latest
2.3GB
Phi-2 is a Transformer with 2.7 billion parameters. It was trained using the same data sources as Phi-1.5, augmented with a new data source that consists of various NLP synthetic texts and filtered websites (source: Microsoft).
3B
105 Pulls Updated 8 months ago
9a654a17bee2 · 2.3GB
-
general.architecturephi2
-
general.file_typeQ6_K
-
general.quantization_version2
-
phi2.attention.head_count32
-
phi2.attention.head_count_kv32
-
phi2.attention.layer_norm_epsilon1e-05
-
phi2.block_count32
-
phi2.context_length2048
-
phi2.embedding_length2560
-
phi2.feed_forward_length10240
-
phi2.rope.dimension_count32
-
tokenizer.ggml.add_bos_tokenfalse
-
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[! " # $ % ...]
-
tokenizer.ggml.unknown_token_id50256
-
NameTypeShape
-
token_embd.weightQ6_K[2560 51200]
-
blk.0.attn_norm.biasF32[2560]
-
blk.0.attn_norm.weightF32[2560]
-
blk.0.attn_qkv.biasF32[7680]
-
blk.0.attn_qkv.weightQ6_K[2560 7680]
-
blk.0.attn_output.biasF32[2560]
-
blk.0.attn_output.weightQ6_K[2560 2560]
-
blk.0.ffn_up.biasF32[10240]
-
blk.0.ffn_up.weightQ6_K[2560 10240]
-
blk.0.ffn_down.biasF32[2560]
-
blk.0.ffn_down.weightQ6_K[10240 2560]
-
blk.1.attn_norm.biasF32[2560]
-
blk.1.attn_norm.weightF32[2560]
-
blk.1.attn_qkv.biasF32[7680]
-
blk.1.attn_qkv.weightQ6_K[2560 7680]
-
blk.1.attn_output.biasF32[2560]
-
blk.1.attn_output.weightQ6_K[2560 2560]
-
blk.1.ffn_up.biasF32[10240]
-
blk.1.ffn_up.weightQ6_K[2560 10240]
-
blk.1.ffn_down.biasF32[2560]
-
blk.1.ffn_down.weightQ6_K[10240 2560]
-
blk.2.attn_norm.biasF32[2560]
-
blk.2.attn_norm.weightF32[2560]
-
blk.2.attn_qkv.biasF32[7680]
-
blk.2.attn_qkv.weightQ6_K[2560 7680]
-
blk.2.attn_output.biasF32[2560]
-
blk.2.attn_output.weightQ6_K[2560 2560]
-
blk.2.ffn_up.biasF32[10240]
-
blk.2.ffn_up.weightQ6_K[2560 10240]
-
blk.2.ffn_down.biasF32[2560]
-
blk.2.ffn_down.weightQ6_K[10240 2560]
-
blk.3.attn_norm.biasF32[2560]
-
blk.3.attn_norm.weightF32[2560]
-
blk.3.attn_qkv.biasF32[7680]
-
blk.3.attn_qkv.weightQ6_K[2560 7680]
-
blk.3.attn_output.biasF32[2560]
-
blk.3.attn_output.weightQ6_K[2560 2560]
-
blk.3.ffn_up.biasF32[10240]
-
blk.3.ffn_up.weightQ6_K[2560 10240]
-
blk.3.ffn_down.biasF32[2560]
-
blk.3.ffn_down.weightQ6_K[10240 2560]
-
blk.4.attn_norm.biasF32[2560]
-
blk.4.attn_norm.weightF32[2560]
-
blk.4.attn_qkv.biasF32[7680]
-
blk.4.attn_qkv.weightQ6_K[2560 7680]
-
blk.4.attn_output.biasF32[2560]
-
blk.4.attn_output.weightQ6_K[2560 2560]
-
blk.4.ffn_up.biasF32[10240]
-
blk.4.ffn_up.weightQ6_K[2560 10240]
-
blk.4.ffn_down.biasF32[2560]
