69 Pulls Updated 6 months ago
abbfaae58dbb · 328MB
-
general.architecturegpt2
-
general.file_type1
-
gpt2.attention.head_count12
-
gpt2.attention.layer_norm_epsilon1e-05
-
gpt2.block_count12
-
gpt2.context_length1024
-
gpt2.embedding_length768
-
gpt2.feed_forward_length3072
-
tokenizer.ggml.bos_token_id50256
-
tokenizer.ggml.eos_token_id50256
-
tokenizer.ggml.merges[Ġ t, Ġ a, h e, i n, r e, ...]
-
tokenizer.ggml.modelgpt2
-
tokenizer.ggml.token_type[1, 1, 1, 1, 1, ...]
-
tokenizer.ggml.tokens[!, ", #, $, %, ...]
-
NameTypeShape
-
token_embd.weightF16[768, 50257]
-
blk.0.attn_norm.biasF32[768]
-
blk.0.attn_norm.weightF32[768]
-
blk.0.attn_output.biasF32[768]
-
blk.0.attn_output.weightF16[768, 768]
-
blk.0.attn_qkv.biasF32[2304]
-
blk.0.attn_qkv.weightF16[768, 2304]
-
blk.0.ffn_down.biasF32[768]
-
blk.0.ffn_down.weightF16[3072, 768]
-
blk.0.ffn_norm.biasF32[768]
-
blk.0.ffn_norm.weightF32[768]
-
blk.0.ffn_up.biasF32[3072]
-
blk.0.ffn_up.weightF16[768, 3072]
-
blk.1.attn_norm.biasF32[768]
-
blk.1.attn_norm.weightF32[768]
-
blk.1.attn_output.biasF32[768]
-
blk.1.attn_output.weightF16[768, 768]
-
blk.1.attn_qkv.biasF32[2304]
-
blk.1.attn_qkv.weightF16[768, 2304]
-
blk.1.ffn_down.biasF32[768]
-
blk.1.ffn_down.weightF16[3072, 768]
-
blk.1.ffn_norm.biasF32[768]
-
blk.1.ffn_norm.weightF32[768]
-
blk.1.ffn_up.biasF32[3072]
-
blk.1.ffn_up.weightF16[768, 3072]
-
blk.2.attn_norm.biasF32[768]
-
blk.2.attn_norm.weightF32[768]
-
blk.2.attn_output.biasF32[768]
-
blk.2.attn_output.weightF16[768, 768]
-
blk.2.attn_qkv.biasF32[2304]
-
blk.2.attn_qkv.weightF16[768, 2304]
-
blk.2.ffn_down.biasF32[768]
-
blk.2.ffn_down.weightF16[3072, 768]
-
blk.2.ffn_norm.biasF32[768]
-
blk.2.ffn_norm.weightF32[768]
-
blk.2.ffn_up.biasF32[3072]
-
blk.2.ffn_up.weightF16[768, 3072]
-
blk.3.attn_norm.biasF32[768]
-
blk.3.attn_norm.weightF32[768]
-
blk.3.attn_output.biasF32[768]
-
blk.3.attn_output.weightF16[768, 768]
-
blk.3.attn_qkv.biasF32[2304]
-
blk.3.attn_qkv.weightF16[768, 2304]
-
blk.3.ffn_down.biasF32[768]
-
blk.3.ffn_down.weightF16[3072, 768]
-
blk.3.ffn_norm.biasF32[768]
-
blk.3.ffn_norm.weightF32[768]
-
blk.3.ffn_up.biasF32[3072]
-
blk.3.ffn_up.weightF16[768, 3072]
-
blk.4.attn_norm.biasF32[768]
-
blk.4.attn_norm.weightF32[768]
-
blk.4.attn_output.biasF32[768]
-
blk.4.attn_output.weightF16[768, 768]
-
blk.4.attn_qkv.biasF32[2304]
-
blk.4.attn_qkv.weightF16[768, 2304]
-
blk.4.ffn_down.biasF32[768]
-
blk.4.ffn_down.weightF16[3072, 768]
-
blk.4.ffn_norm.biasF32[768]
-
blk.4.ffn_norm.weightF32[768]
-
blk.4.ffn_up.biasF32[3072]
-
blk.4.ffn_up.weightF16[768, 3072]
-
blk.5.attn_norm.biasF32[768]
-
blk.5.attn_norm.weightF32[768]
-
blk.5.attn_output.biasF32[768]
-
blk.5.attn_output.weightF16[768, 768]
-
blk.5.attn_qkv.biasF32[2304]
-
blk.5.attn_qkv.weightF16[768, 2304]
-
blk.5.ffn_down.biasF32[768]
