Models
GitHub
Discord
Turbo
Sign in
Download
Models
Download
GitHub
Discord
Sign in
bge-m3
:567m
1.9M
Downloads
Updated
1 year ago
BGE-M3 is a new model from BAAI distinguished for its versatility in Multi-Functionality, Multi-Linguality, and Multi-Granularity.
BGE-M3 is a new model from BAAI distinguished for its versatility in Multi-Functionality, Multi-Linguality, and Multi-Granularity.
Cancel
embedding
567m
bge-m3:567m
...
/
model
daec91ffb5dd · 1.2GB
Metadata
general.architecture
bert
bert
general.file_type
F16
F16
bert.attention.causal
false
false
bert.attention.head_count
16
16
bert.attention.layer_norm_epsilon
1e-05
1e-05
bert.block_count
24
24
bert.context_length
8192
8192
bert.embedding_length
1024
1024
bert.feed_forward_length
4096
4096
bert.pooling_type
CLS
CLS
tokenizer.ggml.add_bos_token
true
true
tokenizer.ggml.add_eos_token
true
true
tokenizer.ggml.add_space_prefix
true
true
tokenizer.ggml.bos_token_id
0
0
tokenizer.ggml.cls_token_id
0
0
tokenizer.ggml.eos_token_id
2
2
tokenizer.ggml.mask_token_id
250001
250001
tokenizer.ggml.model
t5
t5
tokenizer.ggml.padding_token_id
1
1
tokenizer.ggml.pre
default
default
tokenizer.ggml.precompiled_charsmap
[0, 180, 2, 0, 0, ...]
[0, 180, 2, 0, 0, ...]
tokenizer.ggml.remove_extra_whitespaces
true
true
tokenizer.ggml.scores
[0, 0, 0, 0, -3.4635427, ...]
[0, 0, 0, 0, -3.4635427, ...]
tokenizer.ggml.seperator_token_id
2
2
tokenizer.ggml.token_type
[3, 3, 3, 2, 1, ...]
[3, 3, 3, 2, 1, ...]
tokenizer.ggml.token_type_count
1
1
tokenizer.ggml.tokens
[<s>, <pad>, </s>, <unk>, ,, ...]
[<s>, <pad>, </s>, <unk>, ,, ...]
tokenizer.ggml.unknown_token_id
3
3
Tensor
Name
Type
Shape
token_embd.weight
F16
F16
[1024, 250002]
blk.0
blk.0.attn_k.bias
F32
F32
[1024]
blk.0.attn_k.weight
F16
F16
[1024, 1024]
blk.0.attn_output.bias
F32
F32
[1024]
blk.0.attn_output.weight
F16
F16
[1024, 1024]
blk.0.attn_output_norm.bias
F32
F32
[1024]
blk.0.attn_output_norm.weight
F32
F32
[1024]
blk.0.attn_q.bias
F32
F32
[1024]
blk.0.attn_q.weight
F16
F16
[1024, 1024]
blk.0.attn_v.bias
F32
F32
[1024]
blk.0.attn_v.weight
F16
F16
[1024, 1024]
blk.0.ffn_down.bias
F32
F32
[1024]
blk.0.ffn_down.weight
F16
F16
[4096, 1024]
blk.0.ffn_up.bias
F32
F32
[4096]
blk.0.ffn_up.weight
F16
F16
[1024, 4096]
blk.0.layer_output_norm.bias
F32
F32
[1024]
blk.0.layer_output_norm.weight
F32
F32
[1024]
blk.1
blk.1.attn_k.bias
F32
F32
[1024]
blk.1.attn_k.weight
F16
F16
[1024, 1024]
blk.1.attn_output.bias
F32
F32
[1024]
blk.1.attn_output.weight
F16
F16
[1024, 1024]
