Models
Docs
Pricing
Sign in
Download
Models
Download
Docs
Pricing
Sign in
s1gnature
/
tiny-aya
:3b-q8_0
18
Downloads
Updated
2 weeks ago
Cancel
tiny-aya:3b-q8_0
...
/
model
947713a2710e · 3.6GB
Metadata
general.architecture
cohere2
cohere2
general.file_type
Q8_0
Q8_0
cohere2.attention.head_count
16
16
cohere2.attention.head_count_kv
4
4
cohere2.attention.key_length
128
128
cohere2.attention.layer_norm_epsilon
1e-05
1e-05
cohere2.attention.sliding_window
4096
4096
cohere2.attention.value_length
128
128
cohere2.block_count
36
36
cohere2.context_length
500000
500000
cohere2.embedding_length
2048
2048
cohere2.feed_forward_length
11008
11008
cohere2.logit_scale
1
1
cohere2.rope.dimension_count
128
128
cohere2.rope.freq_base
50000
50000
cohere2.rope.scaling.type
none
none
cohere2.vocab_size
262144
262144
tokenizer.ggml.add_bos_token
true
true
tokenizer.ggml.add_eos_token
false
false
tokenizer.ggml.add_sep_token
false
false
tokenizer.ggml.bos_token_id
2
2
tokenizer.ggml.eos_token_id
3
3
tokenizer.ggml.merges
[Ġ Ġ, à ¤, Ġ t, e r, i n, ...]
[Ġ Ġ, à ¤, Ġ t, e r, i n, ...]
tokenizer.ggml.model
gpt2
gpt2
tokenizer.ggml.padding_token_id
0
0
tokenizer.ggml.pre
tiny_aya
tiny_aya
tokenizer.ggml.token_type
[3, 3, 3, 3, 3, ...]
[3, 3, 3, 3, 3, ...]
tokenizer.ggml.tokens
[<PAD>, <MASK_TOKEN>, <BOS_TOKEN>, <EOS_TOKEN>, <UNK>, ...]
[<PAD>, <MASK_TOKEN>, <BOS_TOKEN>, <EOS_TOKEN>, <UNK>, ...]
tokenizer.ggml.unknown_token_id
4
4
Tensor
Name
Type
Shape
token_embd.weight
Q8_0
Q8_0
[2048, 262144]
blk.0
blk.0.attn_k.weight
Q8_0
Q8_0
[2048, 512]
blk.0.attn_norm.weight
F32
F32
[2048]
blk.0.attn_output.weight
Q8_0
Q8_0
[2048, 2048]
blk.0.attn_q.weight
Q8_0
Q8_0
[2048, 2048]
blk.0.attn_v.weight
Q8_0
Q8_0
[2048, 512]
blk.0.ffn_down.weight
Q8_0
Q8_0
[11008, 2048]
blk.0.ffn_gate.weight
Q8_0
Q8_0
[2048, 11008]
blk.0.ffn_up.weight
Q8_0
Q8_0
[2048, 11008]
blk.1
blk.1.attn_k.weight
Q8_0
Q8_0
[2048, 512]
blk.1.attn_norm.weight
F32
F32
[2048]
blk.1.attn_output.weight
Q8_0
Q8_0
[2048, 2048]
blk.1.attn_q.weight
Q8_0
Q8_0
[2048, 2048]
blk.1.attn_v.weight
Q8_0
Q8_0
[2048, 512]
blk.1.ffn_down.weight
Q8_0
Q8_0
[11008, 2048]
blk.1.ffn_gate.weight
Q8_0
Q8_0
[2048, 11008]
blk.1.ffn_up.weight
Q8_0
Q8_0
[2048, 11008]
blk.2
blk.2.attn_k.weight
Q8_0
Q8_0
[2048, 512]
blk.2.attn_norm.weight
F32
F32
[2048]
blk.2.attn_output.weight
Q8_0
Q8_0
[2048, 2048]
blk.2.attn_q.weight
Q8_0
Q8_0
[2048, 2048]
blk.2.attn_v.weight
Q8_0
Q8_0
[2048, 512]
