70 2 hours ago

Text-only decision-model Clef-Flash GGUF files. Default: iq4_xs. TurinTech models, built with Artemis.

28db4324d25d · 4.1GB
    Metadata
  • general.architecture
    clef
  • general.file_type
    Q4_K_M
  • clef.attention.head_count
    16
  • clef.attention.head_count_kv
    4
  • clef.attention.key_length
    256
  • clef.attention.layer_norm_epsilon
    1e-05
  • clef.attention.layer_norm_rms_epsilon
    1e-06
  • clef.attention.recurrent_layers
    [true, true, true, false, true, ...]
  • clef.attention.value_length
    256
  • clef.block_count
    32
  • clef.context_length
    262144
  • clef.decision.block_count
    4
  • clef.decision.head_count
    16
  • clef.decision.routing_block_count
    2
  • clef.decision.type
    clef
  • clef.embedding_length
    4096
  • clef.feed_forward_length
    12288
  • clef.full_attention_interval
    4
  • clef.rope.dimension_count
    64
  • clef.rope.dimension_sections
    [11, 11, 10, 0]
  • clef.rope.freq_base
    1e+07
  • clef.ssm.conv_kernel
    4
  • clef.ssm.group_count
    16
  • clef.ssm.inner_size
    4096
  • clef.ssm.state_size
    128
  • clef.ssm.time_step_rank
    32
  • tokenizer.chat_template.systemone
    {{ "<|im_start|>system\nRead the complete state and schema. Decide every field jointly. Each answer must be exactly one of that field's allowed options.<|im_end|>\n<|im_start|>user\nSTATE:\n" }}{% if images %}{{ sep }}{% for image in images %}{{ image }}{% endfor %}{{ "\n" }}{% endif %}{{ sep }}{{ state if state is string else state | tojson(separators=[',', ':']) }}{{ sep }}{{ "\n\nSCHEMA FIELDS:\n" }}{% for q in questions %}{{ sep }}{{ "\nFIELD " }}{{ loop.index }}{{ "\nID: " }}{{ q.id }}{{ "\nTYPE: " }}{{ q.type }}{{ "\nINSTRUCTION: " }}{{ sep }}{{ mark_question }}{{ q.instructions if q.instructions is string else q.instructions | tojson(separators=[',', ':']) }}{{ sep }}{{ "\nALLOWED OPTIONS:\n" }}{% for o in q.options %}{{ sep }}{{ "OPTION " }}{{ loop.index }}{{ ": " }}{{ sep }}{{ mark_option }}{% set d = o.description %}{% if q.type == 'noul' and d is none %}{% set d = 'The proposition is true or the answer is yes.' if o.key == 'true' else 'The proposition is false or the answer is no.' %}{% endif %}{{ ({'option_id': o.key} if d is none else {'description': d, 'option_id': o.key}) | tojson(separators=[',', ':']) }}{{ sep }}{{ "\n" }}{% endfor %}{{ sep }}{{ "END FIELD\n" }}{% endfor %}{{ sep }}{{ "\n<|im_end|>\n<|im_start|>assistant\n<think>\n\n</think>\n\nJOINT SCHEMA DECISIONS:" }}
  • tokenizer.chat_templates
    [systemone]
  • tokenizer.ggml.add_bos_token
    false
  • tokenizer.ggml.add_eos_token
    false
  • tokenizer.ggml.eos_token_id
    248046
  • tokenizer.ggml.merges
    [Ġ Ġ, ĠĠ ĠĠ, i n, Ġ t, ĠĠĠĠ ĠĠĠĠ, ...]
  • tokenizer.ggml.model
    gpt2
  • tokenizer.ggml.padding_token_id
    248044
  • tokenizer.ggml.pre
    qwen35
  • tokenizer.ggml.token_type
    [1, 1, 1, 1, 1, ...]
  • tokenizer.ggml.tokens
    [!, ", #, $, %, ...]
  • Tensor
  • token_embd.weight
    IQ3_XXS
    [4096, 248320]
  • blk.0
  • blk.0.attn_gate.weight
    Q4_K
    [4096, 4096]
  • blk.0.attn_norm.weight
    F32
    [4096]
  • blk.0.attn_qkv.weight
    Q6_K
    [4096, 8192]
  • blk.0.ffn_down.weight
    IQ4_XS
    [12288, 4096]
  • blk.0.ffn_gate.weight
    IQ3_XXS
    [4096, 12288]
  • blk.0.ffn_up.weight
    IQ3_XXS
    [4096, 12288]
  • blk.0.post_attention_norm.weight
    F32
    [4096]
  • blk.0.ssm_a
    F32
    [32]
  • blk.0.ssm_alpha.weight
    Q4_K
    [4096, 32]
  • blk.0.ssm_beta.weight
    Q4_K
    [4096, 32]
  • blk.0.ssm_conv1d.weight
    F32
    [4, 8192]
  • blk.0.ssm_dt.bias
    F32
    [32]
  • blk.0.ssm_norm.weight
    F32
    [128]
  • blk.0.ssm_out.weight
    Q4_K
    [4096, 4096]
  • blk.1
  • blk.1.attn_gate.weight
    Q4_K
    [4096, 4096]
  • blk.1.attn_norm.weight
    F32
    [4096]
  • blk.1.attn_qkv.weight
    Q4_K
    [4096, 8192]
  • blk.1.ffn_down.weight
    IQ4_XS
    [12288, 4096]
  • blk.1.ffn_gate.weight
    IQ3_XXS
    [4096, 12288]
  • blk.1.ffn_up.weight
    IQ3_XXS
    [4096, 12288]
  • blk.1.post_attention_norm.weight
    F32
    [4096]
  • blk.1.ssm_a
    F32
    [32]