blk.1.attn_output_norm.bias
F32
F32
[1024]
blk.1.attn_output_norm.weight
F32
F32
[1024]
blk.1.attn_q.bias
F32
F32
[1024]
blk.1.attn_q.weight
F16
F16
[1024, 1024]
blk.1.attn_v.bias
F32
F32
[1024]
blk.1.attn_v.weight
F16
F16
[1024, 1024]
blk.1.ffn_down.bias
F32
F32
[1024]
blk.1.ffn_down.weight
F16
F16
[4096, 1024]
blk.1.ffn_up.bias
F32
F32
[4096]
blk.1.ffn_up.weight
F16
F16
[1024, 4096]
blk.1.layer_output_norm.bias
F32
F32
[1024]
blk.1.layer_output_norm.weight
F32
F32
[1024]
blk.2
blk.2.attn_k.bias
F32
F32
[1024]
blk.2.attn_k.weight
F16
F16
[1024, 1024]
blk.2.attn_output.bias
F32
F32
[1024]
blk.2.attn_output.weight
F16
F16
[1024, 1024]
blk.2.attn_output_norm.bias
F32
F32
[1024]
blk.2.attn_output_norm.weight
F32
F32
[1024]
blk.2.attn_q.bias
F32
F32
[1024]
blk.2.attn_q.weight
F16
F16
[1024, 1024]
blk.2.attn_v.bias
F32
F32
[1024]
blk.2.attn_v.weight
F16
F16
[1024, 1024]
blk.2.ffn_down.bias
F32
F32
[1024]
blk.2.ffn_down.weight
F16
F16
[4096, 1024]
blk.2.ffn_up.bias
F32
F32
[4096]
blk.2.ffn_up.weight
F16
F16
[1024, 4096]
blk.2.layer_output_norm.bias
F32
F32
[1024]
blk.2.layer_output_norm.weight
F32
F32
[1024]
blk.3
blk.3.attn_k.bias
F32
F32
[1024]
blk.3.attn_k.weight
F16
F16
[1024, 1024]
blk.3.attn_output.bias
F32
F32
[1024]
blk.3.attn_output.weight
F16
F16
[1024, 1024]
blk.3.attn_output_norm.bias
F32
F32
[1024]
blk.3.attn_output_norm.weight
F32
F32
[1024]
blk.3.attn_q.bias
F32
F32
[1024]
blk.3.attn_q.weight
F16
F16
[1024, 1024]
blk.3.attn_v.bias
F32
F32
[1024]
blk.3.attn_v.weight
F16
F16
[1024, 1024]
blk.3.ffn_down.bias
F32
F32
[1024]
blk.3.ffn_down.weight
F16
F16
[4096, 1024]
blk.3.ffn_up.bias
F32
F32
[4096]
blk.3.ffn_up.weight
F16
F16
[1024, 4096]
blk.3.layer_output_norm.bias
F32
F32
[1024]
blk.3.layer_output_norm.weight
F32
F32
[1024]
blk.4
blk.4.attn_k.bias
F32
F32
[1024]
blk.4.attn_k.weight
F16
F16
[1024, 1024]
blk.4.attn_output.bias
F32
F32
[1024]
blk.4.attn_output.weight
F16
F16
[1024, 1024]
blk.4.attn_output_norm.bias
F32
F32
[1024]
blk.4.attn_output_norm.weight
F32
F32
[1024]
blk.4.attn_q.bias
F32
F32
[1024]
blk.4.attn_q.weight
F16
F16
[1024, 1024]
blk.4.attn_v.bias
F32
F32
[1024]
blk.4.attn_v.weight
F16
F16
[1024, 1024]
blk.4.ffn_down.bias
F32
F32
[1024]
blk.4.ffn_down.weight
F16
F16
[4096, 1024]
blk.4.ffn_up.bias
F32
F32
[4096]
blk.4.ffn_up.weight
F16
F16
[1024, 4096]
blk.4.layer_output_norm.bias
F32
F32
[1024]
blk.4.layer_output_norm.weight
F32
F32
[1024]
blk.5
blk.5.attn_k.bias
F32
F32
[1024]
blk.5.attn_k.weight
F16
F16
[1024, 1024]
blk.5.attn_output.bias
