Models
Docs
Pricing
Sign in
Download
Models
Download
Docs
Pricing
Sign in
peeyush_16
/
supra-router
:latest
6
Downloads
Updated
2 weeks ago
Fast 51M-parameter Llama-based router that classifies prompts and dispatches them to the right downstream model or tool.
Fast 51M-parameter Llama-based router that classifies prompts and dispatches them to the right downstream model or tool.
Cancel
supra-router:latest
...
/
model
d258ecdf5e73 · 105MB
Metadata
general.architecture
llama
llama
general.file_type
F16
F16
llama.attention.head_count
8
8
llama.attention.head_count_kv
4
4
llama.attention.key_length
64
64
llama.attention.layer_norm_rms_epsilon
1e-06
1e-06
llama.attention.value_length
64
64
llama.block_count
12
12
llama.context_length
5120
5120
llama.embedding_length
512
512
llama.feed_forward_length
1408
1408
llama.rope.dimension_count
64
64
llama.rope.freq_base
10000
10000
llama.rope.scaling.factor
1
1
llama.rope.scaling.type
linear
linear
llama.vocab_size
32000
32000
tokenizer.ggml.bos_token_id
0
0
tokenizer.ggml.eos_token_id
2
2
tokenizer.ggml.merges
[Ġ t, Ġ a, h e, i n, r e, ...]
[Ġ t, Ġ a, h e, i n, r e, ...]
tokenizer.ggml.model
gpt2
gpt2
tokenizer.ggml.padding_token_id
1
1
tokenizer.ggml.pre
gpt-2
gpt-2
tokenizer.ggml.token_type
[3, 3, 3, 3, 1, ...]
[3, 3, 3, 3, 1, ...]
tokenizer.ggml.tokens
[<s>, <pad>, </s>, <unk>, <mask>, ...]
[<s>, <pad>, </s>, <unk>, <mask>, ...]
tokenizer.ggml.unknown_token_id
3
3
Tensor
Name
Type
Shape
token_embd.weight
F16
F16
[512, 32000]
blk.0
blk.0.attn_k.weight
F16
F16
[512, 256]
blk.0.attn_norm.weight
F32
F32
[512]
blk.0.attn_output.weight
F16
F16
[512, 512]
blk.0.attn_q.weight
F16
F16
[512, 512]
blk.0.attn_v.weight
F16
F16
[512, 256]
blk.0.ffn_down.weight
F16
F16
[1408, 512]
blk.0.ffn_gate.weight
F16
F16
[512, 1408]
blk.0.ffn_norm.weight
F32
F32
[512]
blk.0.ffn_up.weight
F16
F16
[512, 1408]
blk.1
blk.1.attn_k.weight
F16
F16
[512, 256]
blk.1.attn_norm.weight
F32
F32
[512]
blk.1.attn_output.weight
F16
F16
[512, 512]
blk.1.attn_q.weight
F16
F16
[512, 512]
blk.1.attn_v.weight
F16
F16
[512, 256]
blk.1.ffn_down.weight
F16
F16
[1408, 512]
blk.1.ffn_gate.weight
F16
F16
[512, 1408]
blk.1.ffn_norm.weight
F32
F32
[512]
blk.1.ffn_up.weight
F16
F16
[512, 1408]
blk.2
blk.2.attn_k.weight
F16
F16
[512, 256]
blk.2.attn_norm.weight
F32
F32
[512]
blk.2.attn_output.weight
F16
F16
[512, 512]
blk.2.attn_q.weight
F16
F16
[512, 512]
blk.2.attn_v.weight
F16
F16
[512, 256]
blk.2.ffn_down.weight
F16
F16
[1408, 512]
blk.2.ffn_gate.weight
F16
F16
[512, 1408]
blk.2.ffn_norm.weight
F32
F32
[512]
blk.2.ffn_up.weight
F16
F16
[512, 1408]
blk.3
blk.3.attn_k.weight
F16
F16
[512, 256]
blk.3.attn_norm.weight
F32
F32
[512]
blk.3.attn_output.weight
F16
F16
[512, 512]
blk.3.attn_q.weight
F16
F16
[512, 512]
blk.3.attn_v.weight
F16
F16
[512, 256]
blk.3.ffn_down.weight
F16
F16
[1408, 512]
blk.3.ffn_gate.weight
F16
F16
[512, 1408]
blk.3.ffn_norm.weight
F32
F32
[512]
blk.3.ffn_up.weight
F16
F16
[512, 1408]
blk.4