  • blk.1.ssm_alpha.weight
    Q4_K
    [4096, 32]
  • blk.1.ssm_beta.weight
    Q4_K
    [4096, 32]
  • blk.1.ssm_conv1d.weight
    F32
    [4, 8192]
  • blk.1.ssm_dt.bias
    F32
    [32]
  • blk.1.ssm_norm.weight
    F32
    [128]
  • blk.1.ssm_out.weight
    Q4_K
    [4096, 4096]
  • blk.2
  • blk.2.attn_gate.weight
    Q4_K
    [4096, 4096]
  • blk.2.attn_norm.weight
    F32
    [4096]
  • blk.2.attn_qkv.weight
    Q4_K
    [4096, 8192]
  • blk.2.ffn_down.weight
    IQ4_XS
    [12288, 4096]
  • blk.2.ffn_gate.weight
    IQ3_XXS
    [4096, 12288]
  • blk.2.ffn_up.weight
    IQ3_XXS
    [4096, 12288]
  • blk.2.post_attention_norm.weight
    F32
    [4096]
  • blk.2.ssm_a
    F32
    [32]
  • blk.2.ssm_alpha.weight
    Q4_K
    [4096, 32]
  • blk.2.ssm_beta.weight
    Q4_K
    [4096, 32]
  • blk.2.ssm_conv1d.weight
    F32
    [4, 8192]
  • blk.2.ssm_dt.bias
    F32
    [32]
  • blk.2.ssm_norm.weight
    F32
    [128]
  • blk.2.ssm_out.weight
    Q4_K
    [4096, 4096]
  • blk.3
  • blk.3.attn_k.weight
    Q4_K
    [4096, 1024]
  • blk.3.attn_k_norm.weight
    F32
    [256]
  • blk.3.attn_norm.weight
    F32
    [4096]
  • blk.3.attn_output.weight
    Q4_K
    [4096, 4096]
  • blk.3.attn_q.weight
    Q4_K
    [4096, 8192]
  • blk.3.attn_q_norm.weight
    F32
    [256]
  • blk.3.attn_v.weight
    Q6_K
    [4096, 1024]
  • blk.3.ffn_down.weight
    IQ4_XS
    [12288, 4096]
  • blk.3.ffn_gate.weight
    IQ3_XXS
    [4096, 12288]
  • blk.3.ffn_up.weight
    IQ3_XXS
    [4096, 12288]
  • blk.3.post_attention_norm.weight
    F32
    [4096]
  • blk.4
  • blk.4.attn_gate.weight
    Q4_K
    [4096, 4096]
  • blk.4.attn_norm.weight
    F32
    [4096]
  • blk.4.attn_qkv.weight
    Q4_K
    [4096, 8192]
  • blk.4.ffn_down.weight
    IQ4_XS
    [12288, 4096]
  • blk.4.ffn_gate.weight
    IQ3_XXS
    [4096, 12288]
  • blk.4.ffn_up.weight
    IQ3_XXS
    [4096, 12288]
  • blk.4.post_attention_norm.weight
    F32
    [4096]
  • blk.4.ssm_a
    F32
    [32]
  • blk.4.ssm_alpha.weight
    Q4_K
    [4096, 32]
  • blk.4.ssm_beta.weight
    Q4_K
    [4096, 32]
  • blk.4.ssm_conv1d.weight
    F32
    [4, 8192]
  • blk.4.ssm_dt.bias
    F32
    [32]
  • blk.4.ssm_norm.weight
    F32
    [128]
  • blk.4.ssm_out.weight
    Q4_K
    [4096, 4096]
  • blk.5
  • blk.5.attn_gate.weight
    Q4_K
    [4096, 4096]
  • blk.5.attn_norm.weight
    F32
    [4096]
  • blk.5.attn_qkv.weight
    Q4_K
    [4096, 8192]
  • blk.5.ffn_down.weight
    IQ4_XS
    [12288, 4096]
  • blk.5.ffn_gate.weight
    IQ3_XXS
    [4096, 12288]
  • blk.5.ffn_up.weight
    IQ3_XXS
    [4096, 12288]
  • blk.5.post_attention_norm.weight
    F32
    [4096]
  • blk.5.ssm_a
    F32
    [32]
  • blk.5.ssm_alpha.weight
    Q4_K
    [4096, 32]
  • blk.5.ssm_beta.weight
    Q4_K
    [4096, 32]
  • blk.5.ssm_conv1d.weight
    F32
    [4, 8192]
  • blk.5.ssm_dt.bias
    F32
    [32]
  • blk.5.ssm_norm.weight
    F32
    [128]
  • blk.5.ssm_out.weight
    Q4_K
    [4096, 4096]
  • blk.6
  • blk.6.attn_gate.weight
    Q4_K
    [4096, 4096]
  • blk.6.attn_norm.weight
    F32
    [4096]
  • blk.6.attn_qkv.weight
    Q6_K
    [4096, 8192]
  • blk.6.ffn_down.weight
    IQ4_XS
    [12288, 4096]
  • blk.6.ffn_gate.weight
    IQ3_XXS
    [4096, 12288]
  • blk.6.ffn_up.weight
    IQ3_XXS
    [4096, 12288]
  • blk.6.post_attention_norm.weight
    F32
    [4096]
  • blk.6.ssm_a
    F32
    [32]
  • blk.6.ssm_alpha.weight
    Q4_K
    [4096, 32]
  • blk.6.ssm_beta.weight
    Q4_K
    [4096, 32]
  • blk.6.ssm_conv1d.weight
    F32
    [4, 8192]
  • blk.6.ssm_dt.bias
    F32
    [32]
  • blk.6.ssm_norm.weight
    F32
    [128]
  • blk.6.ssm_out.weight
    Q4_K
    [4096, 4096]
  • blk.7
  • blk.7.attn_k.weight
    Q4_K
    [4096, 1024]
  • blk.7.attn_k_norm.weight
    F32
    [256]
  • blk.7.attn_norm.weight
    F32
    [4096]
  • blk.7.attn_output.weight
    Q4_K
    [4096, 4096]
  • blk.7.attn_q.weight
    Q4_K
    [4096, 8192]
  • blk.7.attn_q_norm.weight
    F32
    [256]
  • blk.7.attn_v.weight
    Q4_K
    [4096, 1024]
  • blk.7.ffn_down.weight
    IQ4_XS
    [12288, 4096]
  • blk.7.ffn_gate.weight
    IQ3_XXS
    [4096, 12288]
  • blk.7.ffn_up.weight
    IQ3_XXS
