222.5K 4 hours ago

Nemotron-3-Nano is a new Standard for Efficient, Open, and Intelligent Agentic Models, now updated with a 4B parameter count model.

tools thinking cloud 4b 30b
527db2cf6c70 · 2.8GB
    Metadata
  • general.architecture
    nemotron_h
  • general.file_type
    Q4_K_M
  • nemotron_h.attention.head_count
    40
  • nemotron_h.attention.head_count_kv
    [0, 0, 0, 0, 0, ...]
  • nemotron_h.attention.key_length
    128
  • nemotron_h.attention.layer_norm_epsilon
    1e-05
  • nemotron_h.attention.layer_norm_rms_epsilon
    1e-05
  • nemotron_h.attention.value_length
    128
  • nemotron_h.block_count
    42
  • nemotron_h.context_length
    262144
  • nemotron_h.embedding_length
    3136
  • nemotron_h.feed_forward_length
    [0, 12544, 0, 12544, 0, ...]
  • nemotron_h.rope.freq_base
    10000
  • nemotron_h.ssm.conv_kernel
    4
  • nemotron_h.ssm.group_count
    8
  • nemotron_h.ssm.inner_size
    7680
  • nemotron_h.ssm.state_size
    128
  • nemotron_h.ssm.time_step_rank
    96
  • tokenizer.ggml.add_bos_token
    false
  • tokenizer.ggml.add_eos_token
    false
  • tokenizer.ggml.add_unknown_token
    false
  • tokenizer.ggml.bos_token_id
    1
  • tokenizer.ggml.eos_token_id
    11
  • tokenizer.ggml.merges
    [Ġ Ġ, Ġ t, e r, i n, Ġ ĠĠĠ, ...]
  • tokenizer.ggml.model
    gpt2
  • tokenizer.ggml.pre
    default
  • tokenizer.ggml.scores
    [0, 1, 2, 3, 4, ...]
  • tokenizer.ggml.token_type
    [3, 3, 3, 3, 3, ...]
  • tokenizer.ggml.tokens
    [<unk>, <s>, </s>, [INST], [/INST], ...]
  • tokenizer.ggml.unknown_token_id
    0
  • Tensor
  • token_embd.weight
    Q5_0
    [3136, 131072]
  • blk.0
  • blk.0.attn_norm.weight
    F32
    [3136]
  • blk.0.ssm_a
    F32
    [1, 96]
  • blk.0.ssm_conv1d.bias
    F32
    [9728]
  • blk.0.ssm_conv1d.weight
    F32
    [4, 9728]
  • blk.0.ssm_d
    F32
    [1, 96]
  • blk.0.ssm_dt.bias
    F32
    [96]
  • blk.0.ssm_in.weight
    Q5_0
    [3136, 17504]
  • blk.0.ssm_norm.weight
    F32
    [960, 8]
  • blk.0.ssm_out.weight
    Q4_K
    [7680, 3136]
  • blk.1
  • blk.1.attn_norm.weight
    F32
    [3136]
  • blk.1.ffn_down.weight
    Q6_K
    [12544, 3136]
  • blk.1.ffn_up.weight
    Q5_0
    [3136, 12544]
  • blk.2
  • blk.2.attn_norm.weight
    F32
    [3136]
  • blk.2.ssm_a
    F32
    [1, 96]
  • blk.2.ssm_conv1d.bias
    F32
    [9728]
  • blk.2.ssm_conv1d.weight
    F32
    [4, 9728]
  • blk.2.ssm_d
    F32
    [1, 96]
  • blk.2.ssm_dt.bias
    F32
    [96]
  • blk.2.ssm_in.weight
    Q5_0
    [3136, 17504]
  • blk.2.ssm_norm.weight
    F32
    [960, 8]
  • blk.2.ssm_out.weight
    Q4_K
    [7680, 3136]
  • blk.3
  • blk.3.attn_norm.weight
    F32
    [3136]
  • blk.3.ffn_down.weight
    Q6_K
    [12544, 3136]
  • blk.3.ffn_up.weight
    Q5_0
    [3136, 12544]
  • blk.4
  • blk.4.attn_norm.weight
    F32
    [3136]
  • blk.4.ssm_a
    F32
    [1, 96]
  • blk.4.ssm_conv1d.bias
    F32
    [9728]
  • blk.4.ssm_conv1d.weight
    F32
    [4, 9728]
  • blk.4.ssm_d
