-
general.architecture
deepseek41
deepseek41
-
general.file_type
(!unknown_file_type 38!)
(!unknown_file_type 38!)
-
deepseek41.attention.compress_ratios
[0, 0, 2, 2, 2, ...]
[0, 0, 2, 2, 2, ...]
-
deepseek41.attention.compress_rope_freq_base
160000
160000
-
deepseek41.attention.head_count
64
64
-
deepseek41.attention.head_count_kv
1
1
-
deepseek41.attention.indexer.candidate_block_size
8
8
-
deepseek41.attention.indexer.candidate_source_layer
20
20
-
deepseek41.attention.indexer.candidate_top_k_blocks
2048
2048
-
deepseek41.attention.indexer.head_count
32
32
-
deepseek41.attention.indexer.key_length
128
128
-
deepseek41.attention.indexer.source_layers
[2, 8, 14, 20, 24, ...]
[2, 8, 14, 20, 24, ...]
-
deepseek41.attention.indexer.top_k
512
512
-
deepseek41.attention.key_length
512
512
-
deepseek41.attention.kv_source_layers
[2, 8, 14, 20]
[2, 8, 14, 20]
-
deepseek41.attention.layer_norm_rms_epsilon
1e-20
1e-20
-
deepseek41.attention.output_group_count
8
8
-
deepseek41.attention.output_lora_rank
1024
1024
-
deepseek41.attention.q_lora_rank
1280
1280
-
deepseek41.attention.sliding_window
128
128
-
deepseek41.attention.value_length
512
512
-
deepseek41.block_count
40
40
-
deepseek41.context_length
1048576
1048576
-
deepseek41.embedding_length
5120
5120
-
deepseek41.embedding_length_per_layer_input
256
256
-
deepseek41.engram.pad_token_id
2
2
-
deepseek41.engram.token_map
[0, 1, 2, 3, 4, ...]
[0, 1, 2, 3, 4, ...]
-
deepseek41.expert_count
384
384
-
deepseek41.expert_feed_forward_length
2304
2304
-
deepseek41.expert_gating_func
4
4
-
deepseek41.expert_shared_count
1
1
-
deepseek41.expert_used_count
6
6
-
deepseek41.expert_weights_norm
true
true
-
deepseek41.expert_weights_scale
1.5
1.5
-
deepseek41.hash_layer_count
0
0
-
deepseek41.hyper_connection.count
4
4
-
deepseek41.hyper_connection.epsilon
1e-06
1e-06
-
deepseek41.hyper_connection.sinkhorn_iterations
20
20
-
deepseek41.ple.head_offsets
[0, 16000057, 32000136, 48000217, 64000314, ...]
[0, 16000057, 32000136, 48000217, 64000314, ...]
-
deepseek41.ple.head_vocab_sizes
[16000057, 16000079, 16000081, 16000097, 16000121, ...]
[16000057, 16000079, 16000081, 16000097, 16000121, ...]
-
deepseek41.ple.heads_per_ngram
8
8
-
deepseek41.ple.layer_multipliers
[76632096046245, 4839876093313, 35959672319349, 73987337458391, 67716810739261, ...]
[76632096046245, 4839876093313, 35959672319349, 73987337458391, 67716810739261, ...]
-
deepseek41.ple.layers
[1, 14]
[1, 14]
-
deepseek41.ple.ngram_size
4
4
-
deepseek41.rope.dimension_count
64
64
-
deepseek41.rope.freq_base
10000
10000
-
deepseek41.rope.scaling.factor
16
16
-
deepseek41.rope.scaling.original_context_length
65536
65536
-
deepseek41.rope.scaling.type
yarn
yarn
-
deepseek41.rope.scaling.yarn_beta_fast
32
32
-
deepseek41.rope.scaling.yarn_beta_slow
1
1
-
deepseek41.swiglu_clamp_exp
[10, 10, 10, 10, 10, ...]
[10, 10, 10, 10, 10, ...]
-
deepseek41.swiglu_clamp_shexp
[10, 10, 10, 10, 10, ...]
[10, 10, 10, 10, 10, ...]
-
tokenizer.ggml.add_bos_token
false
false
-
tokenizer.ggml.add_eos_token
false
false
-
tokenizer.ggml.bos_token_id
0
0
-
tokenizer.ggml.clean_spaces
false
false
-
tokenizer.ggml.eos_token_id
1
1
-
tokenizer.ggml.merges
[Ġ t, Ġ a, i n, Ġ Ġ, h e, ...]
[Ġ t, Ġ a, i n, Ġ Ġ, h e, ...]
-
tokenizer.ggml.model
gpt2
gpt2
-
tokenizer.ggml.padding_token_id
1
1
-
tokenizer.ggml.pre
joyai-llm
joyai-llm
-
tokenizer.ggml.token_type
[3, 3, 3, 1, 1, ...]
[3, 3, 3, 1, 1, ...]
-
tokenizer.ggml.tokens
[<｜begin▁of▁sentence｜>, <｜end▁of▁sentence｜>, <｜▁pad▁｜>, !, ", ...]
[<｜begin▁of▁sentence｜>, <｜end▁of▁sentence｜>, <｜▁pad▁｜>, !, ", ...]
-
9
-
0
-
1043
-
Name
Type
Shape