F32
F32
[1024]
blk.5.attn_output.weight
F16
F16
[1024, 1024]
blk.5.attn_output_norm.bias
F32
F32
[1024]
blk.5.attn_output_norm.weight
F32
F32
[1024]
blk.5.attn_q.bias
F32
F32
[1024]
blk.5.attn_q.weight
F16
F16
[1024, 1024]
blk.5.attn_v.bias
F32
F32
[1024]
blk.5.attn_v.weight
F16
F16
[1024, 1024]
blk.5.ffn_down.bias
F32
F32
[1024]
blk.5.ffn_down.weight
F16
F16
[4096, 1024]
blk.5.ffn_up.bias
F32
F32
[4096]
blk.5.ffn_up.weight
F16
F16
[1024, 4096]
blk.5.layer_output_norm.bias
F32
F32
[1024]
blk.5.layer_output_norm.weight
F32
F32
[1024]
blk.6
blk.6.attn_k.bias
F32
F32
[1024]
blk.6.attn_k.weight
F16
F16
[1024, 1024]
blk.6.attn_output.bias
F32
F32
[1024]
blk.6.attn_output.weight
F16
F16
[1024, 1024]
blk.6.attn_output_norm.bias
F32
F32
[1024]
blk.6.attn_output_norm.weight
F32
F32
[1024]
blk.6.attn_q.bias
F32
F32
[1024]
blk.6.attn_q.weight
F16
F16
[1024, 1024]
blk.6.attn_v.bias
F32
F32
[1024]
blk.6.attn_v.weight
F16
F16
[1024, 1024]
blk.6.ffn_down.bias
F32
F32
[1024]
blk.6.ffn_down.weight
F16
F16
[4096, 1024]
blk.6.ffn_up.bias
F32
F32
[4096]
blk.6.ffn_up.weight
F16
F16
[1024, 4096]
blk.6.layer_output_norm.bias
F32
F32
[1024]
blk.6.layer_output_norm.weight
F32
F32
[1024]
blk.7
blk.7.attn_k.bias
F32
F32
[1024]
blk.7.attn_k.weight
F16
F16
[1024, 1024]
blk.7.attn_output.bias
F32
F32
[1024]
blk.7.attn_output.weight
F16
F16
[1024, 1024]
blk.7.attn_output_norm.bias
F32
F32
[1024]
blk.7.attn_output_norm.weight
F32
F32
[1024]
blk.7.attn_q.bias
F32
F32
[1024]
blk.7.attn_q.weight
F16
F16
[1024, 1024]
blk.7.attn_v.bias
F32
F32
[1024]
blk.7.attn_v.weight
F16
F16
[1024, 1024]
blk.7.ffn_down.bias
F32
F32
[1024]
blk.7.ffn_down.weight
F16
F16
[4096, 1024]
blk.7.ffn_up.bias
F32
F32
[4096]
blk.7.ffn_up.weight
F16
F16
[1024, 4096]
blk.7.layer_output_norm.bias
F32
F32
[1024]
blk.7.layer_output_norm.weight
F32
F32
[1024]
blk.8
blk.8.attn_k.bias
F32
F32
[1024]
blk.8.attn_k.weight
F16
F16
[1024, 1024]
blk.8.attn_output.bias
F32
F32
[1024]
blk.8.attn_output.weight
F16
F16
[1024, 1024]
blk.8.attn_output_norm.bias
F32
F32
[1024]
blk.8.attn_output_norm.weight
F32
F32
[1024]
blk.8.attn_q.bias
F32
F32
[1024]
blk.8.attn_q.weight
F16
F16
[1024, 1024]
blk.8.attn_v.bias
F32
F32
[1024]
blk.8.attn_v.weight
F16
F16
[1024, 1024]
blk.8.ffn_down.bias
F32
F32
[1024]
blk.8.ffn_down.weight
F16
F16
[4096, 1024]
blk.8.ffn_up.bias
F32
F32
[4096]
blk.8.ffn_up.weight
F16
F16
[1024, 4096]
blk.8.layer_output_norm.bias
F32
F32
[1024]
blk.8.layer_output_norm.weight
F32
F32
[1024]
blk.9
blk.9.attn_k.bias
F32