-
blk.4.ffn_down.weightQ6_K[10240 2560]
-
blk.5.attn_norm.biasF32[2560]
-
blk.5.attn_norm.weightF32[2560]
-
blk.5.attn_qkv.biasF32[7680]
-
blk.5.attn_qkv.weightQ6_K[2560 7680]
-
blk.5.attn_output.biasF32[2560]
-
blk.5.attn_output.weightQ6_K[2560 2560]
-
blk.5.ffn_up.biasF32[10240]
-
blk.5.ffn_up.weightQ6_K[2560 10240]
-
blk.5.ffn_down.biasF32[2560]
-
blk.5.ffn_down.weightQ6_K[10240 2560]
-
blk.6.attn_norm.biasF32[2560]
-
blk.6.attn_norm.weightF32[2560]
-
blk.6.attn_qkv.biasF32[7680]
-
blk.6.attn_qkv.weightQ6_K[2560 7680]
-
blk.6.attn_output.biasF32[2560]
-
blk.6.attn_output.weightQ6_K[2560 2560]
-
blk.6.ffn_up.biasF32[10240]
-
blk.6.ffn_up.weightQ6_K[2560 10240]
-
blk.6.ffn_down.biasF32[2560]
-
blk.6.ffn_down.weightQ6_K[10240 2560]
-
blk.7.attn_norm.biasF32[2560]
-
blk.7.attn_norm.weightF32[2560]
-
blk.7.attn_qkv.biasF32[7680]
-
blk.7.attn_qkv.weightQ6_K[2560 7680]
-
blk.7.attn_output.biasF32[2560]
-
blk.7.attn_output.weightQ6_K[2560 2560]
-
blk.7.ffn_up.biasF32[10240]
-
blk.7.ffn_up.weightQ6_K[2560 10240]
-
blk.7.ffn_down.biasF32[2560]
-
blk.7.ffn_down.weightQ6_K[10240 2560]
-
blk.8.attn_norm.biasF32[2560]
-
blk.8.attn_norm.weightF32[2560]
-
blk.8.attn_qkv.biasF32[7680]
-
blk.8.attn_qkv.weightQ6_K[2560 7680]
-
blk.8.attn_output.biasF32[2560]
-
blk.8.attn_output.weightQ6_K[2560 2560]
-
blk.8.ffn_up.biasF32[10240]
-
blk.8.ffn_up.weightQ6_K[2560 10240]
-
blk.8.ffn_down.biasF32[2560]
-
blk.8.ffn_down.weightQ6_K[10240 2560]
-
blk.9.attn_norm.biasF32[2560]
-
blk.9.attn_norm.weightF32[2560]
-
blk.9.attn_qkv.biasF32[7680]
-
blk.9.attn_qkv.weightQ6_K[2560 7680]
-
blk.9.attn_output.biasF32[2560]
-
blk.9.attn_output.weightQ6_K[2560 2560]
-
blk.9.ffn_up.biasF32[10240]
-
blk.9.ffn_up.weightQ6_K[2560 10240]
-
blk.9.ffn_down.biasF32[2560]
-
blk.9.ffn_down.weightQ6_K[10240 2560]
-
blk.10.attn_norm.biasF32[2560]
-
blk.10.attn_norm.weightF32[2560]
-
blk.10.attn_qkv.biasF32[7680]
-
blk.10.attn_qkv.weightQ6_K[2560 7680]
-
blk.10.attn_output.biasF32[2560]
-
blk.10.attn_output.weightQ6_K[2560 2560]
-
blk.10.ffn_up.biasF32[10240]
-
blk.10.ffn_up.weightQ6_K[2560 10240]
-
blk.10.ffn_down.biasF32[2560]
-
blk.10.ffn_down.weightQ6_K[10240 2560]
-
blk.11.attn_norm.biasF32[2560]
-
blk.11.attn_norm.weightF32[2560]
-
blk.11.attn_qkv.biasF32[7680]
-
blk.11.attn_qkv.weightQ6_K[2560 7680]
-
blk.11.attn_output.biasF32[2560]
-
blk.11.attn_output.weightQ6_K[2560 2560]
-
blk.11.ffn_up.biasF32[10240]
-
blk.11.ffn_up.weightQ6_K[2560 10240]
-
blk.11.ffn_down.biasF32[2560]
-
blk.11.ffn_down.weightQ6_K[10240 2560]
