-
general.architecture
bloom
-
general.file_type
F16
-
bloom.attention.head_count
32
-
bloom.attention.head_count_kv
32
-
bloom.attention.layer_norm_epsilon
1e-05
-
bloom.block_count
30
-
bloom.context_length
2048
-
bloom.embedding_length
2560
-
bloom.feed_forward_length
10240
-
tokenizer.ggml.bos_token_id
1
-
tokenizer.ggml.eos_token_id
2
-
tokenizer.ggml.merges
[à ¤ à ¦ Ġ à à ¥ Ġ Ġ ...]
-
tokenizer.ggml.model
gpt2
-
tokenizer.ggml.padding_token_id
3
-
tokenizer.ggml.token_type
[3 3 3 3 1 ...]
-
tokenizer.ggml.tokens
[<unk> <s> </s> <pad> ! ...]
-
tokenizer.ggml.unknown_token_id
0
-
Name
Type
Shape
-
token_embd.weight
F16
[2560 250880]
-
token_embd_norm.bias
F32
[2560]
-
token_embd_norm.weight
F32
[2560]
-
blk.0.attn_norm.bias
F32
[2560]
-
blk.0.attn_norm.weight
F32
[2560]
-
blk.0.ffn_down.bias
F32
[2560]
-
blk.0.ffn_down.weight
F16
[10240 2560]
-
blk.0.ffn_up.bias
F32
[10240]
-
blk.0.ffn_up.weight
F16
[2560 10240]
-
blk.0.ffn_norm.bias
F32
[2560]
-
blk.0.ffn_norm.weight
F32
[2560]
-
blk.0.attn_output.bias
F32
[2560]
-
blk.0.attn_output.weight
F16
[2560 2560]
-
blk.0.attn_qkv.bias
F32
[7680]
-
blk.0.attn_qkv.weight
F16
[2560 7680]
-
blk.1.attn_norm.bias
F32
[2560]
-
blk.1.attn_norm.weight
F32
[2560]
-
blk.1.ffn_down.bias
F32
[2560]
-
blk.1.ffn_down.weight
F16
[10240 2560]
-
blk.1.ffn_up.bias
F32
[10240]
-
blk.1.ffn_up.weight
F16
[2560 10240]
-
blk.1.ffn_norm.bias
F32
[2560]
-
blk.1.ffn_norm.weight
F32
[2560]
-
blk.1.attn_output.bias
F32
[2560]
-
blk.1.attn_output.weight
F16
[2560 2560]
-
blk.1.attn_qkv.bias
F32
[7680]
-
blk.1.attn_qkv.weight
F16
[2560 7680]
-
blk.2.attn_norm.bias
F32
[2560]
-
blk.2.attn_norm.weight
F32
[2560]
-
blk.2.ffn_down.bias
F32
[2560]
-
blk.2.ffn_down.weight
F16
[10240 2560]
-
blk.2.ffn_up.bias
F32
[10240]
-
blk.2.ffn_up.weight
F16
[2560 10240]
-
blk.2.ffn_norm.bias
F32
[2560]
-
blk.2.ffn_norm.weight
F32
[2560]
-
blk.2.attn_output.bias
F32
[2560]
-
blk.2.attn_output.weight
F16
[2560 2560]
-
blk.2.attn_qkv.bias
F32
[7680]
-
blk.2.attn_qkv.weight
F16
[2560 7680]
-
blk.3.attn_norm.bias
F32
[2560]
-
blk.3.attn_norm.weight
F32
[2560]
-
blk.3.ffn_down.bias
F32
[2560]
-
blk.3.ffn_down.weight
F16
[10240 2560]
-
blk.3.ffn_up.bias
F32
[10240]
-
blk.3.ffn_up.weight
F16
[2560 10240]
-
blk.3.ffn_norm.bias
F32
[2560]
-
blk.3.ffn_norm.weight
F32
[2560]
-
blk.3.attn_output.bias
F32
[2560]
-