blk.4.attn_k.weight
F16
F16
[512, 256]
blk.4.attn_norm.weight
F32
F32
[512]
blk.4.attn_output.weight
F16
F16
[512, 512]
blk.4.attn_q.weight
F16
F16
[512, 512]
blk.4.attn_v.weight
F16
F16
[512, 256]
blk.4.ffn_down.weight
F16
F16
[1408, 512]
blk.4.ffn_gate.weight
F16
F16
[512, 1408]
blk.4.ffn_norm.weight
F32
F32
[512]
blk.4.ffn_up.weight
F16
F16
[512, 1408]
blk.5
blk.5.attn_k.weight
F16
F16
[512, 256]
blk.5.attn_norm.weight
F32
F32
[512]
blk.5.attn_output.weight
F16
F16
[512, 512]
blk.5.attn_q.weight
F16
F16
[512, 512]
blk.5.attn_v.weight
F16
F16
[512, 256]
blk.5.ffn_down.weight
F16
F16
[1408, 512]
blk.5.ffn_gate.weight
F16
F16
[512, 1408]
blk.5.ffn_norm.weight
F32
F32
[512]
blk.5.ffn_up.weight
F16
F16
[512, 1408]
blk.6
blk.6.attn_k.weight
F16
F16
[512, 256]
blk.6.attn_norm.weight
F32
F32
[512]
blk.6.attn_output.weight
F16
F16
[512, 512]
blk.6.attn_q.weight
F16
F16
[512, 512]
blk.6.attn_v.weight
F16
F16
[512, 256]
blk.6.ffn_down.weight
F16
F16
[1408, 512]
blk.6.ffn_gate.weight
F16
F16
[512, 1408]
blk.6.ffn_norm.weight
F32
F32
[512]
blk.6.ffn_up.weight
F16
F16
[512, 1408]
blk.7
blk.7.attn_k.weight
F16
F16
[512, 256]
blk.7.attn_norm.weight
F32
F32
[512]
blk.7.attn_output.weight
F16
F16
[512, 512]
blk.7.attn_q.weight
F16
F16
[512, 512]
blk.7.attn_v.weight
F16
F16
[512, 256]
blk.7.ffn_down.weight
F16
F16
[1408, 512]
blk.7.ffn_gate.weight
F16
F16
[512, 1408]
blk.7.ffn_norm.weight
F32
F32
[512]
blk.7.ffn_up.weight
F16
F16
[512, 1408]
blk.8
blk.8.attn_k.weight
F16
F16
[512, 256]
blk.8.attn_norm.weight
F32
F32
[512]
blk.8.attn_output.weight
F16
F16
[512, 512]
blk.8.attn_q.weight
F16
F16
[512, 512]
blk.8.attn_v.weight
F16
F16
[512, 256]
blk.8.ffn_down.weight
F16
F16
[1408, 512]
blk.8.ffn_gate.weight
F16
F16
[512, 1408]
blk.8.ffn_norm.weight
F32
F32
[512]
blk.8.ffn_up.weight
F16
F16
[512, 1408]
blk.9
blk.9.attn_k.weight
F16
F16
[512, 256]
blk.9.attn_norm.weight
F32
F32
[512]
blk.9.attn_output.weight
F16
F16
[512, 512]
blk.9.attn_q.weight
F16
F16
[512, 512]
blk.9.attn_v.weight
F16
F16
[512, 256]
blk.9.ffn_down.weight
F16
F16
[1408, 512]
blk.9.ffn_gate.weight
F16
F16
[512, 1408]
blk.9.ffn_norm.weight
F32
F32
[512]
blk.9.ffn_up.weight
F16
F16
[512, 1408]
blk.10
blk.10.attn_k.weight
F16
F16
[512, 256]
blk.10.attn_norm.weight
F32
F32
[512]
blk.10.attn_output.weight
F16
F16
[512, 512]
blk.10.attn_q.weight
F16
F16
[512, 512]
blk.10.attn_v.weight
F16
F16
[512, 256]
blk.10.ffn_down.weight
F16
F16
[1408, 512]
blk.10.ffn_gate.weight
F16
F16
[512, 1408]
blk.10.ffn_norm.weight
F32
F32
[512]
blk.10.ffn_up.weight
F16
F16
[512, 1408]
blk.11
blk.11.attn_k.weight
F16
F16
[512, 256]
blk.11.attn_norm.weight
F32
F32
[512]
blk.11.attn_output.weight
F16
F16
[512, 512]
blk.11.attn_q.weight
F16
F16
[512, 512]
blk.11.attn_v.weight
F16
F16
[512, 256]
blk.11.ffn_down.weight
F16
F16
[1408, 512]
blk.11.ffn_gate.weight
F16
F16
[512, 1408]
blk.11.ffn_norm.weight
F32
F32
[512]
blk.11.ffn_up.weight
F16
F16
[512, 1408]
output_norm.weight
F32
F32
[512]