mannix / gemma2-9b-simpo

Fine-tuned google/gemma-2-9b-it on princeton-nlp/gemma2-ultrafeedback-armorm with the SimPO objective.

1,021 3 months ago

24 Tags
... /
model
9db5f2794f53 · 5.2GB
    Metadata
  • gemma2.attention.head_count
    16
  • gemma2.attention.head_count_kv
    8
  • gemma2.attention.key_length
    256
  • gemma2.attention.layer_norm_rms_epsilon
    1e-06
  • gemma2.attention.sliding_window
    4096
  • gemma2.attention.value_length
    256
  • gemma2.attn_logit_softcapping
    50
  • gemma2.block_count
    42
  • gemma2.context_length
    8192
  • gemma2.embedding_length
    3584
  • gemma2.feed_forward_length
    14336
  • gemma2.final_logit_softcapping
    30
  • general.architecture
    gemma2
  • general.base_model.0.name
    Gemma 2 9b It
  • general.base_model.0.organization
    Google
  • general.base_model.0.repo_url
    https://huggingface.co/google/gemma-2-9b-it
  • general.base_model.count
    1
  • general.basename
    gemma-2
  • general.datasets
    [princeton-nlp/gemma2-ultrafeedback-armorm]
  • general.file_type
    30
  • general.finetune
    it-SimPO
  • general.license
    mit
  • general.name
    Gemma 2 9b It SimPO
  • general.organization
    Princeton Nlp
  • general.quantization_version
    2
  • general.size_label
    9B
  • general.tags
    [alignment-handbook, generated_from_trainer]
  • general.type
    model
  • quantize.imatrix.chunks_count
    64
  • quantize.imatrix.dataset
    /shared/opt/work_models/_imatrix/calibration_datav3.txt
  • quantize.imatrix.entries_count
    294
  • quantize.imatrix.file
    imatrix.dat
  • tokenizer.ggml.add_bos_token
    true
  • tokenizer.ggml.add_eos_token
    false
  • tokenizer.ggml.add_space_prefix
    false
  • tokenizer.ggml.bos_token_id
    2
  • tokenizer.ggml.eos_token_id
    1
  • tokenizer.ggml.model
    llama
  • tokenizer.ggml.padding_token_id
    0
  • tokenizer.ggml.pre
    default
  • tokenizer.ggml.scores
    [-1000, -1000, -1000, -1000, -1000, ...]
  • tokenizer.ggml.token_type
    [3, 3, 3, 3, 3, ...]
  • tokenizer.ggml.tokens
    [<pad>, <eos>, <bos>, <unk>, <mask>, ...]
  • tokenizer.ggml.unknown_token_id
    3
    • Tensor
  • Name
    Type
    Shape
  • token_embd.weight
    Q6_K
    [3584, 256000]
    • blk.0
  • blk.0.attn_k.weight
    (!unknown_type 23!)
    [3584, 2048]
  • blk.0.attn_norm.weight
    F32
    [3584]
  • blk.0.attn_output.weight
    (!unknown_type 23!)
    [4096, 3584]
  • blk.0.attn_q.weight
    (!unknown_type 23!)
    [3584, 4096]
  • blk.0.attn_v.weight
    (!unknown_type 23!)
    [3584, 2048]
  • blk.0.ffn_down.weight
    (!unknown_type 23!)
    [14336, 3584]
  • blk.0.ffn_gate.weight
    (!unknown_type 23!)
    [3584, 14336]
  • blk.0.ffn_norm.weight
    F32
    [3584]
  • blk.0.ffn_up.weight
    (!unknown_type 23!)
    [3584, 14336]
  • blk.0.post_attention_norm.weight
    F32
    [3584]
  • blk.0.post_ffw_norm.weight
    F32
    [3584]
    • blk.1
  • blk.1.attn_k.weight
    (!unknown_type 23!)
    [3584, 2048]
  • blk.1.attn_norm.weight
    F32
    [3584]
  • blk.1.attn_output.weight
    (!unknown_type 23!)
    [4096, 3584]
  • blk.1.attn_q.weight
    (!unknown_type 23!)
    [3584, 4096]
  • blk.1.attn_v.weight
    (!unknown_type 23!)
    [3584, 2048]
  • blk.1.ffn_down.weight
    (!unknown_type 23!)
    [14336, 3584]
  • blk.1.ffn_gate.weight
    (!unknown_type 23!)
    [3584, 14336]
  • blk.1.ffn_norm.weight
    F32
    [3584]
  • blk.1.ffn_up.weight
    (!unknown_type 23!)
    [3584, 14336]
  • blk.1.post_attention_norm.weight
    F32
    [3584]
  • blk.1.post_ffw_norm.weight
    F32
    [3584]
    • blk.2
  • blk.2.attn_k.weight
    (!unknown_type 23!)
    [3584, 2048]
  • blk.2.attn_norm.weight
    F32
    [3584]
  • blk.2.attn_output.weight
    (!unknown_type 23!)
    [4096, 3584]
  • blk.2.attn_q.weight
    (!unknown_type 23!)
    [3584, 4096]
  • blk.2.attn_v.weight
    (!unknown_type 23!)
    [3584, 2048]
  • blk.2.ffn_down.weight
    (!unknown_type 23!)
    [14336, 3584]
  • blk.2.ffn_gate.weight
    (!unknown_type 23!)
    [3584, 14336]
  • blk.2.ffn_norm.weight
    F32
    [3584]
  • blk.2.ffn_up.weight
    (!unknown_type 23!)