blk.3.attn_output.weight
F16
[2560 2560]
-
blk.3.attn_qkv.bias
F32
[7680]
-
blk.3.attn_qkv.weight
F16
[2560 7680]
-
blk.4.attn_norm.bias
F32
[2560]
-
blk.4.attn_norm.weight
F32
[2560]
-
blk.4.ffn_down.bias
F32
[2560]
-
blk.4.ffn_down.weight
F16
[10240 2560]
-
blk.4.ffn_up.bias
F32
[10240]
-
blk.4.ffn_up.weight
F16
[2560 10240]
-
blk.4.ffn_norm.bias
F32
[2560]
-
blk.4.ffn_norm.weight
F32
[2560]
-
blk.4.attn_output.bias
F32
[2560]
-
blk.4.attn_output.weight
F16
[2560 2560]
-
blk.4.attn_qkv.bias
F32
[7680]
-
blk.4.attn_qkv.weight
F16
[2560 7680]
-
blk.5.attn_norm.bias
F32
[2560]
-
blk.5.attn_norm.weight
F32
[2560]
-
blk.5.ffn_down.bias
F32
[2560]
-
blk.5.ffn_down.weight
F16
[10240 2560]
-
blk.5.ffn_up.bias
F32
[10240]
-
blk.5.ffn_up.weight
F16
[2560 10240]
-
blk.5.ffn_norm.bias
F32
[2560]
-
blk.5.ffn_norm.weight
F32
[2560]
-
blk.5.attn_output.bias
F32
[2560]
-
blk.5.attn_output.weight
F16
[2560 2560]
-
blk.5.attn_qkv.bias
F32
[7680]
-
blk.5.attn_qkv.weight
F16
[2560 7680]
-
blk.6.attn_norm.bias
F32
[2560]
-
blk.6.attn_norm.weight
F32
[2560]
-
blk.6.ffn_down.bias
F32
[2560]
-
blk.6.ffn_down.weight
F16
[10240 2560]
-
blk.6.ffn_up.bias
F32
[10240]
-
blk.6.ffn_up.weight
F16
[2560 10240]
-
blk.6.ffn_norm.bias
F32
[2560]
-
blk.6.ffn_norm.weight
F32
[2560]
-
blk.6.attn_output.bias
F32
[2560]
-
blk.6.attn_output.weight
F16
[2560 2560]
-
blk.6.attn_qkv.bias
F32
[7680]
-
blk.6.attn_qkv.weight
F16
[2560 7680]
-
blk.7.attn_norm.bias
F32
[2560]
-
blk.7.attn_norm.weight
F32
[2560]
-
blk.7.ffn_down.bias
F32
[2560]
-
blk.7.ffn_down.weight
F16
[10240 2560]
-
blk.7.ffn_up.bias
F32
[10240]
-
blk.7.ffn_up.weight
F16
[2560 10240]
-
blk.7.ffn_norm.bias
F32
[2560]
-
blk.7.ffn_norm.weight
F32
[2560]
-
blk.7.attn_output.bias
F32
[2560]
-
blk.7.attn_output.weight
F16
[2560 2560]
-
blk.7.attn_qkv.bias
F32
[7680]
-
blk.7.attn_qkv.weight
F16
[2560 7680]
-
blk.8.attn_norm.bias
F32
[2560]
-
blk.8.attn_norm.weight
F32
[2560]
-
blk.8.ffn_down.bias
F32
[2560]
-
blk.8.ffn_down.weight
F16
[10240 2560]
-
blk.8.ffn_up.bias
F32
[10240]
-
blk.8.ffn_up.weight
F16
[2560 10240]
-
blk.8.ffn_norm.bias
F32
[2560]
-
blk.8.ffn_norm.weight
F32
[2560]
-
blk.8.attn_output.bias
F32
[2560]
-
blk.8.attn_output.weight
F16
[2560 2560]
-
blk.8.attn_qkv.bias
F32
[7680]
-
blk.8.attn_qkv.weight
F16
[2560 7680]
-
blk.9.attn_norm.bias
F32
[2560]
-
blk.9.attn_norm.weight
F32
[2560]