-
blk.12.attn_norm.biasF32[2560]
-
blk.12.attn_norm.weightF32[2560]
-
blk.12.attn_qkv.biasF32[7680]
-
blk.12.attn_qkv.weightQ6_K[2560 7680]
-
blk.12.attn_output.biasF32[2560]
-
blk.12.attn_output.weightQ6_K[2560 2560]
-
blk.12.ffn_up.biasF32[10240]
-
blk.12.ffn_up.weightQ6_K[2560 10240]
-
blk.12.ffn_down.biasF32[2560]
-
blk.12.ffn_down.weightQ6_K[10240 2560]
-
blk.13.attn_norm.biasF32[2560]
-
blk.13.attn_norm.weightF32[2560]
-
blk.13.attn_qkv.biasF32[7680]
-
blk.13.attn_qkv.weightQ6_K[2560 7680]
-
blk.13.attn_output.biasF32[2560]
-
blk.13.attn_output.weightQ6_K[2560 2560]
-
blk.13.ffn_up.biasF32[10240]
-
blk.13.ffn_up.weightQ6_K[2560 10240]
-
blk.13.ffn_down.biasF32[2560]
-
blk.13.ffn_down.weightQ6_K[10240 2560]
-
blk.14.attn_norm.biasF32[2560]
-
blk.14.attn_norm.weightF32[2560]
-
blk.14.attn_qkv.biasF32[7680]
-
blk.14.attn_qkv.weightQ6_K[2560 7680]
-
blk.14.attn_output.biasF32[2560]
-
blk.14.attn_output.weightQ6_K[2560 2560]
-
blk.14.ffn_up.biasF32[10240]
-
blk.14.ffn_up.weightQ6_K[2560 10240]
-
blk.14.ffn_down.biasF32[2560]
-
blk.14.ffn_down.weightQ6_K[10240 2560]
-
blk.15.attn_norm.biasF32[2560]
-
blk.15.attn_norm.weightF32[2560]
-
blk.15.attn_qkv.biasF32[7680]
-
blk.15.attn_qkv.weightQ6_K[2560 7680]
-
blk.15.attn_output.biasF32[2560]
-
blk.15.attn_output.weightQ6_K[2560 2560]
-
blk.15.ffn_up.biasF32[10240]
-
blk.15.ffn_up.weightQ6_K[2560 10240]
-
blk.15.ffn_down.biasF32[2560]
-
blk.15.ffn_down.weightQ6_K[10240 2560]
-
blk.16.attn_norm.biasF32[2560]
-
blk.16.attn_norm.weightF32[2560]
-
blk.16.attn_qkv.biasF32[7680]
-
blk.16.attn_qkv.weightQ6_K[2560 7680]
-
blk.16.attn_output.biasF32[2560]
-
blk.16.attn_output.weightQ6_K[2560 2560]
-
blk.16.ffn_up.biasF32[10240]
-
blk.16.ffn_up.weightQ6_K[2560 10240]
-
blk.16.ffn_down.biasF32[2560]
-
blk.16.ffn_down.weightQ6_K[10240 2560]
-
blk.17.attn_norm.biasF32[2560]
-
blk.17.attn_norm.weightF32[2560]
-
blk.17.attn_qkv.biasF32[7680]
-
blk.17.attn_qkv.weightQ6_K[2560 7680]
-
blk.17.attn_output.biasF32[2560]
-
blk.17.attn_output.weightQ6_K[2560 2560]
-
blk.17.ffn_up.biasF32[10240]
-
blk.17.ffn_up.weightQ6_K[2560 10240]
-
blk.17.ffn_down.biasF32[2560]
-
blk.17.ffn_down.weightQ6_K[10240 2560]
-
blk.18.attn_norm.biasF32[2560]
-
blk.18.attn_norm.weightF32[2560]
-
blk.18.attn_qkv.biasF32[7680]
-
blk.18.attn_qkv.weightQ6_K[2560 7680]
-
blk.18.attn_output.biasF32[2560]
-
blk.18.attn_output.weightQ6_K[2560 2560]
-
blk.18.ffn_up.biasF32[10240]
-
blk.18.ffn_up.weightQ6_K[2560 10240]
-
blk.18.ffn_down.biasF32[2560]
-
blk.18.ffn_down.weightQ6_K[10240 2560]