F32
[1024]
blk.9.attn_k.weight
F16
F16
[1024, 1024]
blk.9.attn_output.bias
F32
F32
[1024]
blk.9.attn_output.weight
F16
F16
[1024, 1024]
blk.9.attn_output_norm.bias
F32
F32
[1024]
blk.9.attn_output_norm.weight
F32
F32
[1024]
blk.9.attn_q.bias
F32
F32
[1024]
blk.9.attn_q.weight
F16
F16
[1024, 1024]
blk.9.attn_v.bias
F32
F32
[1024]
blk.9.attn_v.weight
F16
F16
[1024, 1024]
blk.9.ffn_down.bias
F32
F32
[1024]
blk.9.ffn_down.weight
F16
F16
[4096, 1024]
blk.9.ffn_up.bias
F32
F32
[4096]
blk.9.ffn_up.weight
F16
F16
[1024, 4096]
blk.9.layer_output_norm.bias
F32
F32
[1024]
blk.9.layer_output_norm.weight
F32
F32
[1024]
blk.10
blk.10.attn_k.bias
F32
F32
[1024]
blk.10.attn_k.weight
F16
F16
[1024, 1024]
blk.10.attn_output.bias
F32
F32
[1024]
blk.10.attn_output.weight
F16
F16
[1024, 1024]
blk.10.attn_output_norm.bias
F32
F32
[1024]
blk.10.attn_output_norm.weight
F32
F32
[1024]
blk.10.attn_q.bias
F32
F32
[1024]
blk.10.attn_q.weight
F16
F16
[1024, 1024]
blk.10.attn_v.bias
F32
F32
[1024]
blk.10.attn_v.weight
F16
F16
[1024, 1024]
blk.10.ffn_down.bias
F32
F32
[1024]
blk.10.ffn_down.weight
F16
F16
[4096, 1024]
blk.10.ffn_up.bias
F32
F32
[4096]
blk.10.ffn_up.weight
F16
F16
[1024, 4096]
blk.10.layer_output_norm.bias
F32
F32
[1024]
blk.10.layer_output_norm.weight
F32
F32
[1024]
blk.11
blk.11.attn_k.bias
F32
F32
[1024]
blk.11.attn_k.weight
F16
F16
[1024, 1024]
blk.11.attn_output.bias
F32
F32
[1024]
blk.11.attn_output.weight
F16
F16
[1024, 1024]
blk.11.attn_output_norm.bias
F32
F32
[1024]
blk.11.attn_output_norm.weight
F32
F32
[1024]
blk.11.attn_q.bias
F32
F32
[1024]
blk.11.attn_q.weight
F16
F16
[1024, 1024]
blk.11.attn_v.bias
F32
F32
[1024]
blk.11.attn_v.weight
F16
F16
[1024, 1024]
blk.11.ffn_down.bias
F32
F32
[1024]
blk.11.ffn_down.weight
F16
F16
[4096, 1024]
blk.11.ffn_up.bias
F32
F32
[4096]
blk.11.ffn_up.weight
F16
F16
[1024, 4096]
blk.11.layer_output_norm.bias
F32
F32
[1024]
blk.11.layer_output_norm.weight
F32
F32
[1024]
blk.12
blk.12.attn_k.bias
F32
F32
[1024]
blk.12.attn_k.weight
F16
F16
[1024, 1024]
blk.12.attn_output.bias
F32
F32
[1024]
blk.12.attn_output.weight
F16
F16
[1024, 1024]
blk.12.attn_output_norm.bias
F32
F32
[1024]
blk.12.attn_output_norm.weight
F32
F32
[1024]
blk.12.attn_q.bias
F32
F32
[1024]
blk.12.attn_q.weight
F16
F16
[1024, 1024]
blk.12.attn_v.bias
F32
F32
[1024]
blk.12.attn_v.weight
F16
F16
[1024, 1024]
blk.12.ffn_down.bias
F32
F32
[1024]
blk.12.ffn_down.weight
F16
F16
[4096, 1024]
blk.12.ffn_up.bias
F32
F32
[4096]
blk.12.ffn_up.weight
F16
F16
[1024, 4096]