    [3584, 14336]
  • blk.2.post_attention_norm.weight
    F32
    [3584]
  • blk.2.post_ffw_norm.weight
    F32
    [3584]
    • blk.3
  • blk.3.attn_k.weight
    (!unknown_type 23!)
    [3584, 2048]
  • blk.3.attn_norm.weight
    F32
    [3584]
  • blk.3.attn_output.weight
    (!unknown_type 23!)
    [4096, 3584]
  • blk.3.attn_q.weight
    (!unknown_type 23!)
    [3584, 4096]
  • blk.3.attn_v.weight
    (!unknown_type 23!)
    [3584, 2048]
  • blk.3.ffn_down.weight
    (!unknown_type 23!)
    [14336, 3584]
  • blk.3.ffn_gate.weight
    (!unknown_type 23!)
    [3584, 14336]
  • blk.3.ffn_norm.weight
    F32
    [3584]
  • blk.3.ffn_up.weight
    (!unknown_type 23!)
    [3584, 14336]
  • blk.3.post_attention_norm.weight
    F32
    [3584]
  • blk.3.post_ffw_norm.weight
    F32
    [3584]
    • blk.4
  • blk.4.attn_k.weight
    (!unknown_type 23!)
    [3584, 2048]
  • blk.4.attn_norm.weight
    F32
    [3584]
  • blk.4.attn_output.weight
    (!unknown_type 23!)
    [4096, 3584]
  • blk.4.attn_q.weight
    (!unknown_type 23!)
    [3584, 4096]
  • blk.4.attn_v.weight
    (!unknown_type 23!)
    [3584, 2048]
  • blk.4.ffn_down.weight
    (!unknown_type 23!)
    [14336, 3584]
  • blk.4.ffn_gate.weight
    (!unknown_type 23!)
    [3584, 14336]
  • blk.4.ffn_norm.weight
    F32
    [3584]
  • blk.4.ffn_up.weight
    (!unknown_type 23!)
    [3584, 14336]
  • blk.4.post_attention_norm.weight
    F32
    [3584]
  • blk.4.post_ffw_norm.weight
    F32
    [3584]
    • blk.5
  • blk.5.attn_k.weight
    (!unknown_type 23!)
    [3584, 2048]
  • blk.5.attn_norm.weight
    F32
    [3584]
  • blk.5.attn_output.weight
    (!unknown_type 23!)
    [4096, 3584]
  • blk.5.attn_q.weight
    (!unknown_type 23!)
    [3584, 4096]
  • blk.5.attn_v.weight
    (!unknown_type 23!)
    [3584, 2048]
  • blk.5.ffn_down.weight
    (!unknown_type 23!)
    [14336, 3584]
  • blk.5.ffn_gate.weight
    (!unknown_type 23!)
    [3584, 14336]
  • blk.5.ffn_norm.weight
    F32
    [3584]
  • blk.5.ffn_up.weight
    (!unknown_type 23!)
    [3584, 14336]
  • blk.5.post_attention_norm.weight
    F32
    [3584]
  • blk.5.post_ffw_norm.weight
    F32
    [3584]
    • blk.6
  • blk.6.attn_k.weight
    (!unknown_type 23!)
    [3584, 2048]
  • blk.6.attn_norm.weight
    F32
    [3584]
  • blk.6.attn_output.weight
    (!unknown_type 23!)
    [4096, 3584]
  • blk.6.attn_q.weight
    (!unknown_type 23!)
    [3584, 4096]
  • blk.6.attn_v.weight
    (!unknown_type 23!)
    [3584, 2048]
  • blk.6.ffn_down.weight
    (!unknown_type 23!)
    [14336, 3584]
  • blk.6.ffn_gate.weight
    (!unknown_type 23!)
    [3584, 14336]
  • blk.6.ffn_norm.weight
    F32
    [3584]
  • blk.6.ffn_up.weight
    (!unknown_type 23!)
    [3584, 14336]
  • blk.6.post_attention_norm.weight
    F32
    [3584]
  • blk.6.post_ffw_norm.weight
    F32
    [3584]
    • blk.7
  • blk.7.attn_k.weight
    (!unknown_type 23!)
    [3584, 2048]
  • blk.7.attn_norm.weight
    F32
    [3584]
  • blk.7.attn_output.weight
    (!unknown_type 23!)
    [4096, 3584]
  • blk.7.attn_q.weight
    (!unknown_type 23!)
    [3584, 4096]
  • blk.7.attn_v.weight
    (!unknown_type 23!)
    [3584, 2048]
  • blk.7.ffn_down.weight
    (!unknown_type 23!)
    [14336, 3584]
  • blk.7.ffn_gate.weight
    (!unknown_type 23!)
    [3584, 14336]
  • blk.7.ffn_norm.weight
    F32
    [3584]
  • blk.7.ffn_up.weight
    (!unknown_type 23!)
    [3584, 14336]
  • blk.7.post_attention_norm.weight
    F32
    [3584]
  • blk.7.post_ffw_norm.weight
    F32
    [3584]
    • blk.8
  • blk.8.attn_k.weight
    (!unknown_type 23!)
    [3584, 2048]
  • blk.8.attn_norm.weight
    F32
    [3584]
  • blk.8.attn_output.weight
    (!unknown_type 23!)
    [4096, 3584]
  • blk.8.attn_q.weight
    (!unknown_type 23!)