    [4096, 12288]
  • blk.7.post_attention_norm.weight
    F32
    [4096]
  • blk.8
  • blk.8.attn_gate.weight
    Q4_K
    [4096, 4096]
  • blk.8.attn_norm.weight
    F32
    [4096]
  • blk.8.attn_qkv.weight
    Q4_K
    [4096, 8192]
  • blk.8.ffn_down.weight
    IQ4_XS
    [12288, 4096]
  • blk.8.ffn_gate.weight
    IQ3_XXS
    [4096, 12288]
  • blk.8.ffn_up.weight
    IQ3_XXS
    [4096, 12288]
  • blk.8.post_attention_norm.weight
    F32
    [4096]
  • blk.8.ssm_a
    F32
    [32]
  • blk.8.ssm_alpha.weight
    Q4_K
    [4096, 32]
  • blk.8.ssm_beta.weight
    Q4_K
    [4096, 32]
  • blk.8.ssm_conv1d.weight
    F32
    [4, 8192]
  • blk.8.ssm_dt.bias
    F32
    [32]
  • blk.8.ssm_norm.weight
    F32
    [128]
  • blk.8.ssm_out.weight
    Q4_K
    [4096, 4096]
  • blk.9
  • blk.9.attn_gate.weight
    Q4_K
    [4096, 4096]
  • blk.9.attn_norm.weight
    F32
    [4096]
  • blk.9.attn_qkv.weight
    Q6_K
    [4096, 8192]
  • blk.9.ffn_down.weight
    IQ4_XS
    [12288, 4096]
  • blk.9.ffn_gate.weight
    IQ3_XXS
    [4096, 12288]
  • blk.9.ffn_up.weight
    IQ3_XXS
    [4096, 12288]
  • blk.9.post_attention_norm.weight
    F32
    [4096]
  • blk.9.ssm_a
    F32
    [32]
  • blk.9.ssm_alpha.weight
    Q4_K
    [4096, 32]
  • blk.9.ssm_beta.weight
    Q4_K
    [4096, 32]
  • blk.9.ssm_conv1d.weight
    F32
    [4, 8192]
  • blk.9.ssm_dt.bias
    F32
    [32]
  • blk.9.ssm_norm.weight
    F32
    [128]
  • blk.9.ssm_out.weight
    Q4_K
    [4096, 4096]
  • blk.10
  • blk.10.attn_gate.weight
    Q4_K
    [4096, 4096]
  • blk.10.attn_norm.weight
    F32
    [4096]
  • blk.10.attn_qkv.weight
    Q4_K
    [4096, 8192]
  • blk.10.ffn_down.weight
    IQ4_XS
    [12288, 4096]
  • blk.10.ffn_gate.weight
    IQ3_XXS
    [4096, 12288]
  • blk.10.ffn_up.weight
    IQ3_XXS
    [4096, 12288]
  • blk.10.post_attention_norm.weight
    F32
    [4096]
  • blk.10.ssm_a
    F32
    [32]
  • blk.10.ssm_alpha.weight
    Q4_K
    [4096, 32]
  • blk.10.ssm_beta.weight
    Q4_K
    [4096, 32]
  • blk.10.ssm_conv1d.weight
    F32
    [4, 8192]
  • blk.10.ssm_dt.bias
    F32
    [32]
  • blk.10.ssm_norm.weight
    F32
    [128]
  • blk.10.ssm_out.weight
    Q4_K
    [4096, 4096]
  • blk.11
  • blk.11.attn_k.weight
    Q4_K
    [4096, 1024]
  • blk.11.attn_k_norm.weight
    F32
    [256]
  • blk.11.attn_norm.weight
    F32
    [4096]
  • blk.11.attn_output.weight
    Q4_K
    [4096, 4096]
  • blk.11.attn_q.weight
    Q4_K
    [4096, 8192]
  • blk.11.attn_q_norm.weight
    F32
    [256]
  • blk.11.attn_v.weight
    Q4_K
    [4096, 1024]
  • blk.11.ffn_down.weight
    IQ4_XS
    [12288, 4096]
  • blk.11.ffn_gate.weight
    IQ3_XXS
    [4096, 12288]
  • blk.11.ffn_up.weight
    IQ3_XXS
    [4096, 12288]
  • blk.11.post_attention_norm.weight
    F32
    [4096]
  • blk.12
  • blk.12.attn_gate.weight
    Q4_K
    [4096, 4096]
  • blk.12.attn_norm.weight
    F32
    [4096]
  • blk.12.attn_qkv.weight
    Q6_K
    [4096, 8192]
  • blk.12.ffn_down.weight
    IQ4_XS
    [12288, 4096]
  • blk.12.ffn_gate.weight
    IQ3_XXS
    [4096, 12288]
  • blk.12.ffn_up.weight
    IQ3_XXS
    [4096, 12288]
  • blk.12.post_attention_norm.weight
    F32
    [4096]
  • blk.12.ssm_a
    F32
    [32]
  • blk.12.ssm_alpha.weight
    Q4_K
    [4096, 32]
  • blk.12.ssm_beta.weight
    Q4_K
    [4096, 32]
  • blk.12.ssm_conv1d.weight
    F32
    [4, 8192]
  • blk.12.ssm_dt.bias
    F32
    [32]
  • blk.12.ssm_norm.weight
    F32
    [128]
  • blk.12.ssm_out.weight
    Q4_K
    [4096, 4096]
  • blk.13
  • blk.13.attn_gate.weight
    Q4_K
    [4096, 4096]
  • blk.13.attn_norm.weight
    F32
    [4096]
  • blk.13.attn_qkv.weight
    Q4_K
    [4096, 8192]
  • blk.13.ffn_down.weight
    IQ4_XS
    [12288, 4096]
  • blk.13.ffn_gate.weight
    IQ3_XXS
    [4096, 12288]
  • blk.13.ffn_up.weight
    IQ3_XXS
    [4096, 12288]
  • blk.13.post_attention_norm.weight
    F32
    [4096]
  • blk.13.ssm_a
    F32
    [32]
  • blk.13.ssm_alpha.weight
    Q4_K
    [4096, 32]
  • blk.13.ssm_beta.weight
    Q4_K
    [4096, 32]
  • blk.13.ssm_conv1d.weight
    F32
    [4, 8192]
  • blk.13.ssm_dt.bias
    F32
    [32]
  • blk.13.ssm_norm.weight
    F32
    [128]
  • blk.13.ssm_out.weight
