v.blk.8.attn_v.weight
BF16
BF16
[1280, 1280]
v.blk.8.ffn_down.bias
F32
F32
[1280]
v.blk.8.ffn_down.weight
BF16
BF16
[3420, 1280]
v.blk.8.ffn_gate.bias
F32
F32
[3420]
v.blk.8.ffn_gate.weight
BF16
BF16
[1280, 3420]
v.blk.8.ffn_up.bias
F32
F32
[3420]
v.blk.8.ffn_up.weight
BF16
BF16
[1280, 3420]
v.blk.8.ln1.weight
F32
F32
[1280]
v.blk.8.ln2.weight
F32
F32
[1280]
v.blk.9
v.blk.9.attn_k.bias
F32
F32
[1280]
v.blk.9.attn_k.weight
BF16
BF16
[1280, 1280]
v.blk.9.attn_out.bias
F32
F32
[1280]
v.blk.9.attn_out.weight
BF16
BF16
[1280, 1280]
v.blk.9.attn_q.bias
F32
F32
[1280]
v.blk.9.attn_q.weight
BF16
BF16
[1280, 1280]
v.blk.9.attn_v.bias
F32
F32
[1280]
v.blk.9.attn_v.weight
BF16
BF16
[1280, 1280]
v.blk.9.ffn_down.bias
F32
F32
[1280]
v.blk.9.ffn_down.weight
BF16
BF16
[3420, 1280]
v.blk.9.ffn_gate.bias
F32
F32
[3420]
v.blk.9.ffn_gate.weight
BF16
BF16
[1280, 3420]
v.blk.9.ffn_up.bias
F32
F32
[3420]
v.blk.9.ffn_up.weight
BF16
BF16
[1280, 3420]
v.blk.9.ln1.weight
F32
F32
[1280]
v.blk.9.ln2.weight
F32
F32
[1280]
v.blk.10
v.blk.10.attn_k.bias
F32
F32
[1280]
v.blk.10.attn_k.weight
BF16
BF16
[1280, 1280]
v.blk.10.attn_out.bias
F32
F32
[1280]
v.blk.10.attn_out.weight
BF16
BF16
[1280, 1280]
v.blk.10.attn_q.bias
F32
F32
[1280]
v.blk.10.attn_q.weight
BF16
BF16
[1280, 1280]
v.blk.10.attn_v.bias
F32
F32
[1280]
v.blk.10.attn_v.weight
BF16
BF16
[1280, 1280]
v.blk.10.ffn_down.bias
F32
F32
[1280]
v.blk.10.ffn_down.weight
BF16
BF16
[3420, 1280]
v.blk.10.ffn_gate.bias
F32
F32
[3420]
v.blk.10.ffn_gate.weight
BF16
BF16
[1280, 3420]
v.blk.10.ffn_up.bias
F32
F32
[3420]
v.blk.10.ffn_up.weight
BF16
BF16
[1280, 3420]
v.blk.10.ln1.weight
F32
F32
[1280]
v.blk.10.ln2.weight
F32
F32
[1280]
v.blk.11
v.blk.11.attn_k.bias
F32
F32
[1280]
v.blk.11.attn_k.weight
BF16
BF16
[1280, 1280]
v.blk.11.attn_out.bias
F32
F32
[1280]
v.blk.11.attn_out.weight
BF16
BF16
[1280, 1280]
v.blk.11.attn_q.bias
F32
F32
[1280]
v.blk.11.attn_q.weight
BF16
BF16
[1280, 1280]
v.blk.11.attn_v.bias
F32
F32
[1280]
v.blk.11.attn_v.weight
BF16
BF16
[1280, 1280]
v.blk.11.ffn_down.bias
F32
F32
[1280]
v.blk.11.ffn_down.weight
BF16
BF16
[3420, 1280]
v.blk.11.ffn_gate.bias
F32
F32
[3420]
v.blk.11.ffn_gate.weight
BF16
BF16
[1280, 3420]
v.blk.11.ffn_up.bias
F32
F32
[3420]
v.blk.11.ffn_up.weight
BF16