    F32
    [1, 96]
  • blk.4.ssm_dt.bias
    F32
    [96]
  • blk.4.ssm_in.weight
    Q5_0
    [3136, 17504]
  • blk.4.ssm_norm.weight
    F32
    [960, 8]
  • blk.4.ssm_out.weight
    Q4_K
    [7680, 3136]
  • blk.5
  • blk.5.attn_norm.weight
    F32
    [3136]
  • blk.5.ffn_down.weight
    Q6_K
    [12544, 3136]
  • blk.5.ffn_up.weight
    Q5_0
    [3136, 12544]
  • blk.6
  • blk.6.attn_norm.weight
    F32
    [3136]
  • blk.6.ssm_a
    F32
    [1, 96]
  • blk.6.ssm_conv1d.bias
    F32
    [9728]
  • blk.6.ssm_conv1d.weight
    F32
    [4, 9728]
  • blk.6.ssm_d
    F32
    [1, 96]
  • blk.6.ssm_dt.bias
    F32
    [96]
  • blk.6.ssm_in.weight
    Q5_0
    [3136, 17504]
  • blk.6.ssm_norm.weight
    F32
    [960, 8]
  • blk.6.ssm_out.weight
    Q4_K
    [7680, 3136]
  • blk.7
  • blk.7.attn_norm.weight
    F32
    [3136]
  • blk.7.ssm_a
    F32
    [1, 96]
  • blk.7.ssm_conv1d.bias
    F32
    [9728]
  • blk.7.ssm_conv1d.weight
    F32
    [4, 9728]
  • blk.7.ssm_d
    F32
    [1, 96]
  • blk.7.ssm_dt.bias
    F32
    [96]
  • blk.7.ssm_in.weight
    Q5_0
    [3136, 17504]
  • blk.7.ssm_norm.weight
    F32
    [960, 8]
  • blk.7.ssm_out.weight
    Q4_K
    [7680, 3136]
  • blk.8
  • blk.8.attn_norm.weight
    F32
    [3136]
  • blk.8.ffn_down.weight
    Q6_K
    [12544, 3136]
  • blk.8.ffn_up.weight
    Q5_0
    [3136, 12544]
  • blk.9
  • blk.9.attn_norm.weight
    F32
    [3136]
  • blk.9.ssm_a
    F32
    [1, 96]
  • blk.9.ssm_conv1d.bias
    F32
    [9728]
  • blk.9.ssm_conv1d.weight
    F32
    [4, 9728]
  • blk.9.ssm_d
    F32
    [1, 96]
  • blk.9.ssm_dt.bias
    F32
    [96]
  • blk.9.ssm_in.weight
    Q5_0
    [3136, 17504]
  • blk.9.ssm_norm.weight
    F32
    [960, 8]
  • blk.9.ssm_out.weight
    Q4_K
    [7680, 3136]
  • blk.10
  • blk.10.attn_norm.weight
    F32
    [3136]
  • blk.10.ffn_down.weight
    Q6_K
    [12544, 3136]
  • blk.10.ffn_up.weight
    Q5_0
    [3136, 12544]
  • blk.11
  • blk.11.attn_norm.weight
    F32
    [3136]
  • blk.11.ssm_a
    F32
    [1, 96]
  • blk.11.ssm_conv1d.bias
    F32
    [9728]
  • blk.11.ssm_conv1d.weight
    F32
    [4, 9728]
  • blk.11.ssm_d
    F32
    [1, 96]
  • blk.11.ssm_dt.bias
    F32
    [96]
  • blk.11.ssm_in.weight
    Q5_0
    [3136, 17504]
  • blk.11.ssm_norm.weight
    F32
    [960, 8]
  • blk.11.ssm_out.weight
    Q4_K
    [7680, 3136]
  • blk.12
  • blk.12.attn_k.weight
    Q5_0
    [3136, 1024]
  • blk.12.attn_norm.weight
    F32
    [3136]
  • blk.12.attn_output.weight
    Q4_K
    [5120, 3136]
  • blk.12.attn_q.weight
    Q5_0
    [3136, 5120]
  • blk.12.attn_v.weight
    Q5_0
    [3136, 1024]
  • blk.13
  • blk.13.attn_norm.weight
    F32
    [3136]
  • blk.13.ffn_down.weight
    Q4_K
    [12544, 3136]
  • blk.13.ffn_up.weight
    Q5_0
    [3136, 12544]
  • blk.14
  • blk.14.attn_norm.weight
    F32
    [3136]
  • blk.14.ssm_a
    F32
    [1, 96]
  • blk.14.ssm_conv1d.bias
    F32
    [9728]
  • blk.14.ssm_conv1d.weight
    F32
    [4, 9728]