    Q4_K
    [4096, 4096]
  • blk.14
  • blk.14.attn_gate.weight
    Q4_K
    [4096, 4096]
  • blk.14.attn_norm.weight
    F32
    [4096]
  • blk.14.attn_qkv.weight
    Q4_K
    [4096, 8192]
  • blk.14.ffn_down.weight
    IQ4_XS
    [12288, 4096]
  • blk.14.ffn_gate.weight
    IQ3_XXS
    [4096, 12288]
  • blk.14.ffn_up.weight
    IQ3_XXS
    [4096, 12288]
  • blk.14.post_attention_norm.weight
    F32
    [4096]
  • blk.14.ssm_a
    F32
    [32]
  • blk.14.ssm_alpha.weight
    Q4_K
    [4096, 32]
  • blk.14.ssm_beta.weight
    Q4_K
    [4096, 32]
  • blk.14.ssm_conv1d.weight
    F32
    [4, 8192]
  • blk.14.ssm_dt.bias
    F32
    [32]
  • blk.14.ssm_norm.weight
    F32
    [128]
  • blk.14.ssm_out.weight
    Q4_K
    [4096, 4096]
  • blk.15
  • blk.15.attn_k.weight
    Q4_K
    [4096, 1024]
  • blk.15.attn_k_norm.weight
    F32
    [256]
  • blk.15.attn_norm.weight
    F32
    [4096]
  • blk.15.attn_output.weight
    Q4_K
    [4096, 4096]
  • blk.15.attn_q.weight
    Q4_K
    [4096, 8192]
  • blk.15.attn_q_norm.weight
    F32
    [256]
  • blk.15.attn_v.weight
    Q6_K
    [4096, 1024]
  • blk.15.ffn_down.weight
    IQ4_XS
    [12288, 4096]
  • blk.15.ffn_gate.weight
    IQ3_XXS
    [4096, 12288]
  • blk.15.ffn_up.weight
    IQ3_XXS
    [4096, 12288]
  • blk.15.post_attention_norm.weight
    F32
    [4096]
  • blk.16
  • blk.16.attn_gate.weight
    Q4_K
    [4096, 4096]
  • blk.16.attn_norm.weight
    F32
    [4096]
  • blk.16.attn_qkv.weight
    Q4_K
    [4096, 8192]
  • blk.16.ffn_down.weight
    IQ4_XS
    [12288, 4096]
  • blk.16.ffn_gate.weight
    IQ3_XXS
    [4096, 12288]
  • blk.16.ffn_up.weight
    IQ3_XXS
    [4096, 12288]
  • blk.16.post_attention_norm.weight
    F32
    [4096]
  • blk.16.ssm_a
    F32
    [32]
  • blk.16.ssm_alpha.weight
    Q4_K
    [4096, 32]
  • blk.16.ssm_beta.weight
    Q4_K
    [4096, 32]
  • blk.16.ssm_conv1d.weight
    F32
    [4, 8192]
  • blk.16.ssm_dt.bias
    F32
    [32]
  • blk.16.ssm_norm.weight
    F32
    [128]
  • blk.16.ssm_out.weight
    Q4_K
    [4096, 4096]
  • blk.17
  • blk.17.attn_gate.weight
    Q4_K
    [4096, 4096]
  • blk.17.attn_norm.weight
    F32
    [4096]
  • blk.17.attn_qkv.weight
    Q4_K
    [4096, 8192]
  • blk.17.ffn_down.weight
    IQ4_XS
    [12288, 4096]
  • blk.17.ffn_gate.weight
    IQ3_XXS
    [4096, 12288]
  • blk.17.ffn_up.weight
    IQ3_XXS
    [4096, 12288]
  • blk.17.post_attention_norm.weight
    F32
    [4096]
  • blk.17.ssm_a
    F32
    [32]
  • blk.17.ssm_alpha.weight
    Q4_K
    [4096, 32]
  • blk.17.ssm_beta.weight
    Q4_K
    [4096, 32]
  • blk.17.ssm_conv1d.weight
    F32
    [4, 8192]
  • blk.17.ssm_dt.bias
    F32
    [32]
  • blk.17.ssm_norm.weight
    F32
    [128]
  • blk.17.ssm_out.weight
    Q4_K
    [4096, 4096]
  • blk.18
  • blk.18.attn_gate.weight
    Q4_K
    [4096, 4096]
  • blk.18.attn_norm.weight
    F32
    [4096]
  • blk.18.attn_qkv.weight
    Q6_K
    [4096, 8192]
  • blk.18.ffn_down.weight
    IQ4_XS
    [12288, 4096]
  • blk.18.ffn_gate.weight
    IQ3_XXS
    [4096, 12288]
  • blk.18.ffn_up.weight
    IQ3_XXS
    [4096, 12288]
  • blk.18.post_attention_norm.weight
    F32
    [4096]
  • blk.18.ssm_a
    F32
    [32]
  • blk.18.ssm_alpha.weight
    Q4_K
    [4096, 32]
  • blk.18.ssm_beta.weight
    Q4_K
    [4096, 32]
  • blk.18.ssm_conv1d.weight
    F32
    [4, 8192]
  • blk.18.ssm_dt.bias
    F32
    [32]
  • blk.18.ssm_norm.weight
    F32
    [128]
  • blk.18.ssm_out.weight
    Q4_K
    [4096, 4096]
  • blk.19
  • blk.19.attn_k.weight
    Q4_K
    [4096, 1024]
  • blk.19.attn_k_norm.weight
    F32
    [256]
  • blk.19.attn_norm.weight
    F32
    [4096]
  • blk.19.attn_output.weight
    Q4_K
    [4096, 4096]
  • blk.19.attn_q.weight
    Q4_K
    [4096, 8192]
  • blk.19.attn_q_norm.weight
    F32
    [256]
  • blk.19.attn_v.weight
    Q4_K
    [4096, 1024]
  • blk.19.ffn_down.weight
    IQ4_XS
    [12288, 4096]
  • blk.19.ffn_gate.weight
    IQ3_XXS
    [4096, 12288]
  • blk.19.ffn_up.weight
    IQ3_XXS
    [4096, 12288]
  • blk.19.post_attention_norm.weight
    F32
    [4096]
  • blk.20
  • blk.20.attn_gate.weight
    Q4_K
    [4096, 4096]
  • blk.20.attn_norm.weight