blk.2.ffn_down.weight
Q8_0
Q8_0
[11008, 2048]
blk.2.ffn_gate.weight
Q8_0
Q8_0
[2048, 11008]
blk.2.ffn_up.weight
Q8_0
Q8_0
[2048, 11008]
blk.3
blk.3.attn_k.weight
Q8_0
Q8_0
[2048, 512]
blk.3.attn_norm.weight
F32
F32
[2048]
blk.3.attn_output.weight
Q8_0
Q8_0
[2048, 2048]
blk.3.attn_q.weight
Q8_0
Q8_0
[2048, 2048]
blk.3.attn_v.weight
Q8_0
Q8_0
[2048, 512]
blk.3.ffn_down.weight
Q8_0
Q8_0
[11008, 2048]
blk.3.ffn_gate.weight
Q8_0
Q8_0
[2048, 11008]
blk.3.ffn_up.weight
Q8_0
Q8_0
[2048, 11008]
blk.4
blk.4.attn_k.weight
Q8_0
Q8_0
[2048, 512]
blk.4.attn_norm.weight
F32
F32
[2048]
blk.4.attn_output.weight
Q8_0
Q8_0
[2048, 2048]
blk.4.attn_q.weight
Q8_0
Q8_0
[2048, 2048]
blk.4.attn_v.weight
Q8_0
Q8_0
[2048, 512]
blk.4.ffn_down.weight
Q8_0
Q8_0
[11008, 2048]
blk.4.ffn_gate.weight
Q8_0
Q8_0
[2048, 11008]
blk.4.ffn_up.weight
Q8_0
Q8_0
[2048, 11008]
blk.5
blk.5.attn_k.weight
Q8_0
Q8_0
[2048, 512]
blk.5.attn_norm.weight
F32
F32
[2048]
blk.5.attn_output.weight
Q8_0
Q8_0
[2048, 2048]
blk.5.attn_q.weight
Q8_0
Q8_0
[2048, 2048]
blk.5.attn_v.weight
Q8_0
Q8_0
[2048, 512]
blk.5.ffn_down.weight
Q8_0
Q8_0
[11008, 2048]
blk.5.ffn_gate.weight
Q8_0
Q8_0
[2048, 11008]
blk.5.ffn_up.weight
Q8_0
Q8_0
[2048, 11008]
blk.6
blk.6.attn_k.weight
Q8_0
Q8_0
[2048, 512]
blk.6.attn_norm.weight
F32
F32
[2048]
blk.6.attn_output.weight
Q8_0
Q8_0
[2048, 2048]
blk.6.attn_q.weight
Q8_0
Q8_0
[2048, 2048]
blk.6.attn_v.weight
Q8_0
Q8_0
[2048, 512]
blk.6.ffn_down.weight
Q8_0
Q8_0
[11008, 2048]
blk.6.ffn_gate.weight
Q8_0
Q8_0
[2048, 11008]
blk.6.ffn_up.weight
Q8_0
Q8_0
[2048, 11008]
blk.7
blk.7.attn_k.weight
Q8_0
Q8_0
[2048, 512]
blk.7.attn_norm.weight
F32
F32
[2048]
blk.7.attn_output.weight
Q8_0
Q8_0
[2048, 2048]
blk.7.attn_q.weight
Q8_0
Q8_0
[2048, 2048]
blk.7.attn_v.weight
Q8_0
Q8_0
[2048, 512]
blk.7.ffn_down.weight
Q8_0
Q8_0
[11008, 2048]
blk.7.ffn_gate.weight
Q8_0
Q8_0
[2048, 11008]
blk.7.ffn_up.weight
Q8_0
Q8_0
[2048, 11008]
blk.8
blk.8.attn_k.weight
Q8_0
Q8_0
[2048, 512]
blk.8.attn_norm.weight
F32
F32
[2048]
blk.8.attn_output.weight
Q8_0
Q8_0
[2048, 2048]
blk.8.attn_q.weight
Q8_0
Q8_0
[2048, 2048]
blk.8.attn_v.weight
Q8_0
Q8_0
[2048, 512]
blk.8.ffn_down.weight
Q8_0
Q8_0
[11008, 2048]
blk.8.ffn_gate.weight
Q8_0
Q8_0
[2048, 11008]
blk.8.ffn_up.weight
Q8_0
Q8_0
[2048, 11008]
blk.9
blk.9.attn_k.weight
Q8_0
Q8_0
[2048, 512]
blk.9.attn_norm.weight
F32
F32
[2048]
blk.9.attn_output.weight
Q8_0
Q8_0
[2048, 2048]
blk.9.attn_q.weight
Q8_0