blk.12.layer_output_norm.bias
F32
F32
[1024]
blk.12.layer_output_norm.weight
F32
F32
[1024]
blk.13
blk.13.attn_k.bias
F32
F32
[1024]
blk.13.attn_k.weight
F16
F16
[1024, 1024]
blk.13.attn_output.bias
F32
F32
[1024]
blk.13.attn_output.weight
F16
F16
[1024, 1024]
blk.13.attn_output_norm.bias
F32
F32
[1024]
blk.13.attn_output_norm.weight
F32
F32
[1024]
blk.13.attn_q.bias
F32
F32
[1024]
blk.13.attn_q.weight
F16
F16
[1024, 1024]
blk.13.attn_v.bias
F32
F32
[1024]
blk.13.attn_v.weight
F16
F16
[1024, 1024]
blk.13.ffn_down.bias
F32
F32
[1024]
blk.13.ffn_down.weight
F16
F16
[4096, 1024]
blk.13.ffn_up.bias
F32
F32
[4096]
blk.13.ffn_up.weight
F16
F16
[1024, 4096]
blk.13.layer_output_norm.bias
F32
F32
[1024]
blk.13.layer_output_norm.weight
F32
F32
[1024]
blk.14
blk.14.attn_k.bias
F32
F32
[1024]
blk.14.attn_k.weight
F16
F16
[1024, 1024]
blk.14.attn_output.bias
F32
F32
[1024]
blk.14.attn_output.weight
F16
F16
[1024, 1024]
blk.14.attn_output_norm.bias
F32
F32
[1024]
blk.14.attn_output_norm.weight
F32
F32
[1024]
blk.14.attn_q.bias
F32
F32
[1024]
blk.14.attn_q.weight
F16
F16
[1024, 1024]
blk.14.attn_v.bias
F32
F32
[1024]
blk.14.attn_v.weight
F16
F16
[1024, 1024]
blk.14.ffn_down.bias
F32
F32
[1024]
blk.14.ffn_down.weight
F16
F16
[4096, 1024]
blk.14.ffn_up.bias
F32
F32
[4096]
blk.14.ffn_up.weight
F16
F16
[1024, 4096]
blk.14.layer_output_norm.bias
F32
F32
[1024]
blk.14.layer_output_norm.weight
F32
F32
[1024]
blk.15
blk.15.attn_k.bias
F32
F32
[1024]
blk.15.attn_k.weight
F16
F16
[1024, 1024]
blk.15.attn_output.bias
F32
F32
[1024]
blk.15.attn_output.weight
F16
F16
[1024, 1024]
blk.15.attn_output_norm.bias
F32
F32
[1024]
blk.15.attn_output_norm.weight
F32
F32
[1024]
blk.15.attn_q.bias
F32
F32
[1024]
blk.15.attn_q.weight
F16
F16
[1024, 1024]
blk.15.attn_v.bias
F32
F32
[1024]
blk.15.attn_v.weight
F16
F16
[1024, 1024]
blk.15.ffn_down.bias
F32
F32
[1024]
blk.15.ffn_down.weight
F16
F16
[4096, 1024]
blk.15.ffn_up.bias
F32
F32
[4096]
blk.15.ffn_up.weight
F16
F16
[1024, 4096]
blk.15.layer_output_norm.bias
F32
F32
[1024]
blk.15.layer_output_norm.weight
F32
F32
[1024]
blk.16
blk.16.attn_k.bias
F32
F32
[1024]
blk.16.attn_k.weight
F16
F16
[1024, 1024]
blk.16.attn_output.bias
F32
F32
[1024]
blk.16.attn_output.weight
F16
F16
[1024, 1024]
blk.16.attn_output_norm.bias
F32
F32
[1024]
blk.16.attn_output_norm.weight
F32
F32
[1024]
blk.16.attn_q.bias
F32
F32
[1024]
blk.16.attn_q.weight
F16
F16
[1024, 1024]
blk.16.attn_v.bias
F32
F32
[1024]
blk.16.attn_v.weight
F16
F16
[1024, 1024]
blk.16.ffn_down.bias