    F32
    [4096]
  • blk.20.attn_qkv.weight
    Q4_K
    [4096, 8192]
  • blk.20.ffn_down.weight
    IQ4_XS
    [12288, 4096]
  • blk.20.ffn_gate.weight
    IQ3_XXS
    [4096, 12288]
  • blk.20.ffn_up.weight
    IQ3_XXS
    [4096, 12288]
  • blk.20.post_attention_norm.weight
    F32
    [4096]
  • blk.20.ssm_a
    F32
    [32]
  • blk.20.ssm_alpha.weight
    Q4_K
    [4096, 32]
  • blk.20.ssm_beta.weight
    Q4_K
    [4096, 32]
  • blk.20.ssm_conv1d.weight
    F32
    [4, 8192]
  • blk.20.ssm_dt.bias
    F32
    [32]
  • blk.20.ssm_norm.weight
    F32
    [128]
  • blk.20.ssm_out.weight
    Q4_K
    [4096, 4096]
  • blk.21
  • blk.21.attn_gate.weight
    Q4_K
    [4096, 4096]
  • blk.21.attn_norm.weight
    F32
    [4096]
  • blk.21.attn_qkv.weight
    Q6_K
    [4096, 8192]
  • blk.21.ffn_down.weight
    IQ4_XS
    [12288, 4096]
  • blk.21.ffn_gate.weight
    IQ3_XXS
    [4096, 12288]
  • blk.21.ffn_up.weight
    IQ3_XXS
    [4096, 12288]
  • blk.21.post_attention_norm.weight
    F32
    [4096]
  • blk.21.ssm_a
    F32
    [32]
  • blk.21.ssm_alpha.weight
    Q4_K
    [4096, 32]
  • blk.21.ssm_beta.weight
    Q4_K
    [4096, 32]
  • blk.21.ssm_conv1d.weight
    F32
    [4, 8192]
  • blk.21.ssm_dt.bias
    F32
    [32]
  • blk.21.ssm_norm.weight
    F32
    [128]
  • blk.21.ssm_out.weight
    Q4_K
    [4096, 4096]
  • blk.22
  • blk.22.attn_gate.weight
    Q4_K
    [4096, 4096]
  • blk.22.attn_norm.weight
    F32
    [4096]
  • blk.22.attn_qkv.weight
    Q4_K
    [4096, 8192]
  • blk.22.ffn_down.weight
    IQ4_XS
    [12288, 4096]
  • blk.22.ffn_gate.weight
    IQ3_XXS
    [4096, 12288]
  • blk.22.ffn_up.weight
    IQ3_XXS
    [4096, 12288]
  • blk.22.post_attention_norm.weight
    F32
    [4096]
  • blk.22.ssm_a
    F32
    [32]
  • blk.22.ssm_alpha.weight
    Q4_K
    [4096, 32]
  • blk.22.ssm_beta.weight
    Q4_K
    [4096, 32]
  • blk.22.ssm_conv1d.weight
    F32
    [4, 8192]
  • blk.22.ssm_dt.bias
    F32
    [32]
  • blk.22.ssm_norm.weight
    F32
    [128]
  • blk.22.ssm_out.weight
    Q4_K
    [4096, 4096]
  • blk.23
  • blk.23.attn_k.weight
    Q4_K
    [4096, 1024]
  • blk.23.attn_k_norm.weight
    F32
    [256]
  • blk.23.attn_norm.weight
    F32
    [4096]
  • blk.23.attn_output.weight
    Q4_K
    [4096, 4096]
  • blk.23.attn_q.weight
    Q4_K
    [4096, 8192]
  • blk.23.attn_q_norm.weight
    F32
    [256]
  • blk.23.attn_v.weight
    Q4_K
    [4096, 1024]
  • blk.23.ffn_down.weight
    IQ4_XS
    [12288, 4096]
  • blk.23.ffn_gate.weight
    IQ3_XXS
    [4096, 12288]
  • blk.23.ffn_up.weight
    IQ3_XXS
    [4096, 12288]
  • blk.23.post_attention_norm.weight
    F32
    [4096]
  • blk.24
  • blk.24.attn_gate.weight
    Q4_K
    [4096, 4096]
  • blk.24.attn_norm.weight
    F32
    [4096]
  • blk.24.attn_qkv.weight
    Q6_K
    [4096, 8192]
  • blk.24.ffn_down.weight
    IQ4_XS
    [12288, 4096]
  • blk.24.ffn_gate.weight
    IQ3_XXS
    [4096, 12288]
  • blk.24.ffn_up.weight
    IQ3_XXS
    [4096, 12288]
  • blk.24.post_attention_norm.weight
    F32
    [4096]
  • blk.24.ssm_a
    F32
    [32]
  • blk.24.ssm_alpha.weight
    Q4_K
    [4096, 32]
  • blk.24.ssm_beta.weight
    Q4_K
    [4096, 32]
  • blk.24.ssm_conv1d.weight
    F32
    [4, 8192]
  • blk.24.ssm_dt.bias
    F32
    [32]
  • blk.24.ssm_norm.weight
    F32
    [128]
  • blk.24.ssm_out.weight
    Q4_K
    [4096, 4096]
  • blk.25
  • blk.25.attn_gate.weight
    Q4_K
    [4096, 4096]
  • blk.25.attn_norm.weight
    F32
    [4096]
  • blk.25.attn_qkv.weight
    Q4_K
    [4096, 8192]
  • blk.25.ffn_down.weight
    IQ4_XS
    [12288, 4096]
  • blk.25.ffn_gate.weight
    IQ3_XXS
    [4096, 12288]
  • blk.25.ffn_up.weight
    IQ3_XXS
    [4096, 12288]
  • blk.25.post_attention_norm.weight
    F32
    [4096]
  • blk.25.ssm_a
    F32
    [32]
  • blk.25.ssm_alpha.weight
    Q4_K
    [4096, 32]
  • blk.25.ssm_beta.weight
    Q4_K
    [4096, 32]
  • blk.25.ssm_conv1d.weight
    F32
    [4, 8192]
  • blk.25.ssm_dt.bias
    F32
    [32]
  • blk.25.ssm_norm.weight
    F32
    [128]
  • blk.25.ssm_out.weight
    Q4_K
    [4096, 4096]
  • blk.26
  • blk.26.attn_gate.weight
    Q4_K
    [4096, 4096]
  • blk.26.attn_norm.weight
    F32