Q8_0
[2048, 2048]
blk.9.attn_v.weight
Q8_0
Q8_0
[2048, 512]
blk.9.ffn_down.weight
Q8_0
Q8_0
[11008, 2048]
blk.9.ffn_gate.weight
Q8_0
Q8_0
[2048, 11008]
blk.9.ffn_up.weight
Q8_0
Q8_0
[2048, 11008]
blk.10
blk.10.attn_k.weight
Q8_0
Q8_0
[2048, 512]
blk.10.attn_norm.weight
F32
F32
[2048]
blk.10.attn_output.weight
Q8_0
Q8_0
[2048, 2048]
blk.10.attn_q.weight
Q8_0
Q8_0
[2048, 2048]
blk.10.attn_v.weight
Q8_0
Q8_0
[2048, 512]
blk.10.ffn_down.weight
Q8_0
Q8_0
[11008, 2048]
blk.10.ffn_gate.weight
Q8_0
Q8_0
[2048, 11008]
blk.10.ffn_up.weight
Q8_0
Q8_0
[2048, 11008]
blk.11
blk.11.attn_k.weight
Q8_0
Q8_0
[2048, 512]
blk.11.attn_norm.weight
F32
F32
[2048]
blk.11.attn_output.weight
Q8_0
Q8_0
[2048, 2048]
blk.11.attn_q.weight
Q8_0
Q8_0
[2048, 2048]
blk.11.attn_v.weight
Q8_0
Q8_0
[2048, 512]
blk.11.ffn_down.weight
Q8_0
Q8_0
[11008, 2048]
blk.11.ffn_gate.weight
Q8_0
Q8_0
[2048, 11008]
blk.11.ffn_up.weight
Q8_0
Q8_0
[2048, 11008]
blk.12
blk.12.attn_k.weight
Q8_0
Q8_0
[2048, 512]
blk.12.attn_norm.weight
F32
F32
[2048]
blk.12.attn_output.weight
Q8_0
Q8_0
[2048, 2048]
blk.12.attn_q.weight
Q8_0
Q8_0
[2048, 2048]
blk.12.attn_v.weight
Q8_0
Q8_0
[2048, 512]
blk.12.ffn_down.weight
Q8_0
Q8_0
[11008, 2048]
blk.12.ffn_gate.weight
Q8_0
Q8_0
[2048, 11008]
blk.12.ffn_up.weight
Q8_0
Q8_0
[2048, 11008]
blk.13
blk.13.attn_k.weight
Q8_0
Q8_0
[2048, 512]
blk.13.attn_norm.weight
F32
F32
[2048]
blk.13.attn_output.weight
Q8_0
Q8_0
[2048, 2048]
blk.13.attn_q.weight
Q8_0
Q8_0
[2048, 2048]
blk.13.attn_v.weight
Q8_0
Q8_0
[2048, 512]
blk.13.ffn_down.weight
Q8_0
Q8_0
[11008, 2048]
blk.13.ffn_gate.weight
Q8_0
Q8_0
[2048, 11008]
blk.13.ffn_up.weight
Q8_0
Q8_0
[2048, 11008]
blk.14
blk.14.attn_k.weight
Q8_0
Q8_0
[2048, 512]
blk.14.attn_norm.weight
F32
F32
[2048]
blk.14.attn_output.weight
Q8_0
Q8_0
[2048, 2048]
blk.14.attn_q.weight
Q8_0
Q8_0
[2048, 2048]
blk.14.attn_v.weight
Q8_0
Q8_0
[2048, 512]
blk.14.ffn_down.weight
Q8_0
Q8_0
[11008, 2048]
blk.14.ffn_gate.weight
Q8_0
Q8_0
[2048, 11008]
blk.14.ffn_up.weight
Q8_0
Q8_0
[2048, 11008]
blk.15
blk.15.attn_k.weight
Q8_0
Q8_0
[2048, 512]
blk.15.attn_norm.weight
F32
F32
[2048]
blk.15.attn_output.weight
Q8_0
Q8_0
[2048, 2048]
blk.15.attn_q.weight
Q8_0
Q8_0
[2048, 2048]
blk.15.attn_v.weight
Q8_0
Q8_0
[2048, 512]
blk.15.ffn_down.weight
Q8_0
Q8_0
[11008, 2048]
blk.15.ffn_gate.weight
Q8_0
Q8_0
[2048, 11008]
blk.15.ffn_up.weight
Q8_0
Q8_0
[2048, 11008]
blk.16
blk.16.attn_k.weight
Q8_0
Q8_0