-
blk.9.ffn_down.bias
F32
[2560]
-
blk.9.ffn_down.weight
F16
[10240 2560]
-
blk.9.ffn_up.bias
F32
[10240]
-
blk.9.ffn_up.weight
F16
[2560 10240]
-
blk.9.ffn_norm.bias
F32
[2560]
-
blk.9.ffn_norm.weight
F32
[2560]
-
blk.9.attn_output.bias
F32
[2560]
-
blk.9.attn_output.weight
F16
[2560 2560]
-
blk.9.attn_qkv.bias
F32
[7680]
-
blk.9.attn_qkv.weight
F16
[2560 7680]
-
blk.10.attn_norm.bias
F32
[2560]
-
blk.10.attn_norm.weight
F32
[2560]
-
blk.10.ffn_down.bias
F32
[2560]
-
blk.10.ffn_down.weight
F16
[10240 2560]
-
blk.10.ffn_up.bias
F32
[10240]
-
blk.10.ffn_up.weight
F16
[2560 10240]
-
blk.10.ffn_norm.bias
F32
[2560]
-
blk.10.ffn_norm.weight
F32
[2560]
-
blk.10.attn_output.bias
F32
[2560]
-
blk.10.attn_output.weight
F16
[2560 2560]
-
blk.10.attn_qkv.bias
F32
[7680]
-
blk.10.attn_qkv.weight
F16
[2560 7680]
-
blk.11.attn_norm.bias
F32
[2560]
-
blk.11.attn_norm.weight
F32
[2560]
-
blk.11.ffn_down.bias
F32
[2560]
-
blk.11.ffn_down.weight
F16
[10240 2560]
-
blk.11.ffn_up.bias
F32
[10240]
-
blk.11.ffn_up.weight
F16
[2560 10240]
-
blk.11.ffn_norm.bias
F32
[2560]
-
blk.11.ffn_norm.weight
F32
[2560]
-
blk.11.attn_output.bias
F32
[2560]
-
blk.11.attn_output.weight
F16
[2560 2560]
-
blk.11.attn_qkv.bias
F32
[7680]
-
blk.11.attn_qkv.weight
F16
[2560 7680]
-
blk.12.attn_norm.bias
F32
[2560]
-
blk.12.attn_norm.weight
F32
[2560]
-
blk.12.ffn_down.bias
F32
[2560]
-
blk.12.ffn_down.weight
F16
[10240 2560]
-
blk.12.ffn_up.bias
F32
[10240]
-
blk.12.ffn_up.weight
F16
[2560 10240]
-
blk.12.ffn_norm.bias
F32
[2560]
-
blk.12.ffn_norm.weight
F32
[2560]
-
blk.12.attn_output.bias
F32
[2560]
-
blk.12.attn_output.weight
F16
[2560 2560]
-
blk.12.attn_qkv.bias
F32
[7680]
-
blk.12.attn_qkv.weight
F16
[2560 7680]
-
blk.13.attn_norm.bias
F32
[2560]
-
blk.13.attn_norm.weight
F32
[2560]
-
blk.13.ffn_down.bias
F32
[2560]
-
blk.13.ffn_down.weight
F16
[10240 2560]
-
blk.13.ffn_up.bias
F32
[10240]
-
blk.13.ffn_up.weight
F16
[2560 10240]
-
blk.13.ffn_norm.bias
F32
[2560]
-
blk.13.ffn_norm.weight
F32
[2560]
-
blk.13.attn_output.bias
F32
[2560]
-
blk.13.attn_output.weight
F16
[2560 2560]
-
blk.13.attn_qkv.bias
F32
[7680]
-
blk.13.attn_qkv.weight
F16
[2560 7680]
-
blk.14.attn_norm.bias
F32
[2560]
-
blk.14.attn_norm.weight
F32
[2560]
-
blk.14.ffn_down.bias
F32
[2560]
-
blk.14.ffn_down.weight
F16
[10240 2560]
-
blk.14.ffn_up.bias
F32
[10240]
-
blk.14.ffn_up.weight