    [4096]
  • blk.26.attn_qkv.weight
    Q6_K
    [4096, 8192]
  • blk.26.ffn_down.weight
    IQ4_XS
    [12288, 4096]
  • blk.26.ffn_gate.weight
    IQ3_XXS
    [4096, 12288]
  • blk.26.ffn_up.weight
    IQ3_XXS
    [4096, 12288]
  • blk.26.post_attention_norm.weight
    F32
    [4096]
  • blk.26.ssm_a
    F32
    [32]
  • blk.26.ssm_alpha.weight
    Q4_K
    [4096, 32]
  • blk.26.ssm_beta.weight
    Q4_K
    [4096, 32]
  • blk.26.ssm_conv1d.weight
    F32
    [4, 8192]
  • blk.26.ssm_dt.bias
    F32
    [32]
  • blk.26.ssm_norm.weight
    F32
    [128]
  • blk.26.ssm_out.weight
    Q4_K
    [4096, 4096]
  • blk.27
  • blk.27.attn_k.weight
    Q4_K
    [4096, 1024]
  • blk.27.attn_k_norm.weight
    F32
    [256]
  • blk.27.attn_norm.weight
    F32
    [4096]
  • blk.27.attn_output.weight
    Q4_K
    [4096, 4096]
  • blk.27.attn_q.weight
    Q4_K
    [4096, 8192]
  • blk.27.attn_q_norm.weight
    F32
    [256]
  • blk.27.attn_v.weight
    Q6_K
    [4096, 1024]
  • blk.27.ffn_down.weight
    IQ4_XS
    [12288, 4096]
  • blk.27.ffn_gate.weight
    IQ3_XXS
    [4096, 12288]
  • blk.27.ffn_up.weight
    IQ3_XXS
    [4096, 12288]
  • blk.27.post_attention_norm.weight
    F32
    [4096]
  • blk.28
  • blk.28.attn_gate.weight
    Q4_K
    [4096, 4096]
  • blk.28.attn_norm.weight
    F32
    [4096]
  • blk.28.attn_qkv.weight
    Q6_K
    [4096, 8192]
  • blk.28.ffn_down.weight
    IQ4_XS
    [12288, 4096]
  • blk.28.ffn_gate.weight
    IQ3_XXS
    [4096, 12288]
  • blk.28.ffn_up.weight
    IQ3_XXS
    [4096, 12288]
  • blk.28.post_attention_norm.weight
    F32
    [4096]
  • blk.28.ssm_a
    F32
    [32]
  • blk.28.ssm_alpha.weight
    Q4_K
    [4096, 32]
  • blk.28.ssm_beta.weight
    Q4_K
    [4096, 32]
  • blk.28.ssm_conv1d.weight
    F32
    [4, 8192]
  • blk.28.ssm_dt.bias
    F32
    [32]
  • blk.28.ssm_norm.weight
    F32
    [128]
  • blk.28.ssm_out.weight
    Q4_K
    [4096, 4096]
  • blk.29
  • blk.29.attn_gate.weight
    Q4_K
    [4096, 4096]
  • blk.29.attn_norm.weight
    F32
    [4096]
  • blk.29.attn_qkv.weight
    Q6_K
    [4096, 8192]
  • blk.29.ffn_down.weight
    IQ4_XS
    [12288, 4096]
  • blk.29.ffn_gate.weight
    IQ3_XXS
    [4096, 12288]
  • blk.29.ffn_up.weight
    IQ3_XXS
    [4096, 12288]
  • blk.29.post_attention_norm.weight
    F32
    [4096]
  • blk.29.ssm_a
    F32
    [32]
  • blk.29.ssm_alpha.weight
    Q4_K
    [4096, 32]
  • blk.29.ssm_beta.weight
    Q4_K
    [4096, 32]
  • blk.29.ssm_conv1d.weight
    F32
    [4, 8192]
  • blk.29.ssm_dt.bias
    F32
    [32]
  • blk.29.ssm_norm.weight
    F32
    [128]
  • blk.29.ssm_out.weight
    Q4_K
    [4096, 4096]
  • blk.30
  • blk.30.attn_gate.weight
    Q4_K
    [4096, 4096]
  • blk.30.attn_norm.weight
    F32
    [4096]
  • blk.30.attn_qkv.weight
    Q6_K
    [4096, 8192]
  • blk.30.ffn_down.weight
    IQ4_XS
    [12288, 4096]
  • blk.30.ffn_gate.weight
    IQ3_XXS
    [4096, 12288]
  • blk.30.ffn_up.weight
    IQ3_XXS
    [4096, 12288]
  • blk.30.post_attention_norm.weight
    F32
    [4096]
  • blk.30.ssm_a
    F32
    [32]
  • blk.30.ssm_alpha.weight
    Q4_K
    [4096, 32]
  • blk.30.ssm_beta.weight
    Q4_K
    [4096, 32]
  • blk.30.ssm_conv1d.weight
    F32
    [4, 8192]
  • blk.30.ssm_dt.bias
    F32
    [32]
  • blk.30.ssm_norm.weight
    F32
    [128]
  • blk.30.ssm_out.weight
    Q4_K
    [4096, 4096]
  • blk.31
  • blk.31.attn_k.weight
    Q4_K
    [4096, 1024]
  • blk.31.attn_k_norm.weight
    F32
    [256]
  • blk.31.attn_norm.weight
    F32
    [4096]
  • blk.31.attn_output.weight
    Q4_K
    [4096, 4096]
  • blk.31.attn_q.weight
    Q4_K
    [4096, 8192]
  • blk.31.attn_q_norm.weight
    F32
    [256]
  • blk.31.attn_v.weight
    Q6_K
    [4096, 1024]
  • blk.31.ffn_down.weight
    IQ4_XS
    [12288, 4096]
  • blk.31.ffn_gate.weight
    IQ3_XXS
    [4096, 12288]
  • blk.31.ffn_up.weight
    IQ3_XXS
    [4096, 12288]
  • blk.31.post_attention_norm.weight
    F32
    [4096]
  • dec.blk.0
  • dec.blk.0.cross_attn_k.bias
    F32
    [1024]
  • dec.blk.0.cross_attn_k.weight
    Q4_K
    [1024, 1024]
  • dec.blk.0.cross_attn_norm.bias
    F32
    [1024]
  • dec.blk.0.cross_attn_norm.weight
    F32
