Phi-3-mini-4K-instruct with CPO-SimPO
36 Pulls Updated 4 months ago
e32350736f8c · 2.2GB
-
general.architecturephi3
-
general.file_type14
-
general.namePhi3
-
general.quantization_version2
-
phi3.attention.head_count32
-
phi3.attention.head_count_kv32
-
phi3.attention.layer_norm_rms_epsilon1e-05
-
phi3.block_count32
-
phi3.context_length4096
-
phi3.embedding_length3072
-
phi3.feed_forward_length8192
-
phi3.rope.dimension_count96
-
phi3.rope.freq_base10000
-
phi3.rope.scaling.original_context_length4096
-
quantize.imatrix.chunks_count80
-
quantize.imatrix.dataset/shared/opt/work_models/_imatrix/calibration_datav3.txt
-
quantize.imatrix.entries_count128
-
quantize.imatrix.fileimatrix.dat
-
tokenizer.ggml.add_bos_tokentrue
-
tokenizer.ggml.add_eos_tokenfalse
-
tokenizer.ggml.bos_token_id1
-
tokenizer.ggml.eos_token_id32000
-
tokenizer.ggml.modelllama
-
tokenizer.ggml.padding_token_id32000
-
tokenizer.ggml.predefault
-
tokenizer.ggml.scores[-1000, -1000, -1000, 0, 0, ...]
-
tokenizer.ggml.token_type[3, 3, 4, 6, 6, ...]
-
tokenizer.ggml.tokens[<unk>, <s>, </s>, <0x00>, <0x01>, ...]
-
tokenizer.ggml.unknown_token_id0
-
NameTypeShape
-
token_embd.weightQ4_K[3072, 32064]
-
blk.0.attn_norm.weightF32[3072]
-
blk.0.attn_output.weightQ4_K[3072, 3072]
-
blk.0.attn_qkv.weightQ4_K[3072, 9216]
-
blk.0.ffn_down.weightQ5_K[8192, 3072]
-
blk.0.ffn_norm.weightF32[3072]
-
blk.0.ffn_up.weightQ4_K[3072, 16384]
-
blk.1.attn_norm.weightF32[3072]
-
blk.1.attn_output.weightQ4_K[3072, 3072]
-
blk.1.attn_qkv.weightQ4_K[3072, 9216]
-
blk.1.ffn_down.weightQ5_K[8192, 3072]
-
blk.1.ffn_norm.weightF32[3072]
-
blk.1.ffn_up.weightQ4_K[3072, 16384]
-
blk.2.attn_norm.weightF32[3072]
-
blk.2.attn_output.weightQ4_K[3072, 3072]
-
blk.2.attn_qkv.weightQ4_K[3072, 9216]
-
blk.2.ffn_down.weightQ4_K[8192, 3072]
-
blk.2.ffn_norm.weightF32[3072]
-
blk.2.ffn_up.weightQ4_K[3072, 16384]
-
blk.3.attn_norm.weightF32[3072]
-
blk.3.attn_output.weightQ4_K[3072, 3072]
-
blk.3.attn_qkv.weightQ4_K[3072, 9216]
-
blk.3.ffn_down.weightQ4_K[8192, 3072]
-
blk.3.ffn_norm.weightF32[3072]
-
blk.3.ffn_up.weightQ4_K[3072, 16384]
-
blk.4.attn_norm.weightF32[3072]
-
blk.4.attn_output.weightQ4_K[3072, 3072]
-
blk.4.attn_qkv.weightQ4_K[3072, 9216]
-
blk.4.ffn_down.weightQ4_K[8192, 3072]
-
blk.4.ffn_norm.weightF32[3072]
-
blk.4.ffn_up.weightQ4_K[3072, 16384]
-
blk.5.attn_norm.weightF32[3072]
-
blk.5.attn_output.weightQ4_K[3072, 3072]
-
blk.5.attn_qkv.weightQ4_K[3072, 9216]
-
blk.5.ffn_down.weightQ4_K[8192, 3072]
-
blk.5.ffn_norm.weightF32[3072]
-
blk.5.ffn_up.weightQ4_K[3072, 16384]
-
blk.6.attn_norm.weightF32[3072]
-
blk.6.attn_output.weightQ4_K[3072, 3072]
-
blk.6.attn_qkv.weightQ4_K[3072, 9216]
-
blk.6.ffn_down.weightQ4_K[8192, 3072]
-
blk.6.ffn_norm.weightF32[3072]
-
blk.6.ffn_up.weightQ4_K[3072, 16384]
-
blk.7.attn_norm.weightF32[3072]
-
blk.7.attn_output.weightQ4_K[3072, 3072]
-
blk.7.attn_qkv.weightQ4_K[3072, 9216]
-
blk.7.ffn_down.weightQ4_K[8192, 3072]