BF16
[1280, 3420]
v.blk.11.ln1.weight
F32
F32
[1280]
v.blk.11.ln2.weight
F32
F32
[1280]
v.blk.12
v.blk.12.attn_k.bias
F32
F32
[1280]
v.blk.12.attn_k.weight
BF16
BF16
[1280, 1280]
v.blk.12.attn_out.bias
F32
F32
[1280]
v.blk.12.attn_out.weight
BF16
BF16
[1280, 1280]
v.blk.12.attn_q.bias
F32
F32
[1280]
v.blk.12.attn_q.weight
BF16
BF16
[1280, 1280]
v.blk.12.attn_v.bias
F32
F32
[1280]
v.blk.12.attn_v.weight
BF16
BF16
[1280, 1280]
v.blk.12.ffn_down.bias
F32
F32
[1280]
v.blk.12.ffn_down.weight
BF16
BF16
[3420, 1280]
v.blk.12.ffn_gate.bias
F32
F32
[3420]
v.blk.12.ffn_gate.weight
BF16
BF16
[1280, 3420]
v.blk.12.ffn_up.bias
F32
F32
[3420]
v.blk.12.ffn_up.weight
BF16
BF16
[1280, 3420]
v.blk.12.ln1.weight
F32
F32
[1280]
v.blk.12.ln2.weight
F32
F32
[1280]
v.blk.13
v.blk.13.attn_k.bias
F32
F32
[1280]
v.blk.13.attn_k.weight
BF16
BF16
[1280, 1280]
v.blk.13.attn_out.bias
F32
F32
[1280]
v.blk.13.attn_out.weight
BF16
BF16
[1280, 1280]
v.blk.13.attn_q.bias
F32
F32
[1280]
v.blk.13.attn_q.weight
BF16
BF16
[1280, 1280]
v.blk.13.attn_v.bias
F32
F32
[1280]
v.blk.13.attn_v.weight
BF16
BF16
[1280, 1280]
v.blk.13.ffn_down.bias
F32
F32
[1280]
v.blk.13.ffn_down.weight
BF16
BF16
[3420, 1280]
v.blk.13.ffn_gate.bias
F32
F32
[3420]
v.blk.13.ffn_gate.weight
BF16
BF16
[1280, 3420]
v.blk.13.ffn_up.bias
F32
F32
[3420]
v.blk.13.ffn_up.weight
BF16
BF16
[1280, 3420]
v.blk.13.ln1.weight
F32
F32
[1280]
v.blk.13.ln2.weight
F32
F32
[1280]
v.blk.14
v.blk.14.attn_k.bias
F32
F32
[1280]
v.blk.14.attn_k.weight
BF16
BF16
[1280, 1280]
v.blk.14.attn_out.bias
F32
F32
[1280]
v.blk.14.attn_out.weight
BF16
BF16
[1280, 1280]
v.blk.14.attn_q.bias
F32
F32
[1280]
v.blk.14.attn_q.weight
BF16
BF16
[1280, 1280]
v.blk.14.attn_v.bias
F32
F32
[1280]
v.blk.14.attn_v.weight
BF16
BF16
[1280, 1280]
v.blk.14.ffn_down.bias
F32
F32
[1280]
v.blk.14.ffn_down.weight
BF16
BF16
[3420, 1280]
v.blk.14.ffn_gate.bias
F32
F32
[3420]
v.blk.14.ffn_gate.weight
BF16
BF16
[1280, 3420]
v.blk.14.ffn_up.bias
F32
F32
[3420]
v.blk.14.ffn_up.weight
BF16
BF16
[1280, 3420]
v.blk.14.ln1.weight
F32
F32
[1280]
v.blk.14.ln2.weight
F32
F32
[1280]
v.blk.15
v.blk.15.attn_k.bias
F32
F32
[1280]
v.blk.15.attn_k.weight
BF16
BF16
[1280, 1280]
v.blk.15.attn_out.bias
F32
F32
[1280]
v.blk.15.attn_out.weight
BF16
BF16