-
blk.19.attn_norm.biasF32[2560]
-
blk.19.attn_norm.weightF32[2560]
-
blk.19.attn_qkv.biasF32[7680]
-
blk.19.attn_qkv.weightQ6_K[2560 7680]
-
blk.19.attn_output.biasF32[2560]
-
blk.19.attn_output.weightQ6_K[2560 2560]
-
blk.19.ffn_up.biasF32[10240]
-
blk.19.ffn_up.weightQ6_K[2560 10240]
-
blk.19.ffn_down.biasF32[2560]
-
blk.19.ffn_down.weightQ6_K[10240 2560]
-
blk.20.attn_norm.biasF32[2560]
-
blk.20.attn_norm.weightF32[2560]
-
blk.20.attn_qkv.biasF32[7680]
-
blk.20.attn_qkv.weightQ6_K[2560 7680]
-
blk.20.attn_output.biasF32[2560]
-
blk.20.attn_output.weightQ6_K[2560 2560]
-
blk.20.ffn_up.biasF32[10240]
-
blk.20.ffn_up.weightQ6_K[2560 10240]
-
blk.20.ffn_down.biasF32[2560]
-
blk.20.ffn_down.weightQ6_K[10240 2560]
-
blk.21.attn_norm.biasF32[2560]
-
blk.21.attn_norm.weightF32[2560]
-
blk.21.attn_qkv.biasF32[7680]
-
blk.21.attn_qkv.weightQ6_K[2560 7680]
-
blk.21.attn_output.biasF32[2560]
-
blk.21.attn_output.weightQ6_K[2560 2560]
-
blk.21.ffn_up.biasF32[10240]
-
blk.21.ffn_up.weightQ6_K[2560 10240]
-
blk.21.ffn_down.biasF32[2560]
-
blk.21.ffn_down.weightQ6_K[10240 2560]
-
blk.22.attn_norm.biasF32[2560]
-
blk.22.attn_norm.weightF32[2560]
-
blk.22.attn_qkv.biasF32[7680]
-
blk.22.attn_qkv.weightQ6_K[2560 7680]
-
blk.22.attn_output.biasF32[2560]
-
blk.22.attn_output.weightQ6_K[2560 2560]
-
blk.22.ffn_up.biasF32[10240]
-
blk.22.ffn_up.weightQ6_K[2560 10240]
-
blk.22.ffn_down.biasF32[2560]
-
blk.22.ffn_down.weightQ6_K[10240 2560]
-
blk.23.attn_norm.biasF32[2560]
-
blk.23.attn_norm.weightF32[2560]
-
blk.23.attn_qkv.biasF32[7680]
-
blk.23.attn_qkv.weightQ6_K[2560 7680]
-
blk.23.attn_output.biasF32[2560]
-
blk.23.attn_output.weightQ6_K[2560 2560]
-
blk.23.ffn_up.biasF32[10240]
-
blk.23.ffn_up.weightQ6_K[2560 10240]
-
blk.23.ffn_down.biasF32[2560]
-
blk.23.ffn_down.weightQ6_K[10240 2560]
-
blk.24.attn_norm.biasF32[2560]
-
blk.24.attn_norm.weightF32[2560]
-
blk.24.attn_qkv.biasF32[7680]
-
blk.24.attn_qkv.weightQ6_K[2560 7680]
-
blk.24.attn_output.biasF32[2560]
-
blk.24.attn_output.weightQ6_K[2560 2560]
-
blk.24.ffn_up.biasF32[10240]
-
blk.24.ffn_up.weightQ6_K[2560 10240]
-
blk.24.ffn_down.biasF32[2560]
-
blk.24.ffn_down.weightQ6_K[10240 2560]
-
blk.25.attn_norm.biasF32[2560]
-
blk.25.attn_norm.weightF32[2560]
-
blk.25.attn_qkv.biasF32[7680]
-
blk.25.attn_qkv.weightQ6_K[2560 7680]
-
blk.25.attn_output.biasF32[2560]
-
blk.25.attn_output.weightQ6_K[2560 2560]
-
blk.25.ffn_up.biasF32[10240]
-
blk.25.ffn_up.weightQ6_K[2560 10240]
-
blk.25.ffn_down.biasF32[2560]
-
blk.25.ffn_down.weightQ6_K[10240 2560]