-
blk.7.ffn_norm.weightF32[3072]
-
blk.7.ffn_up.weightQ4_K[3072, 16384]
-
blk.8.attn_norm.weightF32[3072]
-
blk.8.attn_output.weightQ4_K[3072, 3072]
-
blk.8.attn_qkv.weightQ4_K[3072, 9216]
-
blk.8.ffn_down.weightQ4_K[8192, 3072]
-
blk.8.ffn_norm.weightF32[3072]
-
blk.8.ffn_up.weightQ4_K[3072, 16384]
-
blk.9.attn_norm.weightF32[3072]
-
blk.9.attn_output.weightQ4_K[3072, 3072]
-
blk.9.attn_qkv.weightQ4_K[3072, 9216]
-
blk.9.ffn_down.weightQ4_K[8192, 3072]
-
blk.9.ffn_norm.weightF32[3072]
-
blk.9.ffn_up.weightQ4_K[3072, 16384]
-
blk.10.attn_norm.weightF32[3072]
-
blk.10.attn_output.weightQ4_K[3072, 3072]
-
blk.10.attn_qkv.weightQ4_K[3072, 9216]
-
blk.10.ffn_down.weightQ5_K[8192, 3072]
-
blk.10.ffn_norm.weightF32[3072]
-
blk.10.ffn_up.weightQ4_K[3072, 16384]
-
blk.11.attn_norm.weightF32[3072]
-
blk.11.attn_output.weightQ4_K[3072, 3072]
-
blk.11.attn_qkv.weightQ4_K[3072, 9216]
-
blk.11.ffn_down.weightQ5_K[8192, 3072]
-
blk.11.ffn_norm.weightF32[3072]
-
blk.11.ffn_up.weightQ4_K[3072, 16384]
-
blk.12.attn_norm.weightF32[3072]
-
blk.12.attn_output.weightQ4_K[3072, 3072]
-
blk.12.attn_qkv.weightQ4_K[3072, 9216]
-
blk.12.ffn_down.weightQ4_K[8192, 3072]
-
blk.12.ffn_norm.weightF32[3072]
-
blk.12.ffn_up.weightQ4_K[3072, 16384]
-
blk.13.attn_norm.weightF32[3072]
-
blk.13.attn_output.weightQ4_K[3072, 3072]
-
blk.13.attn_qkv.weightQ4_K[3072, 9216]
-
blk.13.ffn_down.weightQ4_K[8192, 3072]
-
blk.13.ffn_norm.weightF32[3072]
-
blk.13.ffn_up.weightQ4_K[3072, 16384]
-
blk.14.attn_norm.weightF32[3072]
-
blk.14.attn_output.weightQ4_K[3072, 3072]
-
blk.14.attn_qkv.weightQ4_K[3072, 9216]
-
blk.14.ffn_down.weightQ4_K[8192, 3072]
-
blk.14.ffn_norm.weightF32[3072]
-
blk.14.ffn_up.weightQ4_K[3072, 16384]
-
blk.15.attn_norm.weightF32[3072]
-
blk.15.attn_output.weightQ4_K[3072, 3072]
-
blk.15.attn_qkv.weightQ4_K[3072, 9216]
-
blk.15.ffn_down.weightQ4_K[8192, 3072]
-
blk.15.ffn_norm.weightF32[3072]
-
blk.15.ffn_up.weightQ4_K[3072, 16384]
-
blk.16.attn_norm.weightF32[3072]
-
blk.16.attn_output.weightQ4_K[3072, 3072]
-
blk.16.attn_qkv.weightQ4_K[3072, 9216]
-
blk.16.ffn_down.weightQ4_K[8192, 3072]
-
blk.16.ffn_norm.weightF32[3072]
-
blk.16.ffn_up.weightQ4_K[3072, 16384]
-
blk.17.attn_norm.weightF32[3072]
-
blk.17.attn_output.weightQ4_K[3072, 3072]
-
blk.17.attn_qkv.weightQ4_K[3072, 9216]
-
blk.17.ffn_down.weightQ4_K[8192, 3072]
-
blk.17.ffn_norm.weightF32[3072]
-
blk.17.ffn_up.weightQ4_K[3072, 16384]
-
blk.18.attn_norm.weightF32[3072]
-
blk.18.attn_output.weightQ4_K[3072, 3072]
-
blk.18.attn_qkv.weightQ4_K[3072, 9216]
-
blk.18.ffn_down.weightQ4_K[8192, 3072]
-
blk.18.ffn_norm.weightF32[3072]
-
blk.18.ffn_up.weightQ4_K[3072, 16384]
-
blk.19.attn_norm.weightF32[3072]
-
blk.19.attn_output.weightQ4_K[3072, 3072]
-
blk.19.attn_qkv.weightQ4_K[3072, 9216]
-
blk.19.ffn_down.weightQ4_K[8192, 3072]
-
blk.19.ffn_norm.weightF32[3072]
-
blk.19.ffn_up.weightQ4_K[3072, 16384]
-
blk.20.attn_norm.weightF32[3072]
-