    [3584, 4096]
  • blk.8.attn_v.weight
    (!unknown_type 23!)
    [3584, 2048]
  • blk.8.ffn_down.weight
    (!unknown_type 23!)
    [14336, 3584]
  • blk.8.ffn_gate.weight
    (!unknown_type 23!)
    [3584, 14336]
  • blk.8.ffn_norm.weight
    F32
    [3584]
  • blk.8.ffn_up.weight
    (!unknown_type 23!)
    [3584, 14336]
  • blk.8.post_attention_norm.weight
    F32
    [3584]
  • blk.8.post_ffw_norm.weight
    F32
    [3584]
    • blk.9
  • blk.9.attn_k.weight
    (!unknown_type 23!)
    [3584, 2048]
  • blk.9.attn_norm.weight
    F32
    [3584]
  • blk.9.attn_output.weight
    (!unknown_type 23!)
    [4096, 3584]
  • blk.9.attn_q.weight
    (!unknown_type 23!)
    [3584, 4096]
  • blk.9.attn_v.weight
    (!unknown_type 23!)
    [3584, 2048]
  • blk.9.ffn_down.weight
    (!unknown_type 23!)
    [14336, 3584]
  • blk.9.ffn_gate.weight
    (!unknown_type 23!)
    [3584, 14336]
  • blk.9.ffn_norm.weight
    F32
    [3584]
  • blk.9.ffn_up.weight
    (!unknown_type 23!)
    [3584, 14336]
  • blk.9.post_attention_norm.weight
    F32
    [3584]
  • blk.9.post_ffw_norm.weight
    F32
    [3584]
    • blk.10
  • blk.10.attn_k.weight
    (!unknown_type 23!)
    [3584, 2048]
  • blk.10.attn_norm.weight
    F32
    [3584]
  • blk.10.attn_output.weight
    (!unknown_type 23!)
    [4096, 3584]
  • blk.10.attn_q.weight
    (!unknown_type 23!)
    [3584, 4096]
  • blk.10.attn_v.weight
    (!unknown_type 23!)
    [3584, 2048]
  • blk.10.ffn_down.weight
    (!unknown_type 23!)
    [14336, 3584]
  • blk.10.ffn_gate.weight
    (!unknown_type 23!)
    [3584, 14336]
  • blk.10.ffn_norm.weight
    F32
    [3584]
  • blk.10.ffn_up.weight
    (!unknown_type 23!)
    [3584, 14336]
  • blk.10.post_attention_norm.weight
    F32
    [3584]
  • blk.10.post_ffw_norm.weight
    F32
    [3584]
    • blk.11
  • blk.11.attn_k.weight
    (!unknown_type 23!)
    [3584, 2048]
  • blk.11.attn_norm.weight
    F32
    [3584]
  • blk.11.attn_output.weight
    (!unknown_type 23!)
    [4096, 3584]
  • blk.11.attn_q.weight
    (!unknown_type 23!)
    [3584, 4096]
  • blk.11.attn_v.weight
    (!unknown_type 23!)
    [3584, 2048]
  • blk.11.ffn_down.weight
    (!unknown_type 23!)
    [14336, 3584]
  • blk.11.ffn_gate.weight
    (!unknown_type 23!)
    [3584, 14336]
  • blk.11.ffn_norm.weight
    F32
    [3584]
  • blk.11.ffn_up.weight
    (!unknown_type 23!)
    [3584, 14336]
  • blk.11.post_attention_norm.weight
    F32
    [3584]
  • blk.11.post_ffw_norm.weight
    F32
    [3584]
    • blk.12
  • blk.12.attn_k.weight
    (!unknown_type 23!)
    [3584, 2048]
  • blk.12.attn_norm.weight
    F32
    [3584]
  • blk.12.attn_output.weight
    (!unknown_type 23!)
    [4096, 3584]
  • blk.12.attn_q.weight
    (!unknown_type 23!)
    [3584, 4096]
  • blk.12.attn_v.weight
    (!unknown_type 23!)
    [3584, 2048]
  • blk.12.ffn_down.weight
    (!unknown_type 23!)
    [14336, 3584]
  • blk.12.ffn_gate.weight
    (!unknown_type 23!)
    [3584, 14336]
  • blk.12.ffn_norm.weight
    F32
    [3584]
  • blk.12.ffn_up.weight
    (!unknown_type 23!)
    [3584, 14336]
  • blk.12.post_attention_norm.weight
    F32
    [3584]
  • blk.12.post_ffw_norm.weight
    F32
    [3584]
    • blk.13
  • blk.13.attn_k.weight
    (!unknown_type 23!)
    [3584, 2048]
  • blk.13.attn_norm.weight
    F32
    [3584]
  • blk.13.attn_output.weight
    (!unknown_type 23!)
    [4096, 3584]
  • blk.13.attn_q.weight
    (!unknown_type 23!)
    [3584, 4096]
  • blk.13.attn_v.weight
    (!unknown_type 23!)
    [3584, 2048]
  • blk.13.ffn_down.weight
    (!unknown_type 23!)
    [14336, 3584]
  • blk.13.ffn_gate.weight
    (!unknown_type 23!)
    [3584, 14336]
  • blk.13.ffn_norm.weight
    F32
    [3584]
  • blk.13.ffn_up.weight
    (!unknown_type 23!)
    [3584, 14336]
  • blk.13.post_attention_norm.weight
    F32
    [3584]
  • blk.13.post_ffw_norm.weight
    F32
    [3584]
    • blk.14
  • blk.14.attn_k.weight
    (!unknown_type 23!)
