Models
Docs
Pricing
Sign in
Download
Models
Download
Docs
Pricing
Sign in
dhiltgen
/
deepseek-ocr
:3b-bf16
13
Downloads
Updated
2 days ago
Cancel
vision
3b
deepseek-ocr:3b-bf16
...
/
model
c40a82e49ba0 · 5.9GB
Metadata
general.architecture
deepseek2-ocr
deepseek2-ocr
general.file_type
F16
F16
deepseek2-ocr.attention.head_count
10
10
deepseek2-ocr.attention.head_count_kv
10
10
deepseek2-ocr.attention.layer_norm_rms_epsilon
1e-06
1e-06
deepseek2-ocr.block_count
12
12
deepseek2-ocr.context_length
8192
8192
deepseek2-ocr.embedding_length
1280
1280
deepseek2-ocr.expert_count
64
64
deepseek2-ocr.expert_feed_forward_length
896
896
deepseek2-ocr.expert_group_count
1
1
deepseek2-ocr.expert_group_used_count
1
1
deepseek2-ocr.expert_shared_count
2
2
deepseek2-ocr.expert_used_count
6
6
deepseek2-ocr.feed_forward_length
6848
6848
deepseek2-ocr.leading_dense_block_count
1
1
deepseek2-ocr.rope.dimension_count
0
0
deepseek2-ocr.vocab_size
129280
129280
tokenizer.ggml.add_bos_token
true
true
tokenizer.ggml.add_eos_token
false
false
tokenizer.ggml.add_padding_token
false
false
tokenizer.ggml.bos_token_id
0
0
tokenizer.ggml.eos_token_id
1
1
tokenizer.ggml.merges
[Ġ t, Ġ a, i n, Ġ Ġ, h e, ...]
[Ġ t, Ġ a, i n, Ġ Ġ, h e, ...]
tokenizer.ggml.model
gpt2
gpt2
tokenizer.ggml.padding_token_id
2
2
tokenizer.ggml.pre
default
default
tokenizer.ggml.scores
[0, 1, 2, 3, 4, ...]
[0, 1, 2, 3, 4, ...]
tokenizer.ggml.token_type
[3, 3, 3, 1, 1, ...]
[3, 3, 3, 1, 1, ...]
tokenizer.ggml.tokens
[<|begin▁of▁sentence|>, <|end▁of▁sentence|>, <|▁pad▁|>, !, ", ...]
[<|begin▁of▁sentence|>, <|end▁of▁sentence|>, <|▁pad▁|>, !, ", ...]
Tensor
Name
Type
Shape
token_embd.weight
BF16
BF16
[1280, 129280]
blk.0
blk.0.attn_k.weight
BF16
BF16
[1280, 1280]
blk.0.attn_norm.weight
F32
F32
[1280]
blk.0.attn_output.weight
BF16
BF16
[1280, 1280]
blk.0.attn_q.weight
BF16
BF16
[1280, 1280]
blk.0.attn_v.weight
BF16
BF16
[1280, 1280]
blk.0.ffn_down.weight
BF16
BF16
[6848, 1280]
blk.0.ffn_gate.weight
BF16
BF16
[1280, 6848]
blk.0.ffn_norm.weight
F32
F32
[1280]
blk.0.ffn_up.weight
BF16
BF16
[1280, 6848]
blk.1
blk.1.attn_k.weight
BF16
BF16
[1280, 1280]
blk.1.attn_norm.weight
F32
F32
[1280]
blk.1.attn_output.weight
BF16
BF16
[1280, 1280]
blk.1.attn_q.weight
BF16
BF16
[1280, 1280]
blk.1.attn_v.weight
BF16
BF16
[1280, 1280]
blk.1.ffn_down_exps.weight
BF16
BF16
[896, 1280, 64]
blk.1.ffn_down_shexp.weight
BF16
BF16
[1792, 1280]
blk.1.ffn_gate_exps.weight
BF16
BF16
[1280, 896, 64]
blk.1.ffn_gate_inp.weight
F32
F32
[1280, 64]
blk.1.ffn_gate_shexp.weight
BF16
BF16
[1280, 1792]
blk.1.ffn_norm.weight
F32
F32
[1280]
blk.1.ffn_up_exps.weight
BF16
BF16
[1280, 896, 64]
blk.1.ffn_up_shexp.weight
BF16
BF16
[1280, 1792]
blk.2
blk.2.attn_k.weight
BF16
BF16
[1280, 1280]
blk.2.attn_norm.weight
F32
F32
[1280]
blk.2.attn_output.weight
BF16
BF16
[1280, 1280]
blk.2.attn_q.weight
BF16
BF16
[1280, 1280]
blk.2.attn_v.weight
BF16
BF16
[1280, 1280]
blk.2.ffn_down_exps.weight
BF16
BF16
[896, 1280, 64]