blk.20.attn_output.weightQ4_K[3072, 3072]
-
blk.20.attn_qkv.weightQ4_K[3072, 9216]
-
blk.20.ffn_down.weightQ4_K[8192, 3072]
-
blk.20.ffn_norm.weightF32[3072]
-
blk.20.ffn_up.weightQ4_K[3072, 16384]
-
blk.21.attn_norm.weightF32[3072]
-
blk.21.attn_output.weightQ4_K[3072, 3072]
-
blk.21.attn_qkv.weightQ4_K[3072, 9216]
-
blk.21.ffn_down.weightQ4_K[8192, 3072]
-
blk.21.ffn_norm.weightF32[3072]
-
blk.21.ffn_up.weightQ4_K[3072, 16384]
-
blk.22.attn_norm.weightF32[3072]
-
blk.22.attn_output.weightQ4_K[3072, 3072]
-
blk.22.attn_qkv.weightQ4_K[3072, 9216]
-
blk.22.ffn_down.weightQ4_K[8192, 3072]
-
blk.22.ffn_norm.weightF32[3072]
-
blk.22.ffn_up.weightQ4_K[3072, 16384]
-
blk.23.attn_norm.weightF32[3072]
-
blk.23.attn_output.weightQ4_K[3072, 3072]
-
blk.23.attn_qkv.weightQ4_K[3072, 9216]
-
blk.23.ffn_down.weightQ4_K[8192, 3072]
-
blk.23.ffn_norm.weightF32[3072]
-
blk.23.ffn_up.weightQ4_K[3072, 16384]
-
blk.24.attn_norm.weightF32[3072]
-
blk.24.attn_output.weightQ4_K[3072, 3072]
-
blk.24.attn_qkv.weightQ4_K[3072, 9216]
-
blk.24.ffn_down.weightQ4_K[8192, 3072]
-
blk.24.ffn_norm.weightF32[3072]
-
blk.24.ffn_up.weightQ4_K[3072, 16384]
-
blk.25.attn_norm.weightF32[3072]
-
blk.25.attn_output.weightQ4_K[3072, 3072]
-
blk.25.attn_qkv.weightQ4_K[3072, 9216]
-
blk.25.ffn_down.weightQ4_K[8192, 3072]
-
blk.25.ffn_norm.weightF32[3072]
-
blk.25.ffn_up.weightQ4_K[3072, 16384]
-
blk.26.attn_norm.weightF32[3072]
-
blk.26.attn_output.weightQ4_K[3072, 3072]
-
blk.26.attn_qkv.weightQ4_K[3072, 9216]
-
blk.26.ffn_down.weightQ4_K[8192, 3072]
-
blk.26.ffn_norm.weightF32[3072]
-
blk.26.ffn_up.weightQ4_K[3072, 16384]
-
blk.27.attn_norm.weightF32[3072]
-
blk.27.attn_output.weightQ4_K[3072, 3072]
-
blk.27.attn_qkv.weightQ4_K[3072, 9216]
-
blk.27.ffn_down.weightQ4_K[8192, 3072]
-
blk.27.ffn_norm.weightF32[3072]
-
blk.27.ffn_up.weightQ4_K[3072, 16384]
-
blk.28.attn_norm.weightF32[3072]
-
blk.28.attn_output.weightQ4_K[3072, 3072]
-
blk.28.attn_qkv.weightQ4_K[3072, 9216]
-
blk.28.ffn_down.weightQ4_K[8192, 3072]
-
blk.28.ffn_norm.weightF32[3072]
-
blk.28.ffn_up.weightQ4_K[3072, 16384]
-
blk.29.attn_norm.weightF32[3072]
-
blk.29.attn_output.weightQ4_K[3072, 3072]
-
blk.29.attn_qkv.weightQ4_K[3072, 9216]
-
blk.29.ffn_down.weightQ4_K[8192, 3072]
-
blk.29.ffn_norm.weightF32[3072]
-
blk.29.ffn_up.weightQ4_K[3072, 16384]
-
blk.30.attn_norm.weightF32[3072]
-
blk.30.attn_output.weightQ4_K[3072, 3072]
-
blk.30.attn_qkv.weightQ4_K[3072, 9216]
-
blk.30.ffn_down.weightQ4_K[8192, 3072]
-
blk.30.ffn_norm.weightF32[3072]
-
blk.30.ffn_up.weightQ4_K[3072, 16384]
-
blk.31.attn_norm.weightF32[3072]
-
blk.31.attn_output.weightQ4_K[3072, 3072]
-
blk.31.attn_qkv.weightQ4_K[3072, 9216]
-
blk.31.ffn_down.weightQ4_K[8192, 3072]
-
blk.31.ffn_norm.weightF32[3072]
-
blk.31.ffn_up.weightQ4_K[3072, 16384]
-
output.weightQ6_K[3072, 32064]
-
output_norm.weightF32[3072]
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
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