    [3584, 2048]
  • blk.14.attn_norm.weight
    F32
    [3584]
  • blk.14.attn_output.weight
    (!unknown_type 23!)
    [4096, 3584]
  • blk.14.attn_q.weight
    (!unknown_type 23!)
    [3584, 4096]
  • blk.14.attn_v.weight
    (!unknown_type 23!)
    [3584, 2048]
  • blk.14.ffn_down.weight
    (!unknown_type 23!)
    [14336, 3584]
  • blk.14.ffn_gate.weight
    (!unknown_type 23!)
    [3584, 14336]
  • blk.14.ffn_norm.weight
    F32
    [3584]
  • blk.14.ffn_up.weight
    (!unknown_type 23!)
    [3584, 14336]
  • blk.14.post_attention_norm.weight
    F32
    [3584]
  • blk.14.post_ffw_norm.weight
    F32
    [3584]
    • blk.15
  • blk.15.attn_k.weight
    (!unknown_type 23!)
    [3584, 2048]
  • blk.15.attn_norm.weight
    F32
    [3584]
  • blk.15.attn_output.weight
    (!unknown_type 23!)
    [4096, 3584]
  • blk.15.attn_q.weight
    (!unknown_type 23!)
    [3584, 4096]
  • blk.15.attn_v.weight
    (!unknown_type 23!)
    [3584, 2048]
  • blk.15.ffn_down.weight
    (!unknown_type 23!)
    [14336, 3584]
  • blk.15.ffn_gate.weight
    (!unknown_type 23!)
    [3584, 14336]
  • blk.15.ffn_norm.weight
    F32
    [3584]
  • blk.15.ffn_up.weight
    (!unknown_type 23!)
    [3584, 14336]
  • blk.15.post_attention_norm.weight
    F32
    [3584]
  • blk.15.post_ffw_norm.weight
    F32
    [3584]
    • blk.16
  • blk.16.attn_k.weight
    (!unknown_type 23!)
    [3584, 2048]
  • blk.16.attn_norm.weight
    F32
    [3584]
  • blk.16.attn_output.weight
    (!unknown_type 23!)
    [4096, 3584]
  • blk.16.attn_q.weight
    (!unknown_type 23!)
    [3584, 4096]
  • blk.16.attn_v.weight
    (!unknown_type 23!)
    [3584, 2048]
  • blk.16.ffn_down.weight
    (!unknown_type 23!)
    [14336, 3584]
  • blk.16.ffn_gate.weight
    (!unknown_type 23!)
    [3584, 14336]
  • blk.16.ffn_norm.weight
    F32
    [3584]
  • blk.16.ffn_up.weight
    (!unknown_type 23!)
    [3584, 14336]
  • blk.16.post_attention_norm.weight
    F32
    [3584]
  • blk.16.post_ffw_norm.weight
    F32
    [3584]
    • blk.17
  • blk.17.attn_k.weight
    (!unknown_type 23!)
    [3584, 2048]
  • blk.17.attn_norm.weight
    F32
    [3584]
  • blk.17.attn_output.weight
    (!unknown_type 23!)
    [4096, 3584]
  • blk.17.attn_q.weight
    (!unknown_type 23!)
    [3584, 4096]
  • blk.17.attn_v.weight
    (!unknown_type 23!)
    [3584, 2048]
  • blk.17.ffn_down.weight
    (!unknown_type 23!)
    [14336, 3584]
  • blk.17.ffn_gate.weight
    (!unknown_type 23!)
    [3584, 14336]
  • blk.17.ffn_norm.weight
    F32
    [3584]
  • blk.17.ffn_up.weight
    (!unknown_type 23!)
    [3584, 14336]
  • blk.17.post_attention_norm.weight
    F32
    [3584]
  • blk.17.post_ffw_norm.weight
    F32
    [3584]
    • blk.18
  • blk.18.attn_k.weight
    (!unknown_type 23!)
    [3584, 2048]
  • blk.18.attn_norm.weight
    F32
    [3584]
  • blk.18.attn_output.weight
    (!unknown_type 23!)
    [4096, 3584]
  • blk.18.attn_q.weight
    (!unknown_type 23!)
    [3584, 4096]
  • blk.18.attn_v.weight
    (!unknown_type 23!)
    [3584, 2048]
  • blk.18.ffn_down.weight
    (!unknown_type 23!)
    [14336, 3584]
  • blk.18.ffn_gate.weight
    (!unknown_type 23!)
    [3584, 14336]
  • blk.18.ffn_norm.weight
    F32
    [3584]
  • blk.18.ffn_up.weight
    (!unknown_type 23!)
    [3584, 14336]
  • blk.18.post_attention_norm.weight
    F32
    [3584]
  • blk.18.post_ffw_norm.weight
    F32
    [3584]
    • blk.19
  • blk.19.attn_k.weight
    (!unknown_type 23!)
    [3584, 2048]
  • blk.19.attn_norm.weight
    F32
    [3584]
  • blk.19.attn_output.weight
    (!unknown_type 23!)
    [4096, 3584]
  • blk.19.attn_q.weight
    (!unknown_type 23!)
    [3584, 4096]
  • blk.19.attn_v.weight
    (!unknown_type 23!)
    [3584, 2048]
  • blk.19.ffn_down.weight
    (!unknown_type 23!)