    [1024]
  • dec.blk.0.cross_attn_norm_kv.bias
    F32
    [1024]
  • dec.blk.0.cross_attn_norm_kv.weight
    F32
    [1024]
  • dec.blk.0.cross_attn_o.bias
    F32
    [1024]
  • dec.blk.0.cross_attn_o.weight
    Q4_K
    [1024, 1024]
  • dec.blk.0.cross_attn_q.bias
    F32
    [1024]
  • dec.blk.0.cross_attn_q.weight
    Q4_K
    [1024, 1024]
  • dec.blk.0.cross_attn_v.bias
    F32
    [1024]
  • dec.blk.0.cross_attn_v.weight
    Q6_K
    [1024, 1024]
  • dec.blk.0.ffn_down.bias
    F32
    [1024]
  • dec.blk.0.ffn_down.weight
    Q6_K
    [4096, 1024]
  • dec.blk.0.ffn_norm.bias
    F32
    [1024]
  • dec.blk.0.ffn_norm.weight
    F32
    [1024]
  • dec.blk.0.ffn_up.bias
    F32
    [4096]
  • dec.blk.0.ffn_up.weight
    Q4_K
    [1024, 4096]
  • dec.blk.1
  • dec.blk.1.cross_attn_k.bias
    F32
    [1024]
  • dec.blk.1.cross_attn_k.weight
    Q4_K
    [1024, 1024]
  • dec.blk.1.cross_attn_norm.bias
    F32
    [1024]
  • dec.blk.1.cross_attn_norm.weight
    F32
    [1024]
  • dec.blk.1.cross_attn_norm_kv.bias
    F32
    [1024]
  • dec.blk.1.cross_attn_norm_kv.weight
    F32
    [1024]
  • dec.blk.1.cross_attn_o.bias
    F32
    [1024]
  • dec.blk.1.cross_attn_o.weight
    Q4_K
    [1024, 1024]
  • dec.blk.1.cross_attn_q.bias
    F32
    [1024]
  • dec.blk.1.cross_attn_q.weight
    Q4_K
    [1024, 1024]
  • dec.blk.1.cross_attn_v.bias
    F32
    [1024]
  • dec.blk.1.cross_attn_v.weight
    Q6_K
    [1024, 1024]
  • dec.blk.1.ffn_down.bias
    F32
    [1024]
  • dec.blk.1.ffn_down.weight
    Q6_K
    [4096, 1024]
  • dec.blk.1.ffn_norm.bias
    F32
    [1024]
  • dec.blk.1.ffn_norm.weight
    F32
    [1024]
  • dec.blk.1.ffn_up.bias
    F32
    [4096]
  • dec.blk.1.ffn_up.weight
    Q4_K
    [1024, 4096]
  • dec.blk.2
  • dec.blk.2.attn_k.bias
    F32
    [1024]
  • dec.blk.2.attn_k.weight
    Q4_K
    [1024, 1024]
  • dec.blk.2.attn_norm.bias
    F32
    [1024]
  • dec.blk.2.attn_norm.weight
    F32
    [1024]
  • dec.blk.2.attn_o.bias
    F32
    [1024]
  • dec.blk.2.attn_o.weight
    Q4_K
    [1024, 1024]
  • dec.blk.2.attn_q.bias
    F32
    [1024]
  • dec.blk.2.attn_q.weight
    Q4_K
    [1024, 1024]
  • dec.blk.2.attn_v.bias
    F32
    [1024]
  • dec.blk.2.attn_v.weight
    Q6_K
    [1024, 1024]
  • dec.blk.2.cross_attn_k.bias
    F32
    [1024]
  • dec.blk.2.cross_attn_k.weight
    Q4_K
    [1024, 1024]
  • dec.blk.2.cross_attn_norm.bias
    F32
    [1024]
  • dec.blk.2.cross_attn_norm.weight
    F32
    [1024]
  • dec.blk.2.cross_attn_o.bias
    F32
    [1024]
  • dec.blk.2.cross_attn_o.weight
    Q4_K
    [1024, 1024]
  • dec.blk.2.cross_attn_q.bias
    F32
    [1024]
  • dec.blk.2.cross_attn_q.weight
    Q4_K
    [1024, 1024]
  • dec.blk.2.cross_attn_v.bias
    F32
    [1024]
  • dec.blk.2.cross_attn_v.weight
    Q6_K
    [1024, 1024]
  • dec.blk.2.ffn_down.bias
    F32
    [1024]
  • dec.blk.2.ffn_down.weight
    Q6_K
    [4096, 1024]
  • dec.blk.2.ffn_norm.bias
    F32
    [1024]
  • dec.blk.2.ffn_norm.weight
    F32
    [1024]
  • dec.blk.2.ffn_up.bias
    F32
    [4096]
  • dec.blk.2.ffn_up.weight
    Q4_K
    [1024, 4096]
  • dec.blk.3
  • dec.blk.3.attn_k.bias
    F32
    [1024]
  • dec.blk.3.attn_k.weight
    Q4_K
    [1024, 1024]
  • dec.blk.3.attn_norm.bias
    F32
    [1024]
  • dec.blk.3.attn_norm.weight
    F32
    [1024]
  • dec.blk.3.attn_o.bias
    F32
    [1024]
  • dec.blk.3.attn_o.weight
    Q4_K
    [1024, 1024]
  • dec.blk.3.attn_q.bias
    F32
    [1024]
  • dec.blk.3.attn_q.weight
    Q4_K
    [1024, 1024]
  • dec.blk.3.attn_v.bias
    F32
    [1024]
  • dec.blk.3.attn_v.weight
    Q6_K
    [1024, 1024]
  • dec.blk.3.cross_attn_k.bias
    F32
    [1024]
  • dec.blk.3.cross_attn_k.weight
    Q4_K
    [1024, 1024]
  • dec.blk.3.cross_attn_norm.bias
    F32
    [1024]
  • dec.blk.3.cross_attn_norm.weight
    F32
    [1024]
  • dec.blk.3.cross_attn_o.bias
    F32
    [1024]
  • dec.blk.3.cross_attn_o.weight
    Q4_K
    [1024, 1024]
  • dec.blk.3.cross_attn_q.bias
    F32
    [1024]
  • dec.blk.3.cross_attn_q.weight
    Q4_K
    [1024, 1024]
  • dec.blk.3.cross_attn_v.bias
    F32
    [1024]
  • dec.blk.3.cross_attn_v.weight
    Q4_K
    [1024, 1024]