[1280, 1280]
v.blk.15.attn_q.bias
F32
F32
[1280]
v.blk.15.attn_q.weight
BF16
BF16
[1280, 1280]
v.blk.15.attn_v.bias
F32
F32
[1280]
v.blk.15.attn_v.weight
BF16
BF16
[1280, 1280]
v.blk.15.ffn_down.bias
F32
F32
[1280]
v.blk.15.ffn_down.weight
BF16
BF16
[3420, 1280]
v.blk.15.ffn_gate.bias
F32
F32
[3420]
v.blk.15.ffn_gate.weight
BF16
BF16
[1280, 3420]
v.blk.15.ffn_up.bias
F32
F32
[3420]
v.blk.15.ffn_up.weight
BF16
BF16
[1280, 3420]
v.blk.15.ln1.weight
F32
F32
[1280]
v.blk.15.ln2.weight
F32
F32
[1280]
v.blk.16
v.blk.16.attn_k.bias
F32
F32
[1280]
v.blk.16.attn_k.weight
BF16
BF16
[1280, 1280]
v.blk.16.attn_out.bias
F32
F32
[1280]
v.blk.16.attn_out.weight
BF16
BF16
[1280, 1280]
v.blk.16.attn_q.bias
F32
F32
[1280]
v.blk.16.attn_q.weight
BF16
BF16
[1280, 1280]
v.blk.16.attn_v.bias
F32
F32
[1280]
v.blk.16.attn_v.weight
BF16
BF16
[1280, 1280]
v.blk.16.ffn_down.bias
F32
F32
[1280]
v.blk.16.ffn_down.weight
BF16
BF16
[3420, 1280]
v.blk.16.ffn_gate.bias
F32
F32
[3420]
v.blk.16.ffn_gate.weight
BF16
BF16
[1280, 3420]
v.blk.16.ffn_up.bias
F32
F32
[3420]
v.blk.16.ffn_up.weight
BF16
BF16
[1280, 3420]
v.blk.16.ln1.weight
F32
F32
[1280]
v.blk.16.ln2.weight
F32
F32
[1280]
v.blk.17
v.blk.17.attn_k.bias
F32
F32
[1280]
v.blk.17.attn_k.weight
BF16
BF16
[1280, 1280]
v.blk.17.attn_out.bias
F32
F32
[1280]
v.blk.17.attn_out.weight
BF16
BF16
[1280, 1280]
v.blk.17.attn_q.bias
F32
F32
[1280]
v.blk.17.attn_q.weight
BF16
BF16
[1280, 1280]
v.blk.17.attn_v.bias
F32
F32
[1280]
v.blk.17.attn_v.weight
BF16
BF16
[1280, 1280]
v.blk.17.ffn_down.bias
F32
F32
[1280]
v.blk.17.ffn_down.weight
BF16
BF16
[3420, 1280]
v.blk.17.ffn_gate.bias
F32
F32
[3420]
v.blk.17.ffn_gate.weight
BF16
BF16
[1280, 3420]
v.blk.17.ffn_up.bias
F32
F32
[3420]
v.blk.17.ffn_up.weight
BF16
BF16
[1280, 3420]
v.blk.17.ln1.weight
F32
F32
[1280]
v.blk.17.ln2.weight
F32
F32
[1280]
v.blk.18
v.blk.18.attn_k.bias
F32
F32
[1280]
v.blk.18.attn_k.weight
BF16
BF16
[1280, 1280]
v.blk.18.attn_out.bias
F32
F32
[1280]
v.blk.18.attn_out.weight
BF16
BF16
[1280, 1280]
v.blk.18.attn_q.bias
F32
F32
[1280]
v.blk.18.attn_q.weight
BF16
BF16
[1280, 1280]
v.blk.18.attn_v.bias
F32
F32
[1280]
v.blk.18.attn_v.weight
BF16
BF16
[1280, 1280]
v.blk.18.ffn_down.bias
F32
F32
[1280]
v.blk.18.ffn_down.weight
BF16
BF16
[3420, 1280]