    [14336, 3584]
  • blk.19.ffn_gate.weight
    (!unknown_type 23!)
    [3584, 14336]
  • blk.19.ffn_norm.weight
    F32
    [3584]
  • blk.19.ffn_up.weight
    (!unknown_type 23!)
    [3584, 14336]
  • blk.19.post_attention_norm.weight
    F32
    [3584]
  • blk.19.post_ffw_norm.weight
    F32
    [3584]
    • blk.20
  • blk.20.attn_k.weight
    (!unknown_type 23!)
    [3584, 2048]
  • blk.20.attn_norm.weight
    F32
    [3584]
  • blk.20.attn_output.weight
    (!unknown_type 23!)
    [4096, 3584]
  • blk.20.attn_q.weight
    (!unknown_type 23!)
    [3584, 4096]
  • blk.20.attn_v.weight
    (!unknown_type 23!)
    [3584, 2048]
  • blk.20.ffn_down.weight
    (!unknown_type 23!)
    [14336, 3584]
  • blk.20.ffn_gate.weight
    (!unknown_type 23!)
    [3584, 14336]
  • blk.20.ffn_norm.weight
    F32
    [3584]
  • blk.20.ffn_up.weight
    (!unknown_type 23!)
    [3584, 14336]
  • blk.20.post_attention_norm.weight
    F32
    [3584]
  • blk.20.post_ffw_norm.weight
    F32
    [3584]
    • blk.21
  • blk.21.attn_k.weight
    (!unknown_type 23!)
    [3584, 2048]
  • blk.21.attn_norm.weight
    F32
    [3584]
  • blk.21.attn_output.weight
    (!unknown_type 23!)
    [4096, 3584]
  • blk.21.attn_q.weight
    (!unknown_type 23!)
    [3584, 4096]
  • blk.21.attn_v.weight
    (!unknown_type 23!)
    [3584, 2048]
  • blk.21.ffn_down.weight
    (!unknown_type 23!)
    [14336, 3584]
  • blk.21.ffn_gate.weight
    (!unknown_type 23!)
    [3584, 14336]
  • blk.21.ffn_norm.weight
    F32
    [3584]
  • blk.21.ffn_up.weight
    (!unknown_type 23!)
    [3584, 14336]
  • blk.21.post_attention_norm.weight
    F32
    [3584]
  • blk.21.post_ffw_norm.weight
    F32
    [3584]
    • blk.22
  • blk.22.attn_k.weight
    (!unknown_type 23!)
    [3584, 2048]
  • blk.22.attn_norm.weight
    F32
    [3584]
  • blk.22.attn_output.weight
    (!unknown_type 23!)
    [4096, 3584]
  • blk.22.attn_q.weight
    (!unknown_type 23!)
    [3584, 4096]
  • blk.22.attn_v.weight
    (!unknown_type 23!)
    [3584, 2048]
  • blk.22.ffn_down.weight
    (!unknown_type 23!)
    [14336, 3584]
  • blk.22.ffn_gate.weight
    (!unknown_type 23!)
    [3584, 14336]
  • blk.22.ffn_norm.weight
    F32
    [3584]
  • blk.22.ffn_up.weight
    (!unknown_type 23!)
    [3584, 14336]
  • blk.22.post_attention_norm.weight
    F32
    [3584]
  • blk.22.post_ffw_norm.weight
    F32
    [3584]
    • blk.23
  • blk.23.attn_k.weight
    (!unknown_type 23!)
    [3584, 2048]
  • blk.23.attn_norm.weight
    F32
    [3584]
  • blk.23.attn_output.weight
    (!unknown_type 23!)
    [4096, 3584]
  • blk.23.attn_q.weight
    (!unknown_type 23!)
    [3584, 4096]
  • blk.23.attn_v.weight
    (!unknown_type 23!)
    [3584, 2048]
  • blk.23.ffn_down.weight
    (!unknown_type 23!)
    [14336, 3584]
  • blk.23.ffn_gate.weight
    (!unknown_type 23!)
    [3584, 14336]
  • blk.23.ffn_norm.weight
    F32
    [3584]
  • blk.23.ffn_up.weight
    (!unknown_type 23!)
    [3584, 14336]
  • blk.23.post_attention_norm.weight
    F32
    [3584]
  • blk.23.post_ffw_norm.weight
    F32
    [3584]
    • blk.24
  • blk.24.attn_k.weight
    (!unknown_type 23!)
    [3584, 2048]
  • blk.24.attn_norm.weight
    F32
    [3584]
  • blk.24.attn_output.weight
    (!unknown_type 23!)
    [4096, 3584]
  • blk.24.attn_q.weight
    (!unknown_type 23!)
    [3584, 4096]
  • blk.24.attn_v.weight
    (!unknown_type 23!)
    [3584, 2048]
  • blk.24.ffn_down.weight
    (!unknown_type 23!)
    [14336, 3584]
  • blk.24.ffn_gate.weight
    (!unknown_type 23!)
    [3584, 14336]
  • blk.24.ffn_norm.weight
    F32
    [3584]
  • blk.24.ffn_up.weight
    (!unknown_type 23!)
    [3584, 14336]
  • blk.24.post_attention_norm.weight
    F32
    [3584]
  • blk.24.post_ffw_norm.weight
    F32
    [3584]
    • blk.25
  • blk.25.attn_k.weight
    (!unknown_type 23!)
    [3584, 2048]
  • blk.25.attn_norm.weight
    F32
    [3584]
  • blk.25.attn_output.weight
    (!unknown_type 23!)