[2048, 512]
blk.16.attn_norm.weight
F32
F32
[2048]
blk.16.attn_output.weight
Q8_0
Q8_0
[2048, 2048]
blk.16.attn_q.weight
Q8_0
Q8_0
[2048, 2048]
blk.16.attn_v.weight
Q8_0
Q8_0
[2048, 512]
blk.16.ffn_down.weight
Q8_0
Q8_0
[11008, 2048]
blk.16.ffn_gate.weight
Q8_0
Q8_0
[2048, 11008]
blk.16.ffn_up.weight
Q8_0
Q8_0
[2048, 11008]
blk.17
blk.17.attn_k.weight
Q8_0
Q8_0
[2048, 512]
blk.17.attn_norm.weight
F32
F32
[2048]
blk.17.attn_output.weight
Q8_0
Q8_0
[2048, 2048]
blk.17.attn_q.weight
Q8_0
Q8_0
[2048, 2048]
blk.17.attn_v.weight
Q8_0
Q8_0
[2048, 512]
blk.17.ffn_down.weight
Q8_0
Q8_0
[11008, 2048]
blk.17.ffn_gate.weight
Q8_0
Q8_0
[2048, 11008]
blk.17.ffn_up.weight
Q8_0
Q8_0
[2048, 11008]
blk.18
blk.18.attn_k.weight
Q8_0
Q8_0
[2048, 512]
blk.18.attn_norm.weight
F32
F32
[2048]
blk.18.attn_output.weight
Q8_0
Q8_0
[2048, 2048]
blk.18.attn_q.weight
Q8_0
Q8_0
[2048, 2048]
blk.18.attn_v.weight
Q8_0
Q8_0
[2048, 512]
blk.18.ffn_down.weight
Q8_0
Q8_0
[11008, 2048]
blk.18.ffn_gate.weight
Q8_0
Q8_0
[2048, 11008]
blk.18.ffn_up.weight
Q8_0
Q8_0
[2048, 11008]
blk.19
blk.19.attn_k.weight
Q8_0
Q8_0
[2048, 512]
blk.19.attn_norm.weight
F32
F32
[2048]
blk.19.attn_output.weight
Q8_0
Q8_0
[2048, 2048]
blk.19.attn_q.weight
Q8_0
Q8_0
[2048, 2048]
blk.19.attn_v.weight
Q8_0
Q8_0
[2048, 512]
blk.19.ffn_down.weight
Q8_0
Q8_0
[11008, 2048]
blk.19.ffn_gate.weight
Q8_0
Q8_0
[2048, 11008]
blk.19.ffn_up.weight
Q8_0
Q8_0
[2048, 11008]
blk.20
blk.20.attn_k.weight
Q8_0
Q8_0
[2048, 512]
blk.20.attn_norm.weight
F32
F32
[2048]
blk.20.attn_output.weight
Q8_0
Q8_0
[2048, 2048]
blk.20.attn_q.weight
Q8_0
Q8_0
[2048, 2048]
blk.20.attn_v.weight
Q8_0
Q8_0
[2048, 512]
blk.20.ffn_down.weight
Q8_0
Q8_0
[11008, 2048]
blk.20.ffn_gate.weight
Q8_0
Q8_0
[2048, 11008]
blk.20.ffn_up.weight
Q8_0
Q8_0
[2048, 11008]
blk.21
blk.21.attn_k.weight
Q8_0
Q8_0
[2048, 512]
blk.21.attn_norm.weight
F32
F32
[2048]
blk.21.attn_output.weight
Q8_0
Q8_0
[2048, 2048]
blk.21.attn_q.weight
Q8_0
Q8_0
[2048, 2048]
blk.21.attn_v.weight
Q8_0
Q8_0
[2048, 512]
blk.21.ffn_down.weight
Q8_0
Q8_0
[11008, 2048]
blk.21.ffn_gate.weight
Q8_0
Q8_0
[2048, 11008]
blk.21.ffn_up.weight
Q8_0
Q8_0
[2048, 11008]
blk.22
blk.22.attn_k.weight
Q8_0
Q8_0
[2048, 512]
blk.22.attn_norm.weight
F32
F32
[2048]
blk.22.attn_output.weight
Q8_0
Q8_0
[2048, 2048]
blk.22.attn_q.weight
Q8_0
Q8_0
[2048, 2048]
blk.22.attn_v.weight
Q8_0
Q8_0
[2048, 512]
blk.22.ffn_down.weight
Q8_0
Q8_0
[11008, 2048]