blk.2.ffn_down_shexp.weight
BF16
BF16
[1792, 1280]
blk.2.ffn_gate_exps.weight
BF16
BF16
[1280, 896, 64]
blk.2.ffn_gate_inp.weight
F32
F32
[1280, 64]
blk.2.ffn_gate_shexp.weight
BF16
BF16
[1280, 1792]
blk.2.ffn_norm.weight
F32
F32
[1280]
blk.2.ffn_up_exps.weight
BF16
BF16
[1280, 896, 64]
blk.2.ffn_up_shexp.weight
BF16
BF16
[1280, 1792]
blk.3
blk.3.attn_k.weight
BF16
BF16
[1280, 1280]
blk.3.attn_norm.weight
F32
F32
[1280]
blk.3.attn_output.weight
BF16
BF16
[1280, 1280]
blk.3.attn_q.weight
BF16
BF16
[1280, 1280]
blk.3.attn_v.weight
BF16
BF16
[1280, 1280]
blk.3.ffn_down_exps.weight
BF16
BF16
[896, 1280, 64]
blk.3.ffn_down_shexp.weight
BF16
BF16
[1792, 1280]
blk.3.ffn_gate_exps.weight
BF16
BF16
[1280, 896, 64]
blk.3.ffn_gate_inp.weight
F32
F32
[1280, 64]
blk.3.ffn_gate_shexp.weight
BF16
BF16
[1280, 1792]
blk.3.ffn_norm.weight
F32
F32
[1280]
blk.3.ffn_up_exps.weight
BF16
BF16
[1280, 896, 64]
blk.3.ffn_up_shexp.weight
BF16
BF16
[1280, 1792]
blk.4
blk.4.attn_k.weight
BF16
BF16
[1280, 1280]
blk.4.attn_norm.weight
F32
F32
[1280]
blk.4.attn_output.weight
BF16
BF16
[1280, 1280]
blk.4.attn_q.weight
BF16
BF16
[1280, 1280]
blk.4.attn_v.weight
BF16
BF16
[1280, 1280]
blk.4.ffn_down_exps.weight
BF16
BF16
[896, 1280, 64]
blk.4.ffn_down_shexp.weight
BF16
BF16
[1792, 1280]
blk.4.ffn_gate_exps.weight
BF16
BF16
[1280, 896, 64]
blk.4.ffn_gate_inp.weight
F32
F32
[1280, 64]
blk.4.ffn_gate_shexp.weight
BF16
BF16
[1280, 1792]
blk.4.ffn_norm.weight
F32
F32
[1280]
blk.4.ffn_up_exps.weight
BF16
BF16
[1280, 896, 64]
blk.4.ffn_up_shexp.weight
BF16
BF16
[1280, 1792]
blk.5
blk.5.attn_k.weight
BF16
BF16
[1280, 1280]
blk.5.attn_norm.weight
F32
F32
[1280]
blk.5.attn_output.weight
BF16
BF16
[1280, 1280]
blk.5.attn_q.weight
BF16
BF16
[1280, 1280]
blk.5.attn_v.weight
BF16
BF16
[1280, 1280]
blk.5.ffn_down_exps.weight
BF16
BF16
[896, 1280, 64]
blk.5.ffn_down_shexp.weight
BF16
BF16
[1792, 1280]
blk.5.ffn_gate_exps.weight
BF16
BF16
[1280, 896, 64]
blk.5.ffn_gate_inp.weight
F32
F32
[1280, 64]
blk.5.ffn_gate_shexp.weight
BF16
BF16
[1280, 1792]
blk.5.ffn_norm.weight
F32
F32
[1280]
blk.5.ffn_up_exps.weight
BF16
BF16
[1280, 896, 64]
blk.5.ffn_up_shexp.weight
BF16
BF16
[1280, 1792]
blk.6
blk.6.attn_k.weight
BF16
BF16
[1280, 1280]
blk.6.attn_norm.weight
F32
F32
[1280]
blk.6.attn_output.weight
BF16
BF16
[1280, 1280]
blk.6.attn_q.weight
BF16
BF16
[1280, 1280]
blk.6.attn_v.weight
BF16
BF16
[1280, 1280]
blk.6.ffn_down_exps.weight
BF16
BF16
[896, 1280, 64]
blk.6.ffn_down_shexp.weight
BF16
BF16
[1792, 1280]
blk.6.ffn_gate_exps.weight
BF16
BF16
[1280, 896, 64]
blk.6.ffn_gate_inp.weight
F32
F32
[1280, 64]
blk.6.ffn_gate_shexp.weight
BF16
BF16
[1280, 1792]
blk.6.ffn_norm.weight
F32
F32
[1280]
blk.6.ffn_up_exps.weight
BF16
BF16
[1280, 896, 64]
blk.6.ffn_up_shexp.weight
BF16
BF16
[1280, 1792]
blk.7
blk.7.attn_k.weight
BF16
BF16
[1280, 1280]
blk.7.attn_norm.weight
F32
F32
[1280]
blk.7.attn_output.weight
BF16
BF16
[1280, 1280]
blk.7.attn_q.weight
BF16
BF16
[1280, 1280]
blk.7.attn_v.weight