    [4096, 3584]
  • blk.25.attn_q.weight
    (!unknown_type 23!)
    [3584, 4096]
  • blk.25.attn_v.weight
    (!unknown_type 23!)
    [3584, 2048]
  • blk.25.ffn_down.weight
    (!unknown_type 23!)
    [14336, 3584]
  • blk.25.ffn_gate.weight
    (!unknown_type 23!)
    [3584, 14336]
  • blk.25.ffn_norm.weight
    F32
    [3584]
  • blk.25.ffn_up.weight
    (!unknown_type 23!)
    [3584, 14336]
  • blk.25.post_attention_norm.weight
    F32
    [3584]
  • blk.25.post_ffw_norm.weight
    F32
    [3584]
    • blk.26
  • blk.26.attn_k.weight
    (!unknown_type 23!)
    [3584, 2048]
  • blk.26.attn_norm.weight
    F32
    [3584]
  • blk.26.attn_output.weight
    (!unknown_type 23!)
    [4096, 3584]
  • blk.26.attn_q.weight
    (!unknown_type 23!)
    [3584, 4096]
  • blk.26.attn_v.weight
    (!unknown_type 23!)
    [3584, 2048]
  • blk.26.ffn_down.weight
    (!unknown_type 23!)
    [14336, 3584]
  • blk.26.ffn_gate.weight
    (!unknown_type 23!)
    [3584, 14336]
  • blk.26.ffn_norm.weight
    F32
    [3584]
  • blk.26.ffn_up.weight
    (!unknown_type 23!)
    [3584, 14336]
  • blk.26.post_attention_norm.weight
    F32
    [3584]
  • blk.26.post_ffw_norm.weight
    F32
    [3584]
    • blk.27
  • blk.27.attn_k.weight
    (!unknown_type 23!)
    [3584, 2048]
  • blk.27.attn_norm.weight
    F32
    [3584]
  • blk.27.attn_output.weight
    (!unknown_type 23!)
    [4096, 3584]
  • blk.27.attn_q.weight
    (!unknown_type 23!)
    [3584, 4096]
  • blk.27.attn_v.weight
    (!unknown_type 23!)
    [3584, 2048]
  • blk.27.ffn_down.weight
    (!unknown_type 23!)
    [14336, 3584]
  • blk.27.ffn_gate.weight
    (!unknown_type 23!)
    [3584, 14336]
  • blk.27.ffn_norm.weight
    F32
    [3584]
  • blk.27.ffn_up.weight
    (!unknown_type 23!)
    [3584, 14336]
  • blk.27.post_attention_norm.weight
    F32
    [3584]
  • blk.27.post_ffw_norm.weight
    F32
    [3584]
    • blk.28
  • blk.28.attn_k.weight
    (!unknown_type 23!)
    [3584, 2048]
  • blk.28.attn_norm.weight
    F32
    [3584]
  • blk.28.attn_output.weight
    (!unknown_type 23!)
    [4096, 3584]
  • blk.28.attn_q.weight
    (!unknown_type 23!)
    [3584, 4096]
  • blk.28.attn_v.weight
    (!unknown_type 23!)
    [3584, 2048]
  • blk.28.ffn_down.weight
    (!unknown_type 23!)
    [14336, 3584]
  • blk.28.ffn_gate.weight
    (!unknown_type 23!)
    [3584, 14336]
  • blk.28.ffn_norm.weight
    F32
    [3584]
  • blk.28.ffn_up.weight
    (!unknown_type 23!)
    [3584, 14336]
  • blk.28.post_attention_norm.weight
    F32
    [3584]
  • blk.28.post_ffw_norm.weight
    F32
    [3584]
    • blk.29
  • blk.29.attn_k.weight
    (!unknown_type 23!)
    [3584, 2048]
  • blk.29.attn_norm.weight
    F32
    [3584]
  • blk.29.attn_output.weight
    (!unknown_type 23!)
    [4096, 3584]
  • blk.29.attn_q.weight
    (!unknown_type 23!)
    [3584, 4096]
  • blk.29.attn_v.weight
    (!unknown_type 23!)
    [3584, 2048]
  • blk.29.ffn_down.weight
    (!unknown_type 23!)
    [14336, 3584]
  • blk.29.ffn_gate.weight
    (!unknown_type 23!)
    [3584, 14336]
  • blk.29.ffn_norm.weight
    F32
    [3584]
  • blk.29.ffn_up.weight
    (!unknown_type 23!)
    [3584, 14336]
  • blk.29.post_attention_norm.weight
    F32
    [3584]
  • blk.29.post_ffw_norm.weight
    F32
    [3584]
    • blk.30
  • blk.30.attn_k.weight
    (!unknown_type 23!)
    [3584, 2048]
  • blk.30.attn_norm.weight
    F32
    [3584]
  • blk.30.attn_output.weight
    (!unknown_type 23!)
    [4096, 3584]
  • blk.30.attn_q.weight
    (!unknown_type 23!)
    [3584, 4096]
  • blk.30.attn_v.weight
    (!unknown_type 23!)
    [3584, 2048]
  • blk.30.ffn_down.weight
    (!unknown_type 23!)
    [14336, 3584]
  • blk.30.ffn_gate.weight
    (!unknown_type 23!)
    [3584, 14336]
  • blk.30.ffn_norm.weight
    F32
    [3584]
  • blk.30.ffn_up.weight
    (!unknown_type 23!)