v.blk.18.ffn_gate.bias
F32
F32
[3420]
v.blk.18.ffn_gate.weight
BF16
BF16
[1280, 3420]
v.blk.18.ffn_up.bias
F32
F32
[3420]
v.blk.18.ffn_up.weight
BF16
BF16
[1280, 3420]
v.blk.18.ln1.weight
F32
F32
[1280]
v.blk.18.ln2.weight
F32
F32
[1280]
v.blk.19
v.blk.19.attn_k.bias
F32
F32
[1280]
v.blk.19.attn_k.weight
BF16
BF16
[1280, 1280]
v.blk.19.attn_out.bias
F32
F32
[1280]
v.blk.19.attn_out.weight
BF16
BF16
[1280, 1280]
v.blk.19.attn_q.bias
F32
F32
[1280]
v.blk.19.attn_q.weight
BF16
BF16
[1280, 1280]
v.blk.19.attn_v.bias
F32
F32
[1280]
v.blk.19.attn_v.weight
BF16
BF16
[1280, 1280]
v.blk.19.ffn_down.bias
F32
F32
[1280]
v.blk.19.ffn_down.weight
BF16
BF16
[3420, 1280]
v.blk.19.ffn_gate.bias
F32
F32
[3420]
v.blk.19.ffn_gate.weight
BF16
BF16
[1280, 3420]
v.blk.19.ffn_up.bias
F32
F32
[3420]
v.blk.19.ffn_up.weight
BF16
BF16
[1280, 3420]
v.blk.19.ln1.weight
F32
F32
[1280]
v.blk.19.ln2.weight
F32
F32
[1280]
v.blk.20
v.blk.20.attn_k.bias
F32
F32
[1280]
v.blk.20.attn_k.weight
BF16
BF16
[1280, 1280]
v.blk.20.attn_out.bias
F32
F32
[1280]
v.blk.20.attn_out.weight
BF16
BF16
[1280, 1280]
v.blk.20.attn_q.bias
F32
F32
[1280]
v.blk.20.attn_q.weight
BF16
BF16
[1280, 1280]
v.blk.20.attn_v.bias
F32
F32
[1280]
v.blk.20.attn_v.weight
BF16
BF16
[1280, 1280]
v.blk.20.ffn_down.bias
F32
F32
[1280]
v.blk.20.ffn_down.weight
BF16
BF16
[3420, 1280]
v.blk.20.ffn_gate.bias
F32
F32
[3420]
v.blk.20.ffn_gate.weight
BF16
BF16
[1280, 3420]
v.blk.20.ffn_up.bias
F32
F32
[3420]
v.blk.20.ffn_up.weight
BF16
BF16
[1280, 3420]
v.blk.20.ln1.weight
F32
F32
[1280]
v.blk.20.ln2.weight
F32
F32
[1280]
v.blk.21
v.blk.21.attn_k.bias
F32
F32
[1280]
v.blk.21.attn_k.weight
BF16
BF16
[1280, 1280]
v.blk.21.attn_out.bias
F32
F32
[1280]
v.blk.21.attn_out.weight
BF16
BF16
[1280, 1280]
v.blk.21.attn_q.bias
F32
F32
[1280]
v.blk.21.attn_q.weight
BF16
BF16
[1280, 1280]
v.blk.21.attn_v.bias
F32
F32
[1280]
v.blk.21.attn_v.weight
BF16
BF16
[1280, 1280]
v.blk.21.ffn_down.bias
F32
F32
[1280]
v.blk.21.ffn_down.weight
BF16
BF16
[3420, 1280]
v.blk.21.ffn_gate.bias
F32
F32
[3420]
v.blk.21.ffn_gate.weight
BF16
BF16
[1280, 3420]
v.blk.21.ffn_up.bias
F32
F32
[3420]
v.blk.21.ffn_up.weight
BF16
BF16
[1280, 3420]
v.blk.21.ln1.weight
F32
F32
[1280]
v.blk.21.ln2.weight
F32
F32
[1280]
v.blk.22