blk.22.ffn_gate.weight
Q8_0
Q8_0
[2048, 11008]
blk.22.ffn_up.weight
Q8_0
Q8_0
[2048, 11008]
blk.23
blk.23.attn_k.weight
Q8_0
Q8_0
[2048, 512]
blk.23.attn_norm.weight
F32
F32
[2048]
blk.23.attn_output.weight
Q8_0
Q8_0
[2048, 2048]
blk.23.attn_q.weight
Q8_0
Q8_0
[2048, 2048]
blk.23.attn_v.weight
Q8_0
Q8_0
[2048, 512]
blk.23.ffn_down.weight
Q8_0
Q8_0
[11008, 2048]
blk.23.ffn_gate.weight
Q8_0
Q8_0
[2048, 11008]
blk.23.ffn_up.weight
Q8_0
Q8_0
[2048, 11008]
blk.24
blk.24.attn_k.weight
Q8_0
Q8_0
[2048, 512]
blk.24.attn_norm.weight
F32
F32
[2048]
blk.24.attn_output.weight
Q8_0
Q8_0
[2048, 2048]
blk.24.attn_q.weight
Q8_0
Q8_0
[2048, 2048]
blk.24.attn_v.weight
Q8_0
Q8_0
[2048, 512]
blk.24.ffn_down.weight
Q8_0
Q8_0
[11008, 2048]
blk.24.ffn_gate.weight
Q8_0
Q8_0
[2048, 11008]
blk.24.ffn_up.weight
Q8_0
Q8_0
[2048, 11008]
blk.25
blk.25.attn_k.weight
Q8_0
Q8_0
[2048, 512]
blk.25.attn_norm.weight
F32
F32
[2048]
blk.25.attn_output.weight
Q8_0
Q8_0
[2048, 2048]
blk.25.attn_q.weight
Q8_0
Q8_0
[2048, 2048]
blk.25.attn_v.weight
Q8_0
Q8_0
[2048, 512]
blk.25.ffn_down.weight
Q8_0
Q8_0
[11008, 2048]
blk.25.ffn_gate.weight
Q8_0
Q8_0
[2048, 11008]
blk.25.ffn_up.weight
Q8_0
Q8_0
[2048, 11008]
blk.26
blk.26.attn_k.weight
Q8_0
Q8_0
[2048, 512]
blk.26.attn_norm.weight
F32
F32
[2048]
blk.26.attn_output.weight
Q8_0
Q8_0
[2048, 2048]
blk.26.attn_q.weight
Q8_0
Q8_0
[2048, 2048]
blk.26.attn_v.weight
Q8_0
Q8_0
[2048, 512]
blk.26.ffn_down.weight
Q8_0
Q8_0
[11008, 2048]
blk.26.ffn_gate.weight
Q8_0
Q8_0
[2048, 11008]
blk.26.ffn_up.weight
Q8_0
Q8_0
[2048, 11008]
blk.27
blk.27.attn_k.weight
Q8_0
Q8_0
[2048, 512]
blk.27.attn_norm.weight
F32
F32
[2048]
blk.27.attn_output.weight
Q8_0
Q8_0
[2048, 2048]
blk.27.attn_q.weight
Q8_0
Q8_0
[2048, 2048]
blk.27.attn_v.weight
Q8_0
Q8_0
[2048, 512]
blk.27.ffn_down.weight
Q8_0
Q8_0
[11008, 2048]
blk.27.ffn_gate.weight
Q8_0
Q8_0
[2048, 11008]
blk.27.ffn_up.weight
Q8_0
Q8_0
[2048, 11008]
blk.28
blk.28.attn_k.weight
Q8_0
Q8_0
[2048, 512]
blk.28.attn_norm.weight
F32
F32
[2048]
blk.28.attn_output.weight
Q8_0
Q8_0
[2048, 2048]
blk.28.attn_q.weight
Q8_0
Q8_0
[2048, 2048]
blk.28.attn_v.weight
Q8_0
Q8_0
[2048, 512]
blk.28.ffn_down.weight
Q8_0
Q8_0
[11008, 2048]
blk.28.ffn_gate.weight
Q8_0
Q8_0
[2048, 11008]
blk.28.ffn_up.weight
Q8_0
Q8_0
[2048, 11008]
blk.29
blk.29.attn_k.weight
Q8_0
Q8_0
[2048, 512]
blk.29.attn_norm.weight
F32
F32
[2048]
blk.29.attn_output.weight
Q8_0
Q8_0
[2048, 2048]
blk.29.attn_q.weight