    [3584, 14336]
  • blk.30.post_attention_norm.weight
    F32
    [3584]
  • blk.30.post_ffw_norm.weight
    F32
    [3584]
    • blk.31
  • blk.31.attn_k.weight
    (!unknown_type 23!)
    [3584, 2048]
  • blk.31.attn_norm.weight
    F32
    [3584]
  • blk.31.attn_output.weight
    (!unknown_type 23!)
    [4096, 3584]
  • blk.31.attn_q.weight
    (!unknown_type 23!)
    [3584, 4096]
  • blk.31.attn_v.weight
    (!unknown_type 23!)
    [3584, 2048]
  • blk.31.ffn_down.weight
    (!unknown_type 23!)
    [14336, 3584]
  • blk.31.ffn_gate.weight
    (!unknown_type 23!)
    [3584, 14336]
  • blk.31.ffn_norm.weight
    F32
    [3584]
  • blk.31.ffn_up.weight
    (!unknown_type 23!)
    [3584, 14336]
  • blk.31.post_attention_norm.weight
    F32
    [3584]
  • blk.31.post_ffw_norm.weight
    F32
    [3584]
    • blk.32
  • blk.32.attn_k.weight
    (!unknown_type 23!)
    [3584, 2048]
  • blk.32.attn_norm.weight
    F32
    [3584]
  • blk.32.attn_output.weight
    (!unknown_type 23!)
    [4096, 3584]
  • blk.32.attn_q.weight
    (!unknown_type 23!)
    [3584, 4096]
  • blk.32.attn_v.weight
    (!unknown_type 23!)
    [3584, 2048]
  • blk.32.ffn_down.weight
    (!unknown_type 23!)
    [14336, 3584]
  • blk.32.ffn_gate.weight
    (!unknown_type 23!)
    [3584, 14336]
  • blk.32.ffn_norm.weight
    F32
    [3584]
  • blk.32.ffn_up.weight
    (!unknown_type 23!)
    [3584, 14336]
  • blk.32.post_attention_norm.weight
    F32
    [3584]
  • blk.32.post_ffw_norm.weight
    F32
    [3584]
    • blk.33
  • blk.33.attn_k.weight
    (!unknown_type 23!)
    [3584, 2048]
  • blk.33.attn_norm.weight
    F32
    [3584]
  • blk.33.attn_output.weight
    (!unknown_type 23!)
    [4096, 3584]
  • blk.33.attn_q.weight
    (!unknown_type 23!)
    [3584, 4096]
  • blk.33.attn_v.weight
    (!unknown_type 23!)
    [3584, 2048]
  • blk.33.ffn_down.weight
    (!unknown_type 23!)
    [14336, 3584]
  • blk.33.ffn_gate.weight
    (!unknown_type 23!)
    [3584, 14336]
  • blk.33.ffn_norm.weight
    F32
    [3584]
  • blk.33.ffn_up.weight
    (!unknown_type 23!)
    [3584, 14336]
  • blk.33.post_attention_norm.weight
    F32
    [3584]
  • blk.33.post_ffw_norm.weight
    F32
    [3584]
    • blk.34
  • blk.34.attn_k.weight
    (!unknown_type 23!)
    [3584, 2048]
  • blk.34.attn_norm.weight
    F32
    [3584]
  • blk.34.attn_output.weight
    (!unknown_type 23!)
    [4096, 3584]
  • blk.34.attn_q.weight
    (!unknown_type 23!)
    [3584, 4096]
  • blk.34.attn_v.weight
    (!unknown_type 23!)
    [3584, 2048]
  • blk.34.ffn_down.weight
    (!unknown_type 23!)
    [14336, 3584]
  • blk.34.ffn_gate.weight
    (!unknown_type 23!)
    [3584, 14336]
  • blk.34.ffn_norm.weight
    F32
    [3584]
  • blk.34.ffn_up.weight
    (!unknown_type 23!)
    [3584, 14336]
  • blk.34.post_attention_norm.weight
    F32
    [3584]
  • blk.34.post_ffw_norm.weight
    F32
    [3584]
    • blk.35
  • blk.35.attn_k.weight
    (!unknown_type 23!)
    [3584, 2048]
  • blk.35.attn_norm.weight
    F32
    [3584]
  • blk.35.attn_output.weight
    (!unknown_type 23!)
    [4096, 3584]
  • blk.35.attn_q.weight
    (!unknown_type 23!)
    [3584, 4096]
  • blk.35.attn_v.weight
    (!unknown_type 23!)
    [3584, 2048]
  • blk.35.ffn_down.weight
    (!unknown_type 23!)
    [14336, 3584]
  • blk.35.ffn_gate.weight
    (!unknown_type 23!)
    [3584, 14336]
  • blk.35.ffn_norm.weight
    F32
    [3584]
  • blk.35.ffn_up.weight
    (!unknown_type 23!)
    [3584, 14336]
  • blk.35.post_attention_norm.weight
    F32
    [3584]
  • blk.35.post_ffw_norm.weight
    F32
    [3584]
    • blk.36
  • blk.36.attn_k.weight
    (!unknown_type 23!)
    [3584, 2048]
  • blk.36.attn_norm.weight
    F32
    [3584]
  • blk.36.attn_output.weight
    (!unknown_type 23!)
    [4096, 3584]
  • blk.36.attn_q.weight
    (!unknown_type 23!)
    [3584, 4096]
  • blk.36.attn_v.weight
    (!unknown_type 23!)