  • dec.blk.3.ffn_down.bias
    F32
    [1024]
  • dec.blk.3.ffn_down.weight
    Q6_K
    [4096, 1024]
  • dec.blk.3.ffn_norm.bias
    F32
    [1024]
  • dec.blk.3.ffn_norm.weight
    F32
    [1024]
  • dec.blk.3.ffn_up.bias
    F32
    [4096]
  • dec.blk.3.ffn_up.weight
    Q4_K
    [1024, 4096]
  • dec.blk.4
  • dec.blk.4.attn_k.bias
    F32
    [1024]
  • dec.blk.4.attn_k.weight
    Q4_K
    [1024, 1024]
  • dec.blk.4.attn_norm.bias
    F32
    [1024]
  • dec.blk.4.attn_norm.weight
    F32
    [1024]
  • dec.blk.4.attn_o.bias
    F32
    [1024]
  • dec.blk.4.attn_o.weight
    Q4_K
    [1024, 1024]
  • dec.blk.4.attn_q.bias
    F32
    [1024]
  • dec.blk.4.attn_q.weight
    Q4_K
    [1024, 1024]
  • dec.blk.4.attn_v.bias
    F32
    [1024]
  • dec.blk.4.attn_v.weight
    Q4_K
    [1024, 1024]
  • dec.blk.4.cross_attn_k.bias
    F32
    [1024]
  • dec.blk.4.cross_attn_k.weight
    Q4_K
    [1024, 1024]
  • dec.blk.4.cross_attn_norm.bias
    F32
    [1024]
  • dec.blk.4.cross_attn_norm.weight
    F32
    [1024]
  • dec.blk.4.cross_attn_o.bias
    F32
    [1024]
  • dec.blk.4.cross_attn_o.weight
    Q4_K
    [1024, 1024]
  • dec.blk.4.cross_attn_q.bias
    F32
    [1024]
  • dec.blk.4.cross_attn_q.weight
    Q4_K
    [1024, 1024]
  • dec.blk.4.cross_attn_v.bias
    F32
    [1024]
  • dec.blk.4.cross_attn_v.weight
    Q6_K
    [1024, 1024]
  • dec.blk.4.ffn_down.bias
    F32
    [1024]
  • dec.blk.4.ffn_down.weight
    Q4_K
    [4096, 1024]
  • dec.blk.4.ffn_norm.bias
    F32
    [1024]
  • dec.blk.4.ffn_norm.weight
    F32
    [1024]
  • dec.blk.4.ffn_up.bias
    F32
    [4096]
  • dec.blk.4.ffn_up.weight
    Q4_K
    [1024, 4096]
  • dec.blk.5
  • dec.blk.5.attn_k.bias
    F32
    [1024]
  • dec.blk.5.attn_k.weight
    Q4_K
    [1024, 1024]
  • dec.blk.5.attn_norm.bias
    F32
    [1024]
  • dec.blk.5.attn_norm.weight
    F32
    [1024]
  • dec.blk.5.attn_o.bias
    F32
    [1024]
  • dec.blk.5.attn_o.weight
    Q4_K
    [1024, 1024]
  • dec.blk.5.attn_q.bias
    F32
    [1024]
  • dec.blk.5.attn_q.weight
    Q4_K
    [1024, 1024]
  • dec.blk.5.attn_v.bias
    F32
    [1024]
  • dec.blk.5.attn_v.weight
    Q4_K
    [1024, 1024]
  • dec.blk.5.cross_attn_k.bias
    F32
    [1024]
  • dec.blk.5.cross_attn_k.weight
    Q4_K
    [1024, 1024]
  • dec.blk.5.cross_attn_norm.bias
    F32
    [1024]
  • dec.blk.5.cross_attn_norm.weight
    F32
    [1024]
  • dec.blk.5.cross_attn_o.bias
    F32
    [1024]
  • dec.blk.5.cross_attn_o.weight
    Q4_K
    [1024, 1024]
  • dec.blk.5.cross_attn_q.bias
    F32
    [1024]
  • dec.blk.5.cross_attn_q.weight
    Q4_K
    [1024, 1024]
  • dec.blk.5.cross_attn_v.bias
    F32
    [1024]
  • dec.blk.5.cross_attn_v.weight
    Q4_K
    [1024, 1024]
  • dec.blk.5.ffn_down.bias
    F32
    [1024]
  • dec.blk.5.ffn_down.weight
    Q4_K
    [4096, 1024]
  • dec.blk.5.ffn_norm.bias
    F32
    [1024]
  • dec.blk.5.ffn_norm.weight
    F32
    [1024]
  • dec.blk.5.ffn_up.bias
    F32
    [4096]
  • dec.blk.5.ffn_up.weight
    Q4_K
    [1024, 4096]
  • decision.field_norm.bias
    F32
    [1024]
  • decision.field_norm.weight
    F32
    [1024]
  • decision.hidden_norm.bias
    F32
    [4096]
  • decision.hidden_norm.weight
    F32
    [4096]
  • decision.option_norm.bias
    F32
    [1024]
  • decision.option_norm.weight
    F32
    [1024]
  • decision.option_summary_norm.bias
    F32
    [1024]
  • decision.option_summary_norm.weight
    F32
    [1024]
  • decision.proj_global.weight
    Q4_K
    [4096, 1024]
  • decision.proj_memory.weight
    Q4_K
    [4096, 1024]
  • decision.proj_option_context.weight
    Q4_K
    [4096, 1024]
  • decision.proj_option_lexical.weight
    Q4_K
    [4096, 1024]
  • decision.proj_option_question.weight
    Q4_K
    [4096, 1024]
  • decision.proj_question.weight
    Q4_K
    [4096, 1024]
  • decision.scales
    F32
    [3]
  • decision.scorer.bias
    F32
    [1024]
  • decision.scorer.weight
    Q4_K
    [4096, 1024]
  • decision.scorer_out.bias
    F32
    [1]
  • decision.scorer_out.weight
    BF16
    [1024]
  • output.weight
    IQ2_XS
    [4096, 248320]
  • token_types.weight
    F32
    [1024, 3]
  • output_norm.weight
    F32
    [4096]