-
blk.26.attn_norm.biasF32[2560]
-
blk.26.attn_norm.weightF32[2560]
-
blk.26.attn_qkv.biasF32[7680]
-
blk.26.attn_qkv.weightQ6_K[2560 7680]
-
blk.26.attn_output.biasF32[2560]
-
blk.26.attn_output.weightQ6_K[2560 2560]
-
blk.26.ffn_up.biasF32[10240]
-
blk.26.ffn_up.weightQ6_K[2560 10240]
-
blk.26.ffn_down.biasF32[2560]
-
blk.26.ffn_down.weightQ6_K[10240 2560]
-
blk.27.attn_norm.biasF32[2560]
-
blk.27.attn_norm.weightF32[2560]
-
blk.27.attn_qkv.biasF32[7680]
-
blk.27.attn_qkv.weightQ6_K[2560 7680]
-
blk.27.attn_output.biasF32[2560]
-
blk.27.attn_output.weightQ6_K[2560 2560]
-
blk.27.ffn_up.biasF32[10240]
-
blk.27.ffn_up.weightQ6_K[2560 10240]
-
blk.27.ffn_down.biasF32[2560]
-
blk.27.ffn_down.weightQ6_K[10240 2560]
-
blk.28.attn_norm.biasF32[2560]
-
blk.28.attn_norm.weightF32[2560]
-
blk.28.attn_qkv.biasF32[7680]
-
blk.28.attn_qkv.weightQ6_K[2560 7680]
-
blk.28.attn_output.biasF32[2560]
-
blk.28.attn_output.weightQ6_K[2560 2560]
-
blk.28.ffn_up.biasF32[10240]
-
blk.28.ffn_up.weightQ6_K[2560 10240]
-
blk.28.ffn_down.biasF32[2560]
-
blk.28.ffn_down.weightQ6_K[10240 2560]
-
blk.29.attn_norm.biasF32[2560]
-
blk.29.attn_norm.weightF32[2560]
-
blk.29.attn_qkv.biasF32[7680]
-
blk.29.attn_qkv.weightQ6_K[2560 7680]
-
blk.29.attn_output.biasF32[2560]
-
blk.29.attn_output.weightQ6_K[2560 2560]
-
blk.29.ffn_up.biasF32[10240]
-
blk.29.ffn_up.weightQ6_K[2560 10240]
-
blk.29.ffn_down.biasF32[2560]
-
blk.29.ffn_down.weightQ6_K[10240 2560]
-
blk.30.attn_norm.biasF32[2560]
-
blk.30.attn_norm.weightF32[2560]
-
blk.30.attn_qkv.biasF32[7680]
-
blk.30.attn_qkv.weightQ6_K[2560 7680]
-
blk.30.attn_output.biasF32[2560]
-
blk.30.attn_output.weightQ6_K[2560 2560]
-
blk.30.ffn_up.biasF32[10240]
-
blk.30.ffn_up.weightQ6_K[2560 10240]
-
blk.30.ffn_down.biasF32[2560]
-
blk.30.ffn_down.weightQ6_K[10240 2560]
-
blk.31.attn_norm.biasF32[2560]
-
blk.31.attn_norm.weightF32[2560]
-
blk.31.attn_qkv.biasF32[7680]
-
blk.31.attn_qkv.weightQ6_K[2560 7680]
-
blk.31.attn_output.biasF32[2560]
-
blk.31.attn_output.weightQ6_K[2560 2560]
-
blk.31.ffn_up.biasF32[10240]
-
blk.31.ffn_up.weightQ6_K[2560 10240]
-
blk.31.ffn_down.biasF32[2560]
-
blk.31.ffn_down.weightQ6_K[10240 2560]
-
output.biasF32[51200]
-
output.weightQ6_K[2560 51200]
-
output_norm.biasF32[2560]
-
output_norm.weightF32[2560]
Metadata
Tensors
blk.0
blk.1
blk.2
blk.3
blk.4
blk.5
blk.6
blk.7
blk.8
blk.9
blk.10
blk.11
blk.12
blk.13
blk.14
blk.15
blk.16
blk.17
blk.18
blk.19
blk.20
blk.21
blk.22
blk.23
blk.24
blk.25
blk.26
blk.27
blk.28
blk.29
blk.30
blk.31