v.blk.22.attn_k.bias
F32
F32
[1280]
v.blk.22.attn_k.weight
BF16
BF16
[1280, 1280]
v.blk.22.attn_out.bias
F32
F32
[1280]
v.blk.22.attn_out.weight
BF16
BF16
[1280, 1280]
v.blk.22.attn_q.bias
F32
F32
[1280]
v.blk.22.attn_q.weight
BF16
BF16
[1280, 1280]
v.blk.22.attn_v.bias
F32
F32
[1280]
v.blk.22.attn_v.weight
BF16
BF16
[1280, 1280]
v.blk.22.ffn_down.bias
F32
F32
[1280]
v.blk.22.ffn_down.weight
BF16
BF16
[3420, 1280]
v.blk.22.ffn_gate.bias
F32
F32
[3420]
v.blk.22.ffn_gate.weight
BF16
BF16
[1280, 3420]
v.blk.22.ffn_up.bias
F32
F32
[3420]
v.blk.22.ffn_up.weight
BF16
BF16
[1280, 3420]
v.blk.22.ln1.weight
F32
F32
[1280]
v.blk.22.ln2.weight
F32
F32
[1280]
v.blk.23
v.blk.23.attn_k.bias
F32
F32
[1280]
v.blk.23.attn_k.weight
BF16
BF16
[1280, 1280]
v.blk.23.attn_out.bias
F32
F32
[1280]
v.blk.23.attn_out.weight
BF16
BF16
[1280, 1280]
v.blk.23.attn_q.bias
F32
F32
[1280]
v.blk.23.attn_q.weight
BF16
BF16
[1280, 1280]
v.blk.23.attn_v.bias
F32
F32
[1280]
v.blk.23.attn_v.weight
BF16
BF16
[1280, 1280]
v.blk.23.ffn_down.bias
F32
F32
[1280]
v.blk.23.ffn_down.weight
BF16
BF16
[3420, 1280]
v.blk.23.ffn_gate.bias
F32
F32
[3420]
v.blk.23.ffn_gate.weight
BF16
BF16
[1280, 3420]
v.blk.23.ffn_up.bias
F32
F32
[3420]
v.blk.23.ffn_up.weight
BF16
BF16
[1280, 3420]
v.blk.23.ln1.weight
F32
F32
[1280]
v.blk.23.ln2.weight
F32
F32
[1280]
v.blk.24
v.blk.24.attn_k.bias
F32
F32
[1280]
v.blk.24.attn_k.weight
BF16
BF16
[1280, 1280]
v.blk.24.attn_out.bias
F32
F32
[1280]
v.blk.24.attn_out.weight
BF16
BF16
[1280, 1280]
v.blk.24.attn_q.bias
F32
F32
[1280]
v.blk.24.attn_q.weight
BF16
BF16
[1280, 1280]
v.blk.24.attn_v.bias
F32
F32
[1280]
v.blk.24.attn_v.weight
BF16
BF16
[1280, 1280]
v.blk.24.ffn_down.bias
F32
F32
[1280]
v.blk.24.ffn_down.weight
BF16
BF16
[3420, 1280]
v.blk.24.ffn_gate.bias
F32
F32
[3420]
v.blk.24.ffn_gate.weight
BF16
BF16
[1280, 3420]
v.blk.24.ffn_up.bias
F32
F32
[3420]
v.blk.24.ffn_up.weight
BF16
BF16
[1280, 3420]
v.blk.24.ln1.weight
F32
F32
[1280]
v.blk.24.ln2.weight
F32
F32
[1280]
v.blk.25
v.blk.25.attn_k.bias
F32
F32
[1280]
v.blk.25.attn_k.weight
BF16
BF16
[1280, 1280]
v.blk.25.attn_out.bias
F32
F32
[1280]
v.blk.25.attn_out.weight
BF16
BF16
[1280, 1280]
v.blk.25.attn_q.bias
F32
F32
[1280]
v.blk.25.attn_q.weight
BF16
BF16
[1280, 1280]
v.blk.25.attn_v.bias