F16
[2560 10240]
-
blk.14.ffn_norm.bias
F32
[2560]
-
blk.14.ffn_norm.weight
F32
[2560]
-
blk.14.attn_output.bias
F32
[2560]
-
blk.14.attn_output.weight
F16
[2560 2560]
-
blk.14.attn_qkv.bias
F32
[7680]
-
blk.14.attn_qkv.weight
F16
[2560 7680]
-
blk.15.attn_norm.bias
F32
[2560]
-
blk.15.attn_norm.weight
F32
[2560]
-
blk.15.ffn_down.bias
F32
[2560]
-
blk.15.ffn_down.weight
F16
[10240 2560]
-
blk.15.ffn_up.bias
F32
[10240]
-
blk.15.ffn_up.weight
F16
[2560 10240]
-
blk.15.ffn_norm.bias
F32
[2560]
-
blk.15.ffn_norm.weight
F32
[2560]
-
blk.15.attn_output.bias
F32
[2560]
-
blk.15.attn_output.weight
F16
[2560 2560]
-
blk.15.attn_qkv.bias
F32
[7680]
-
blk.15.attn_qkv.weight
F16
[2560 7680]
-
blk.16.attn_norm.bias
F32
[2560]
-
blk.16.attn_norm.weight
F32
[2560]
-
blk.16.ffn_down.bias
F32
[2560]
-
blk.16.ffn_down.weight
F16
[10240 2560]
-
blk.16.ffn_up.bias
F32
[10240]
-
blk.16.ffn_up.weight
F16
[2560 10240]
-
blk.16.ffn_norm.bias
F32
[2560]
-
blk.16.ffn_norm.weight
F32
[2560]
-
blk.16.attn_output.bias
F32
[2560]
-
blk.16.attn_output.weight
F16
[2560 2560]
-
blk.16.attn_qkv.bias
F32
[7680]
-
blk.16.attn_qkv.weight
F16
[2560 7680]
-
blk.17.attn_norm.bias
F32
[2560]
-
blk.17.attn_norm.weight
F32
[2560]
-
blk.17.ffn_down.bias
F32
[2560]
-
blk.17.ffn_down.weight
F16
[10240 2560]
-
blk.17.ffn_up.bias
F32
[10240]
-
blk.17.ffn_up.weight
F16
[2560 10240]
-
blk.17.ffn_norm.bias
F32
[2560]
-
blk.17.ffn_norm.weight
F32
[2560]
-
blk.17.attn_output.bias
F32
[2560]
-
blk.17.attn_output.weight
F16
[2560 2560]
-
blk.17.attn_qkv.bias
F32
[7680]
-
blk.17.attn_qkv.weight
F16
[2560 7680]
-
blk.18.attn_norm.bias
F32
[2560]
-
blk.18.attn_norm.weight
F32
[2560]
-
blk.18.ffn_down.bias
F32
[2560]
-
blk.18.ffn_down.weight
F16
[10240 2560]
-
blk.18.ffn_up.bias
F32
[10240]
-
blk.18.ffn_up.weight
F16
[2560 10240]
-
blk.18.ffn_norm.bias
F32
[2560]
-
blk.18.ffn_norm.weight
F32
[2560]
-
blk.18.attn_output.bias
F32
[2560]
-
blk.18.attn_output.weight
F16
[2560 2560]
-
blk.18.attn_qkv.bias
F32
[7680]
-
blk.18.attn_qkv.weight
F16
[2560 7680]
-
blk.19.attn_norm.bias
F32
[2560]
-
blk.19.attn_norm.weight
F32
[2560]
-
blk.19.ffn_down.bias
F32
[2560]
-
blk.19.ffn_down.weight
F16
[10240 2560]
-
blk.19.ffn_up.bias
F32
[10240]
-
blk.19.ffn_up.weight
F16
[2560 10240]
-
blk.19.ffn_norm.bias
F32
[2560]
-
blk.19.ffn_norm.weight
F32
[2560]
-
blk.19.attn_output.bias