Q8_0
Q8_0
[2048, 2048]
blk.29.attn_v.weight
Q8_0
Q8_0
[2048, 512]
blk.29.ffn_down.weight
Q8_0
Q8_0
[11008, 2048]
blk.29.ffn_gate.weight
Q8_0
Q8_0
[2048, 11008]
blk.29.ffn_up.weight
Q8_0
Q8_0
[2048, 11008]
blk.30
blk.30.attn_k.weight
Q8_0
Q8_0
[2048, 512]
blk.30.attn_norm.weight
F32
F32
[2048]
blk.30.attn_output.weight
Q8_0
Q8_0
[2048, 2048]
blk.30.attn_q.weight
Q8_0
Q8_0
[2048, 2048]
blk.30.attn_v.weight
Q8_0
Q8_0
[2048, 512]
blk.30.ffn_down.weight
Q8_0
Q8_0
[11008, 2048]
blk.30.ffn_gate.weight
Q8_0
Q8_0
[2048, 11008]
blk.30.ffn_up.weight
Q8_0
Q8_0
[2048, 11008]
blk.31
blk.31.attn_k.weight
Q8_0
Q8_0
[2048, 512]
blk.31.attn_norm.weight
F32
F32
[2048]
blk.31.attn_output.weight
Q8_0
Q8_0
[2048, 2048]
blk.31.attn_q.weight
Q8_0
Q8_0
[2048, 2048]
blk.31.attn_v.weight
Q8_0
Q8_0
[2048, 512]
blk.31.ffn_down.weight
Q8_0
Q8_0
[11008, 2048]
blk.31.ffn_gate.weight
Q8_0
Q8_0
[2048, 11008]
blk.31.ffn_up.weight
Q8_0
Q8_0
[2048, 11008]
blk.32
blk.32.attn_k.weight
Q8_0
Q8_0
[2048, 512]
blk.32.attn_norm.weight
F32
F32
[2048]
blk.32.attn_output.weight
Q8_0
Q8_0
[2048, 2048]
blk.32.attn_q.weight
Q8_0
Q8_0
[2048, 2048]
blk.32.attn_v.weight
Q8_0
Q8_0
[2048, 512]
blk.32.ffn_down.weight
Q8_0
Q8_0
[11008, 2048]
blk.32.ffn_gate.weight
Q8_0
Q8_0
[2048, 11008]
blk.32.ffn_up.weight
Q8_0
Q8_0
[2048, 11008]
blk.33
blk.33.attn_k.weight
Q8_0
Q8_0
[2048, 512]
blk.33.attn_norm.weight
F32
F32
[2048]
blk.33.attn_output.weight
Q8_0
Q8_0
[2048, 2048]
blk.33.attn_q.weight
Q8_0
Q8_0
[2048, 2048]
blk.33.attn_v.weight
Q8_0
Q8_0
[2048, 512]
blk.33.ffn_down.weight
Q8_0
Q8_0
[11008, 2048]
blk.33.ffn_gate.weight
Q8_0
Q8_0
[2048, 11008]
blk.33.ffn_up.weight
Q8_0
Q8_0
[2048, 11008]
blk.34
blk.34.attn_k.weight
Q8_0
Q8_0
[2048, 512]
blk.34.attn_norm.weight
F32
F32
[2048]
blk.34.attn_output.weight
Q8_0
Q8_0
[2048, 2048]
blk.34.attn_q.weight
Q8_0
Q8_0
[2048, 2048]
blk.34.attn_v.weight
Q8_0
Q8_0
[2048, 512]
blk.34.ffn_down.weight
Q8_0
Q8_0
[11008, 2048]
blk.34.ffn_gate.weight
Q8_0
Q8_0
[2048, 11008]
blk.34.ffn_up.weight
Q8_0
Q8_0
[2048, 11008]
blk.35
blk.35.attn_k.weight
Q8_0
Q8_0
[2048, 512]
blk.35.attn_norm.weight
F32
F32
[2048]
blk.35.attn_output.weight
Q8_0
Q8_0
[2048, 2048]
blk.35.attn_q.weight
Q8_0
Q8_0
[2048, 2048]
blk.35.attn_v.weight
Q8_0
Q8_0
[2048, 512]
blk.35.ffn_down.weight
Q8_0
Q8_0
[11008, 2048]
blk.35.ffn_gate.weight
Q8_0
Q8_0
[2048, 11008]
blk.35.ffn_up.weight
Q8_0
Q8_0
[2048, 11008]
output_norm.weight
F32
F32
[2048]