-
blk.5.ffn_down.weightF16[3072, 768]
-
blk.5.ffn_norm.biasF32[768]
-
blk.5.ffn_norm.weightF32[768]
-
blk.5.ffn_up.biasF32[3072]
-
blk.5.ffn_up.weightF16[768, 3072]
-
blk.6.attn_norm.biasF32[768]
-
blk.6.attn_norm.weightF32[768]
-
blk.6.attn_output.biasF32[768]
-
blk.6.attn_output.weightF16[768, 768]
-
blk.6.attn_qkv.biasF32[2304]
-
blk.6.attn_qkv.weightF16[768, 2304]
-
blk.6.ffn_down.biasF32[768]
-
blk.6.ffn_down.weightF16[3072, 768]
-
blk.6.ffn_norm.biasF32[768]
-
blk.6.ffn_norm.weightF32[768]
-
blk.6.ffn_up.biasF32[3072]
-
blk.6.ffn_up.weightF16[768, 3072]
-
blk.7.attn_norm.biasF32[768]
-
blk.7.attn_norm.weightF32[768]
-
blk.7.attn_output.biasF32[768]
-
blk.7.attn_output.weightF16[768, 768]
-
blk.7.attn_qkv.biasF32[2304]
-
blk.7.attn_qkv.weightF16[768, 2304]
-
blk.7.ffn_down.biasF32[768]
-
blk.7.ffn_down.weightF16[3072, 768]
-
blk.7.ffn_norm.biasF32[768]
-
blk.7.ffn_norm.weightF32[768]
-
blk.7.ffn_up.biasF32[3072]
-
blk.7.ffn_up.weightF16[768, 3072]
-
blk.8.attn_norm.biasF32[768]
-
blk.8.attn_norm.weightF32[768]
-
blk.8.attn_output.biasF32[768]
-
blk.8.attn_output.weightF16[768, 768]
-
blk.8.attn_qkv.biasF32[2304]
-
blk.8.attn_qkv.weightF16[768, 2304]
-
blk.8.ffn_down.biasF32[768]
-
blk.8.ffn_down.weightF16[3072, 768]
-
blk.8.ffn_norm.biasF32[768]
-
blk.8.ffn_norm.weightF32[768]
-
blk.8.ffn_up.biasF32[3072]
-
blk.8.ffn_up.weightF16[768, 3072]
-
blk.9.attn_norm.biasF32[768]
-
blk.9.attn_norm.weightF32[768]
-
blk.9.attn_output.biasF32[768]
-
blk.9.attn_output.weightF16[768, 768]
-
blk.9.attn_qkv.biasF32[2304]
-
blk.9.attn_qkv.weightF16[768, 2304]
-
blk.9.ffn_down.biasF32[768]
-
blk.9.ffn_down.weightF16[3072, 768]
-
blk.9.ffn_norm.biasF32[768]
-
blk.9.ffn_norm.weightF32[768]
-
blk.9.ffn_up.biasF32[3072]
-
blk.9.ffn_up.weightF16[768, 3072]
-
blk.10.attn_norm.biasF32[768]
-
blk.10.attn_norm.weightF32[768]
-
blk.10.attn_output.biasF32[768]
-
blk.10.attn_output.weightF16[768, 768]
-
blk.10.attn_qkv.biasF32[2304]
-
blk.10.attn_qkv.weightF16[768, 2304]
-
blk.10.ffn_down.biasF32[768]
-
blk.10.ffn_down.weightF16[3072, 768]
-
blk.10.ffn_norm.biasF32[768]
-
blk.10.ffn_norm.weightF32[768]
-
blk.10.ffn_up.biasF32[3072]
-
blk.10.ffn_up.weightF16[768, 3072]
-
blk.11.attn_norm.biasF32[768]
-
blk.11.attn_norm.weightF32[768]
-
blk.11.attn_output.biasF32[768]
-
blk.11.attn_output.weightF16[768, 768]
-
blk.11.attn_qkv.biasF32[2304]
-
blk.11.attn_qkv.weightF16[768, 2304]
-
blk.11.ffn_down.biasF32[768]
-
blk.11.ffn_down.weightF16[3072, 768]
-
blk.11.ffn_norm.biasF32[768]
-
blk.11.ffn_norm.weightF32[768]
-
blk.11.ffn_up.biasF32[3072]
-
blk.11.ffn_up.weightF16[768, 3072]
-
output.weightF16[768, 50257]
-
output_norm.biasF32[768]
-
position_embd.weightF16[768, 1024]
-
output_norm.weightF32[768]
Metadata
Tensor
blk.0
blk.1
blk.2
blk.3
blk.4
blk.5
blk.6
blk.7
blk.8
blk.9
blk.10
blk.11