F32
[2560]
-
blk.19.attn_output.weight
F16
[2560 2560]
-
blk.19.attn_qkv.bias
F32
[7680]
-
blk.19.attn_qkv.weight
F16
[2560 7680]
-
blk.20.attn_norm.bias
F32
[2560]
-
blk.20.attn_norm.weight
F32
[2560]
-
blk.20.ffn_down.bias
F32
[2560]
-
blk.20.ffn_down.weight
F16
[10240 2560]
-
blk.20.ffn_up.bias
F32
[10240]
-
blk.20.ffn_up.weight
F16
[2560 10240]
-
blk.20.ffn_norm.bias
F32
[2560]
-
blk.20.ffn_norm.weight
F32
[2560]
-
blk.20.attn_output.bias
F32
[2560]
-
blk.20.attn_output.weight
F16
[2560 2560]
-
blk.20.attn_qkv.bias
F32
[7680]
-
blk.20.attn_qkv.weight
F16
[2560 7680]
-
blk.21.attn_norm.bias
F32
[2560]
-
blk.21.attn_norm.weight
F32
[2560]
-
blk.21.ffn_down.bias
F32
[2560]
-
blk.21.ffn_down.weight
F16
[10240 2560]
-
blk.21.ffn_up.bias
F32
[10240]
-
blk.21.ffn_up.weight
F16
[2560 10240]
-
blk.21.ffn_norm.bias
F32
[2560]
-
blk.21.ffn_norm.weight
F32
[2560]
-
blk.21.attn_output.bias
F32
[2560]
-
blk.21.attn_output.weight
F16
[2560 2560]
-
blk.21.attn_qkv.bias
F32
[7680]
-
blk.21.attn_qkv.weight
F16
[2560 7680]
-
blk.22.attn_norm.bias
F32
[2560]
-
blk.22.attn_norm.weight
F32
[2560]
-
blk.22.ffn_down.bias
F32
[2560]
-
blk.22.ffn_down.weight
F16
[10240 2560]
-
blk.22.ffn_up.bias
F32
[10240]
-
blk.22.ffn_up.weight
F16
[2560 10240]
-
blk.22.ffn_norm.bias
F32
[2560]
-
blk.22.ffn_norm.weight
F32
[2560]
-
blk.22.attn_output.bias
F32
[2560]
-
blk.22.attn_output.weight
F16
[2560 2560]
-
blk.22.attn_qkv.bias
F32
[7680]
-
blk.22.attn_qkv.weight
F16
[2560 7680]
-
blk.23.attn_norm.bias
F32
[2560]
-
blk.23.attn_norm.weight
F32
[2560]
-
blk.23.ffn_down.bias
F32
[2560]
-
blk.23.ffn_down.weight
F16
[10240 2560]
-
blk.23.ffn_up.bias
F32
[10240]
-
blk.23.ffn_up.weight
F16
[2560 10240]
-
blk.23.ffn_norm.bias
F32
[2560]
-
blk.23.ffn_norm.weight
F32
[2560]
-
blk.23.attn_output.bias
F32
[2560]
-
blk.23.attn_output.weight
F16
[2560 2560]
-
blk.23.attn_qkv.bias
F32
[7680]
-
blk.23.attn_qkv.weight
F16
[2560 7680]
-
blk.24.attn_norm.bias
F32
[2560]
-
blk.24.attn_norm.weight
F32
[2560]
-
blk.24.ffn_down.bias
F32
[2560]
-
blk.24.ffn_down.weight
F16
[10240 2560]
-
blk.24.ffn_up.bias
F32
[10240]
-
blk.24.ffn_up.weight
F16
[2560 10240]
-
blk.24.ffn_norm.bias
F32
[2560]
-
blk.24.ffn_norm.weight
F32
[2560]
-
blk.24.attn_output.bias
F32
[2560]
-
blk.24.attn_output.weight
F16
[2560 2560]
-
blk.24.attn_qkv.bias
F32
[7680]
-
blk.24.attn_qkv.weight