F32
F32
[1024]
blk.16.ffn_down.weight
F16
F16
[4096, 1024]
blk.16.ffn_up.bias
F32
F32
[4096]
blk.16.ffn_up.weight
F16
F16
[1024, 4096]
blk.16.layer_output_norm.bias
F32
F32
[1024]
blk.16.layer_output_norm.weight
F32
F32
[1024]
blk.17
blk.17.attn_k.bias
F32
F32
[1024]
blk.17.attn_k.weight
F16
F16
[1024, 1024]
blk.17.attn_output.bias
F32
F32
[1024]
blk.17.attn_output.weight
F16
F16
[1024, 1024]
blk.17.attn_output_norm.bias
F32
F32
[1024]
blk.17.attn_output_norm.weight
F32
F32
[1024]
blk.17.attn_q.bias
F32
F32
[1024]
blk.17.attn_q.weight
F16
F16
[1024, 1024]
blk.17.attn_v.bias
F32
F32
[1024]
blk.17.attn_v.weight
F16
F16
[1024, 1024]
blk.17.ffn_down.bias
F32
F32
[1024]
blk.17.ffn_down.weight
F16
F16
[4096, 1024]
blk.17.ffn_up.bias
F32
F32
[4096]
blk.17.ffn_up.weight
F16
F16
[1024, 4096]
blk.17.layer_output_norm.bias
F32
F32
[1024]
blk.17.layer_output_norm.weight
F32
F32
[1024]
blk.18
blk.18.attn_k.bias
F32
F32
[1024]
blk.18.attn_k.weight
F16
F16
[1024, 1024]
blk.18.attn_output.bias
F32
F32
[1024]
blk.18.attn_output.weight
F16
F16
[1024, 1024]
blk.18.attn_output_norm.bias
F32
F32
[1024]
blk.18.attn_output_norm.weight
F32
F32
[1024]
blk.18.attn_q.bias
F32
F32
[1024]
blk.18.attn_q.weight
F16
F16
[1024, 1024]
blk.18.attn_v.bias
F32
F32
[1024]
blk.18.attn_v.weight
F16
F16
[1024, 1024]
blk.18.ffn_down.bias
F32
F32
[1024]
blk.18.ffn_down.weight
F16
F16
[4096, 1024]
blk.18.ffn_up.bias
F32
F32
[4096]
blk.18.ffn_up.weight
F16
F16
[1024, 4096]
blk.18.layer_output_norm.bias
F32
F32
[1024]
blk.18.layer_output_norm.weight
F32
F32
[1024]
blk.19
blk.19.attn_k.bias
F32
F32
[1024]
blk.19.attn_k.weight
F16
F16
[1024, 1024]
blk.19.attn_output.bias
F32
F32
[1024]
blk.19.attn_output.weight
F16
F16
[1024, 1024]
blk.19.attn_output_norm.bias
F32
F32
[1024]
blk.19.attn_output_norm.weight
F32
F32
[1024]
blk.19.attn_q.bias
F32
F32
[1024]
blk.19.attn_q.weight
F16
F16
[1024, 1024]
blk.19.attn_v.bias
F32
F32
[1024]
blk.19.attn_v.weight
F16
F16
[1024, 1024]
blk.19.ffn_down.bias
F32
F32
[1024]
blk.19.ffn_down.weight
F16
F16
[4096, 1024]
blk.19.ffn_up.bias
F32
F32
[4096]
blk.19.ffn_up.weight
F16
F16
[1024, 4096]
blk.19.layer_output_norm.bias
F32
F32
[1024]
blk.19.layer_output_norm.weight
F32
F32
[1024]
blk.20
blk.20.attn_k.bias
F32
F32
[1024]
blk.20.attn_k.weight
F16
F16
[1024, 1024]
blk.20.attn_output.bias
F32
F32
[1024]
blk.20.attn_output.weight
F16
F16
[1024, 1024]
blk.20.attn_output_norm.bias
F32
F32
[1024]
blk.20.attn_output_norm.weight
F32
F32
[1024]
blk.20.attn_q.bias
F32
F32