F32
F32
[1280]
v.blk.25.attn_v.weight
BF16
BF16
[1280, 1280]
v.blk.25.ffn_down.bias
F32
F32
[1280]
v.blk.25.ffn_down.weight
BF16
BF16
[3420, 1280]
v.blk.25.ffn_gate.bias
F32
F32
[3420]
v.blk.25.ffn_gate.weight
BF16
BF16
[1280, 3420]
v.blk.25.ffn_up.bias
F32
F32
[3420]
v.blk.25.ffn_up.weight
BF16
BF16
[1280, 3420]
v.blk.25.ln1.weight
F32
F32
[1280]
v.blk.25.ln2.weight
F32
F32
[1280]
v.blk.26
v.blk.26.attn_k.bias
F32
F32
[1280]
v.blk.26.attn_k.weight
BF16
BF16
[1280, 1280]
v.blk.26.attn_out.bias
F32
F32
[1280]
v.blk.26.attn_out.weight
BF16
BF16
[1280, 1280]
v.blk.26.attn_q.bias
F32
F32
[1280]
v.blk.26.attn_q.weight
BF16
BF16
[1280, 1280]
v.blk.26.attn_v.bias
F32
F32
[1280]
v.blk.26.attn_v.weight
BF16
BF16
[1280, 1280]
v.blk.26.ffn_down.bias
F32
F32
[1280]
v.blk.26.ffn_down.weight
BF16
BF16
[3420, 1280]
v.blk.26.ffn_gate.bias
F32
F32
[3420]
v.blk.26.ffn_gate.weight
BF16
BF16
[1280, 3420]
v.blk.26.ffn_up.bias
F32
F32
[3420]
v.blk.26.ffn_up.weight
BF16
BF16
[1280, 3420]
v.blk.26.ln1.weight
F32
F32
[1280]
v.blk.26.ln2.weight
F32
F32
[1280]
v.blk.27
v.blk.27.attn_k.bias
F32
F32
[1280]
v.blk.27.attn_k.weight
BF16
BF16
[1280, 1280]
v.blk.27.attn_out.bias
F32
F32
[1280]
v.blk.27.attn_out.weight
BF16
BF16
[1280, 1280]
v.blk.27.attn_q.bias
F32
F32
[1280]
v.blk.27.attn_q.weight
BF16
BF16
[1280, 1280]
v.blk.27.attn_v.bias
F32
F32
[1280]
v.blk.27.attn_v.weight
BF16
BF16
[1280, 1280]
v.blk.27.ffn_down.bias
F32
F32
[1280]
v.blk.27.ffn_down.weight
BF16
BF16
[3420, 1280]
v.blk.27.ffn_gate.bias
F32
F32
[3420]
v.blk.27.ffn_gate.weight
BF16
BF16
[1280, 3420]
v.blk.27.ffn_up.bias
F32
F32
[3420]
v.blk.27.ffn_up.weight
BF16
BF16
[1280, 3420]
v.blk.27.ln1.weight
F32
F32
[1280]
v.blk.27.ln2.weight
F32
F32
[1280]
v.blk.28
v.blk.28.attn_k.bias
F32
F32
[1280]
v.blk.28.attn_k.weight
BF16
BF16
[1280, 1280]
v.blk.28.attn_out.bias
F32
F32
[1280]
v.blk.28.attn_out.weight
BF16
BF16
[1280, 1280]
v.blk.28.attn_q.bias
F32
F32
[1280]
v.blk.28.attn_q.weight
BF16
BF16
[1280, 1280]
v.blk.28.attn_v.bias
F32
F32
[1280]
v.blk.28.attn_v.weight
BF16
BF16
[1280, 1280]
v.blk.28.ffn_down.bias
F32
F32
[1280]
v.blk.28.ffn_down.weight
BF16
BF16
[3420, 1280]
v.blk.28.ffn_gate.bias
F32
F32
[3420]
v.blk.28.ffn_gate.weight
BF16
BF16
[1280, 3420]
v.blk.28.ffn_up.bias