F16
[2560 7680]
-
blk.25.attn_norm.bias
F32
[2560]
-
blk.25.attn_norm.weight
F32
[2560]
-
blk.25.ffn_down.bias
F32
[2560]
-
blk.25.ffn_down.weight
F16
[10240 2560]
-
blk.25.ffn_up.bias
F32
[10240]
-
blk.25.ffn_up.weight
F16
[2560 10240]
-
blk.25.ffn_norm.bias
F32
[2560]
-
blk.25.ffn_norm.weight
F32
[2560]
-
blk.25.attn_output.bias
F32
[2560]
-
blk.25.attn_output.weight
F16
[2560 2560]
-
blk.25.attn_qkv.bias
F32
[7680]
-
blk.25.attn_qkv.weight
F16
[2560 7680]
-
blk.26.attn_norm.bias
F32
[2560]
-
blk.26.attn_norm.weight
F32
[2560]
-
blk.26.ffn_down.bias
F32
[2560]
-
blk.26.ffn_down.weight
F16
[10240 2560]
-
blk.26.ffn_up.bias
F32
[10240]
-
blk.26.ffn_up.weight
F16
[2560 10240]
-
blk.26.ffn_norm.bias
F32
[2560]
-
blk.26.ffn_norm.weight
F32
[2560]
-
blk.26.attn_output.bias
F32
[2560]
-
blk.26.attn_output.weight
F16
[2560 2560]
-
blk.26.attn_qkv.bias
F32
[7680]
-
blk.26.attn_qkv.weight
F16
[2560 7680]
-
blk.27.attn_norm.bias
F32
[2560]
-
blk.27.attn_norm.weight
F32
[2560]
-
blk.27.ffn_down.bias
F32
[2560]
-
blk.27.ffn_down.weight
F16
[10240 2560]
-
blk.27.ffn_up.bias
F32
[10240]
-
blk.27.ffn_up.weight
F16
[2560 10240]
-
blk.27.ffn_norm.bias
F32
[2560]
-
blk.27.ffn_norm.weight
F32
[2560]
-
blk.27.attn_output.bias
F32
[2560]
-
blk.27.attn_output.weight
F16
[2560 2560]
-
blk.27.attn_qkv.bias
F32
[7680]
-
blk.27.attn_qkv.weight
F16
[2560 7680]
-
blk.28.attn_norm.bias
F32
[2560]
-
blk.28.attn_norm.weight
F32
[2560]
-
blk.28.ffn_down.bias
F32
[2560]
-
blk.28.ffn_down.weight
F16
[10240 2560]
-
blk.28.ffn_up.bias
F32
[10240]
-
blk.28.ffn_up.weight
F16
[2560 10240]
-
blk.28.ffn_norm.bias
F32
[2560]
-
blk.28.ffn_norm.weight
F32
[2560]
-
blk.28.attn_output.bias
F32
[2560]
-
blk.28.attn_output.weight
F16
[2560 2560]
-
blk.28.attn_qkv.bias
F32
[7680]
-
blk.28.attn_qkv.weight
F16
[2560 7680]
-
blk.29.attn_norm.bias
F32
[2560]
-
blk.29.attn_norm.weight
F32
[2560]
-
blk.29.ffn_down.bias
F32
[2560]
-
blk.29.ffn_down.weight
F16
[10240 2560]
-
blk.29.ffn_up.bias
F32
[10240]
-
blk.29.ffn_up.weight
F16
[2560 10240]
-
blk.29.ffn_norm.bias
F32
[2560]
-
blk.29.ffn_norm.weight
F32
[2560]
-
blk.29.attn_output.bias
F32
[2560]
-
blk.29.attn_output.weight
F16
[2560 2560]
-
blk.29.attn_qkv.bias
F32
[7680]
-
blk.29.attn_qkv.weight
F16
[2560 7680]
-
output_norm.bias
F32
[2560]
-
output_norm.weight
F32
[2560]
-
output.weight
F16
[2560 250880]