    [3584, 2048]
  • blk.36.ffn_down.weight
    (!unknown_type 23!)
    [14336, 3584]
  • blk.36.ffn_gate.weight
    (!unknown_type 23!)
    [3584, 14336]
  • blk.36.ffn_norm.weight
    F32
    [3584]
  • blk.36.ffn_up.weight
    (!unknown_type 23!)
    [3584, 14336]
  • blk.36.post_attention_norm.weight
    F32
    [3584]
  • blk.36.post_ffw_norm.weight
    F32
    [3584]
    • blk.37
  • blk.37.attn_k.weight
    (!unknown_type 23!)
    [3584, 2048]
  • blk.37.attn_norm.weight
    F32
    [3584]
  • blk.37.attn_output.weight
    (!unknown_type 23!)
    [4096, 3584]
  • blk.37.attn_q.weight
    (!unknown_type 23!)
    [3584, 4096]
  • blk.37.attn_v.weight
    (!unknown_type 23!)
    [3584, 2048]
  • blk.37.ffn_down.weight
    (!unknown_type 23!)
    [14336, 3584]
  • blk.37.ffn_gate.weight
    (!unknown_type 23!)
    [3584, 14336]
  • blk.37.ffn_norm.weight
    F32
    [3584]
  • blk.37.ffn_up.weight
    (!unknown_type 23!)
    [3584, 14336]
  • blk.37.post_attention_norm.weight
    F32
    [3584]
  • blk.37.post_ffw_norm.weight
    F32
    [3584]
    • blk.38
  • blk.38.attn_k.weight
    (!unknown_type 23!)
    [3584, 2048]
  • blk.38.attn_norm.weight
    F32
    [3584]
  • blk.38.attn_output.weight
    (!unknown_type 23!)
    [4096, 3584]
  • blk.38.attn_q.weight
    (!unknown_type 23!)
    [3584, 4096]
  • blk.38.attn_v.weight
    (!unknown_type 23!)
    [3584, 2048]
  • blk.38.ffn_down.weight
    (!unknown_type 23!)
    [14336, 3584]
  • blk.38.ffn_gate.weight
    (!unknown_type 23!)
    [3584, 14336]
  • blk.38.ffn_norm.weight
    F32
    [3584]
  • blk.38.ffn_up.weight
    (!unknown_type 23!)
    [3584, 14336]
  • blk.38.post_attention_norm.weight
    F32
    [3584]
  • blk.38.post_ffw_norm.weight
    F32
    [3584]
    • blk.39
  • blk.39.attn_k.weight
    (!unknown_type 23!)
    [3584, 2048]
  • blk.39.attn_norm.weight
    F32
    [3584]
  • blk.39.attn_output.weight
    (!unknown_type 23!)
    [4096, 3584]
  • blk.39.attn_q.weight
    (!unknown_type 23!)
    [3584, 4096]
  • blk.39.attn_v.weight
    (!unknown_type 23!)
    [3584, 2048]
  • blk.39.ffn_down.weight
    (!unknown_type 23!)
    [14336, 3584]
  • blk.39.ffn_gate.weight
    (!unknown_type 23!)
    [3584, 14336]
  • blk.39.ffn_norm.weight
    F32
    [3584]
  • blk.39.ffn_up.weight
    (!unknown_type 23!)
    [3584, 14336]
  • blk.39.post_attention_norm.weight
    F32
    [3584]
  • blk.39.post_ffw_norm.weight
    F32
    [3584]
    • blk.40
  • blk.40.attn_k.weight
    (!unknown_type 23!)
    [3584, 2048]
  • blk.40.attn_norm.weight
    F32
    [3584]
  • blk.40.attn_output.weight
    (!unknown_type 23!)
    [4096, 3584]
  • blk.40.attn_q.weight
    (!unknown_type 23!)
    [3584, 4096]
  • blk.40.attn_v.weight
    (!unknown_type 23!)
    [3584, 2048]
  • blk.40.ffn_down.weight
    (!unknown_type 23!)
    [14336, 3584]
  • blk.40.ffn_gate.weight
    (!unknown_type 23!)
    [3584, 14336]
  • blk.40.ffn_norm.weight
    F32
    [3584]
  • blk.40.ffn_up.weight
    (!unknown_type 23!)
    [3584, 14336]
  • blk.40.post_attention_norm.weight
    F32
    [3584]
  • blk.40.post_ffw_norm.weight
    F32
    [3584]
    • blk.41
  • blk.41.attn_k.weight
    (!unknown_type 23!)
    [3584, 2048]
  • blk.41.attn_norm.weight
    F32
    [3584]
  • blk.41.attn_output.weight
    (!unknown_type 23!)
    [4096, 3584]
  • blk.41.attn_q.weight
    (!unknown_type 23!)
    [3584, 4096]
  • blk.41.attn_v.weight
    (!unknown_type 23!)
    [3584, 2048]
  • blk.41.ffn_down.weight
    (!unknown_type 23!)
    [14336, 3584]
  • blk.41.ffn_gate.weight
    (!unknown_type 23!)
    [3584, 14336]
  • blk.41.ffn_norm.weight
    F32
    [3584]
  • blk.41.ffn_up.weight
    (!unknown_type 23!)
    [3584, 14336]
  • blk.41.post_attention_norm.weight
    F32
    [3584]
  • blk.41.post_ffw_norm.weight
    F32
    [3584]
  • output_norm.weight
    F32
    [3584]