F32
F32
[3420]
v.blk.28.ffn_up.weight
BF16
BF16
[1280, 3420]
v.blk.28.ln1.weight
F32
F32
[1280]
v.blk.28.ln2.weight
F32
F32
[1280]
v.blk.29
v.blk.29.attn_k.bias
F32
F32
[1280]
v.blk.29.attn_k.weight
BF16
BF16
[1280, 1280]
v.blk.29.attn_out.bias
F32
F32
[1280]
v.blk.29.attn_out.weight
BF16
BF16
[1280, 1280]
v.blk.29.attn_q.bias
F32
F32
[1280]
v.blk.29.attn_q.weight
BF16
BF16
[1280, 1280]
v.blk.29.attn_v.bias
F32
F32
[1280]
v.blk.29.attn_v.weight
BF16
BF16
[1280, 1280]
v.blk.29.ffn_down.bias
F32
F32
[1280]
v.blk.29.ffn_down.weight
BF16
BF16
[3420, 1280]
v.blk.29.ffn_gate.bias
F32
F32
[3420]
v.blk.29.ffn_gate.weight
BF16
BF16
[1280, 3420]
v.blk.29.ffn_up.bias
F32
F32
[3420]
v.blk.29.ffn_up.weight
BF16
BF16
[1280, 3420]
v.blk.29.ln1.weight
F32
F32
[1280]
v.blk.29.ln2.weight
F32
F32
[1280]
v.blk.30
v.blk.30.attn_k.bias
F32
F32
[1280]
v.blk.30.attn_k.weight
BF16
BF16
[1280, 1280]
v.blk.30.attn_out.bias
F32
F32
[1280]
v.blk.30.attn_out.weight
BF16
BF16
[1280, 1280]
v.blk.30.attn_q.bias
F32
F32
[1280]
v.blk.30.attn_q.weight
BF16
BF16
[1280, 1280]
v.blk.30.attn_v.bias
F32
F32
[1280]
v.blk.30.attn_v.weight
BF16
BF16
[1280, 1280]
v.blk.30.ffn_down.bias
F32
F32
[1280]
v.blk.30.ffn_down.weight
BF16
BF16
[3420, 1280]
v.blk.30.ffn_gate.bias
F32
F32
[3420]
v.blk.30.ffn_gate.weight
BF16
BF16
[1280, 3420]
v.blk.30.ffn_up.bias
F32
F32
[3420]
v.blk.30.ffn_up.weight
BF16
BF16
[1280, 3420]
v.blk.30.ln1.weight
F32
F32
[1280]
v.blk.30.ln2.weight
F32
F32
[1280]
v.blk.31
v.blk.31.attn_k.bias
F32
F32
[1280]
v.blk.31.attn_k.weight
BF16
BF16
[1280, 1280]
v.blk.31.attn_out.bias
F32
F32
[1280]
v.blk.31.attn_out.weight
BF16
BF16
[1280, 1280]
v.blk.31.attn_q.bias
F32
F32
[1280]
v.blk.31.attn_q.weight
BF16
BF16
[1280, 1280]
v.blk.31.attn_v.bias
F32
F32
[1280]
v.blk.31.attn_v.weight
BF16
BF16
[1280, 1280]
v.blk.31.ffn_down.bias
F32
F32
[1280]
v.blk.31.ffn_down.weight
BF16
BF16
[3420, 1280]
v.blk.31.ffn_gate.bias
F32
F32
[3420]
v.blk.31.ffn_gate.weight
BF16
BF16
[1280, 3420]
v.blk.31.ffn_up.bias
F32
F32
[3420]
v.blk.31.ffn_up.weight
BF16
BF16
[1280, 3420]
v.blk.31.ln1.weight
F32
F32
[1280]
v.blk.31.ln2.weight
F32
F32
[1280]
v.patch_embd.weight
F32
F32
[14, 14, 3, 1280]
v.patch_embd.weight.1
F32
F32
[14, 14, 3, 1280]
v.post_ln.weight
F32
F32
[1280]