[1024]
blk.20.attn_q.weight
F16
F16
[1024, 1024]
blk.20.attn_v.bias
F32
F32
[1024]
blk.20.attn_v.weight
F16
F16
[1024, 1024]
blk.20.ffn_down.bias
F32
F32
[1024]
blk.20.ffn_down.weight
F16
F16
[4096, 1024]
blk.20.ffn_up.bias
F32
F32
[4096]
blk.20.ffn_up.weight
F16
F16
[1024, 4096]
blk.20.layer_output_norm.bias
F32
F32
[1024]
blk.20.layer_output_norm.weight
F32
F32
[1024]
blk.21
blk.21.attn_k.bias
F32
F32
[1024]
blk.21.attn_k.weight
F16
F16
[1024, 1024]
blk.21.attn_output.bias
F32
F32
[1024]
blk.21.attn_output.weight
F16
F16
[1024, 1024]
blk.21.attn_output_norm.bias
F32
F32
[1024]
blk.21.attn_output_norm.weight
F32
F32
[1024]
blk.21.attn_q.bias
F32
F32
[1024]
blk.21.attn_q.weight
F16
F16
[1024, 1024]
blk.21.attn_v.bias
F32
F32
[1024]
blk.21.attn_v.weight
F16
F16
[1024, 1024]
blk.21.ffn_down.bias
F32
F32
[1024]
blk.21.ffn_down.weight
F16
F16
[4096, 1024]
blk.21.ffn_up.bias
F32
F32
[4096]
blk.21.ffn_up.weight
F16
F16
[1024, 4096]
blk.21.layer_output_norm.bias
F32
F32
[1024]
blk.21.layer_output_norm.weight
F32
F32
[1024]
blk.22
blk.22.attn_k.bias
F32
F32
[1024]
blk.22.attn_k.weight
F16
F16
[1024, 1024]
blk.22.attn_output.bias
F32
F32
[1024]
blk.22.attn_output.weight
F16
F16
[1024, 1024]
blk.22.attn_output_norm.bias
F32
F32
[1024]
blk.22.attn_output_norm.weight
F32
F32
[1024]
blk.22.attn_q.bias
F32
F32
[1024]
blk.22.attn_q.weight
F16
F16
[1024, 1024]
blk.22.attn_v.bias
F32
F32
[1024]
blk.22.attn_v.weight
F16
F16
[1024, 1024]
blk.22.ffn_down.bias
F32
F32
[1024]
blk.22.ffn_down.weight
F16
F16
[4096, 1024]
blk.22.ffn_up.bias
F32
F32
[4096]
blk.22.ffn_up.weight
F16
F16
[1024, 4096]
blk.22.layer_output_norm.bias
F32
F32
[1024]
blk.22.layer_output_norm.weight
F32
F32
[1024]
blk.23
blk.23.attn_k.bias
F32
F32
[1024]
blk.23.attn_k.weight
F16
F16
[1024, 1024]
blk.23.attn_output.bias
F32
F32
[1024]
blk.23.attn_output.weight
F16
F16
[1024, 1024]
blk.23.attn_output_norm.bias
F32
F32
[1024]
blk.23.attn_output_norm.weight
F32
F32
[1024]
blk.23.attn_q.bias
F32
F32
[1024]
blk.23.attn_q.weight
F16
F16
[1024, 1024]
blk.23.attn_v.bias
F32
F32
[1024]
blk.23.attn_v.weight
F16
F16
[1024, 1024]
blk.23.ffn_down.bias
F32
F32
[1024]
blk.23.ffn_down.weight
F16
F16
[4096, 1024]
blk.23.ffn_up.bias
F32
F32
[4096]
blk.23.ffn_up.weight
F16
F16
[1024, 4096]
blk.23.layer_output_norm.bias
F32
F32
[1024]
blk.23.layer_output_norm.weight
F32
F32
[1024]
position_embd.weight
F32
F32
[1024, 8192]
token_embd_norm.bias
F32
F32
[1024]
token_embd_norm.weight
F32
F32
[1024]
token_types.weight
F32
F32
[1024]