BF16
BF16
[1280, 1280]
blk.7.ffn_down_exps.weight
BF16
BF16
[896, 1280, 64]
blk.7.ffn_down_shexp.weight
BF16
BF16
[1792, 1280]
blk.7.ffn_gate_exps.weight
BF16
BF16
[1280, 896, 64]
blk.7.ffn_gate_inp.weight
F32
F32
[1280, 64]
blk.7.ffn_gate_shexp.weight
BF16
BF16
[1280, 1792]
blk.7.ffn_norm.weight
F32
F32
[1280]
blk.7.ffn_up_exps.weight
BF16
BF16
[1280, 896, 64]
blk.7.ffn_up_shexp.weight
BF16
BF16
[1280, 1792]
blk.8
blk.8.attn_k.weight
BF16
BF16
[1280, 1280]
blk.8.attn_norm.weight
F32
F32
[1280]
blk.8.attn_output.weight
BF16
BF16
[1280, 1280]
blk.8.attn_q.weight
BF16
BF16
[1280, 1280]
blk.8.attn_v.weight
BF16
BF16
[1280, 1280]
blk.8.ffn_down_exps.weight
BF16
BF16
[896, 1280, 64]
blk.8.ffn_down_shexp.weight
BF16
BF16
[1792, 1280]
blk.8.ffn_gate_exps.weight
BF16
BF16
[1280, 896, 64]
blk.8.ffn_gate_inp.weight
F32
F32
[1280, 64]
blk.8.ffn_gate_shexp.weight
BF16
BF16
[1280, 1792]
blk.8.ffn_norm.weight
F32
F32
[1280]
blk.8.ffn_up_exps.weight
BF16
BF16
[1280, 896, 64]
blk.8.ffn_up_shexp.weight
BF16
BF16
[1280, 1792]
blk.9
blk.9.attn_k.weight
BF16
BF16
[1280, 1280]
blk.9.attn_norm.weight
F32
F32
[1280]
blk.9.attn_output.weight
BF16
BF16
[1280, 1280]
blk.9.attn_q.weight
BF16
BF16
[1280, 1280]
blk.9.attn_v.weight
BF16
BF16
[1280, 1280]
blk.9.ffn_down_exps.weight
BF16
BF16
[896, 1280, 64]
blk.9.ffn_down_shexp.weight
BF16
BF16
[1792, 1280]
blk.9.ffn_gate_exps.weight
BF16
BF16
[1280, 896, 64]
blk.9.ffn_gate_inp.weight
F32
F32
[1280, 64]
blk.9.ffn_gate_shexp.weight
BF16
BF16
[1280, 1792]
blk.9.ffn_norm.weight
F32
F32
[1280]
blk.9.ffn_up_exps.weight
BF16
BF16
[1280, 896, 64]
blk.9.ffn_up_shexp.weight
BF16
BF16
[1280, 1792]
blk.10
blk.10.attn_k.weight
BF16
BF16
[1280, 1280]
blk.10.attn_norm.weight
F32
F32
[1280]
blk.10.attn_output.weight
BF16
BF16
[1280, 1280]
blk.10.attn_q.weight
BF16
BF16
[1280, 1280]
blk.10.attn_v.weight
BF16
BF16
[1280, 1280]
blk.10.ffn_down_exps.weight
BF16
BF16
[896, 1280, 64]
blk.10.ffn_down_shexp.weight
BF16
BF16
[1792, 1280]
blk.10.ffn_gate_exps.weight
BF16
BF16
[1280, 896, 64]
blk.10.ffn_gate_inp.weight
F32
F32
[1280, 64]
blk.10.ffn_gate_shexp.weight
BF16
BF16
[1280, 1792]
blk.10.ffn_norm.weight
F32
F32
[1280]
blk.10.ffn_up_exps.weight
BF16
BF16
[1280, 896, 64]
blk.10.ffn_up_shexp.weight
BF16
BF16
[1280, 1792]
blk.11
blk.11.attn_k.weight
BF16
BF16
[1280, 1280]
blk.11.attn_norm.weight
F32
F32
[1280]
blk.11.attn_output.weight
BF16
BF16
[1280, 1280]
blk.11.attn_q.weight
BF16
BF16
[1280, 1280]
blk.11.attn_v.weight
BF16
BF16
[1280, 1280]
blk.11.ffn_down_exps.weight
BF16
BF16
[896, 1280, 64]
blk.11.ffn_down_shexp.weight
BF16
BF16
[1792, 1280]
blk.11.ffn_gate_exps.weight
BF16
BF16
[1280, 896, 64]
blk.11.ffn_gate_inp.weight
F32
F32
[1280, 64]
blk.11.ffn_gate_shexp.weight
BF16
BF16
[1280, 1792]
blk.11.ffn_norm.weight
F32
F32
[1280]
blk.11.ffn_up_exps.weight
BF16
BF16
[1280, 896, 64]
blk.11.ffn_up_shexp.weight
BF16
BF16
[1280, 1792]
output.weight
BF16
BF16
[1280, 129280]
output_norm.weight
F32
F32
[1280]