  • blk.14.ssm_d
    F32
    [1, 96]
  • blk.14.ssm_dt.bias
    F32
    [96]
  • blk.14.ssm_in.weight
    Q5_0
    [3136, 17504]
  • blk.14.ssm_norm.weight
    F32
    [960, 8]
  • blk.14.ssm_out.weight
    Q4_K
    [7680, 3136]
  • blk.15
  • blk.15.attn_norm.weight
    F32
    [3136]
  • blk.15.ffn_down.weight
    Q4_K
    [12544, 3136]
  • blk.15.ffn_up.weight
    Q5_0
    [3136, 12544]
  • blk.16
  • blk.16.attn_norm.weight
    F32
    [3136]
  • blk.16.ssm_a
    F32
    [1, 96]
  • blk.16.ssm_conv1d.bias
    F32
    [9728]
  • blk.16.ssm_conv1d.weight
    F32
    [4, 9728]
  • blk.16.ssm_d
    F32
    [1, 96]
  • blk.16.ssm_dt.bias
    F32
    [96]
  • blk.16.ssm_in.weight
    Q5_0
    [3136, 17504]
  • blk.16.ssm_norm.weight
    F32
    [960, 8]
  • blk.16.ssm_out.weight
    Q4_K
    [7680, 3136]
  • blk.17
  • blk.17.attn_k.weight
    Q5_0
    [3136, 1024]
  • blk.17.attn_norm.weight
    F32
    [3136]
  • blk.17.attn_output.weight
    Q4_K
    [5120, 3136]
  • blk.17.attn_q.weight
    Q5_0
    [3136, 5120]
  • blk.17.attn_v.weight
    Q5_0
    [3136, 1024]
  • blk.18
  • blk.18.attn_norm.weight
    F32
    [3136]
  • blk.18.ffn_down.weight
    Q6_K
    [12544, 3136]
  • blk.18.ffn_up.weight
    Q5_0
    [3136, 12544]
  • blk.19
  • blk.19.attn_norm.weight
    F32
    [3136]
  • blk.19.ssm_a
    F32
    [1, 96]
  • blk.19.ssm_conv1d.bias
    F32
    [9728]
  • blk.19.ssm_conv1d.weight
    F32
    [4, 9728]
  • blk.19.ssm_d
    F32
    [1, 96]
  • blk.19.ssm_dt.bias
    F32
    [96]
  • blk.19.ssm_in.weight
    Q5_0
    [3136, 17504]
  • blk.19.ssm_norm.weight
    F32
    [960, 8]
  • blk.19.ssm_out.weight
    Q4_K
    [7680, 3136]
  • blk.20
  • blk.20.attn_norm.weight
    F32
    [3136]
  • blk.20.ffn_down.weight
    Q4_K
    [12544, 3136]
  • blk.20.ffn_up.weight
    Q5_0
    [3136, 12544]
  • blk.21
  • blk.21.attn_norm.weight
    F32
    [3136]
  • blk.21.ssm_a
    F32
    [1, 96]
  • blk.21.ssm_conv1d.bias
    F32
    [9728]
  • blk.21.ssm_conv1d.weight
    F32
    [4, 9728]
  • blk.21.ssm_d
    F32
    [1, 96]
  • blk.21.ssm_dt.bias
    F32
    [96]
  • blk.21.ssm_in.weight
    Q5_0
    [3136, 17504]
  • blk.21.ssm_norm.weight
    F32
    [960, 8]
  • blk.21.ssm_out.weight
    Q4_K
    [7680, 3136]
  • blk.22
  • blk.22.attn_norm.weight
    F32
    [3136]
  • blk.22.ffn_down.weight
    Q4_K
    [12544, 3136]
  • blk.22.ffn_up.weight
    Q5_0
    [3136, 12544]
  • blk.23
  • blk.23.attn_norm.weight
    F32
    [3136]
  • blk.23.ssm_a
    F32
    [1, 96]
  • blk.23.ssm_conv1d.bias
    F32
    [9728]
  • blk.23.ssm_conv1d.weight
    F32
    [4, 9728]
  • blk.23.ssm_d
    F32
    [1, 96]
  • blk.23.ssm_dt.bias
    F32
    [96]
  • blk.23.ssm_in.weight
    Q5_0
    [3136, 17504]
  • blk.23.ssm_norm.weight
    F32
    [960, 8]
  • blk.23.ssm_out.weight
    Q4_K
    [7680, 3136]
  • blk.24
  • blk.24.attn_k.weight
    Q5_0
    [3136, 1024]
  • blk.24.attn_norm.weight
    F32
    [3136]
  • blk.24.attn_output.weight
    Q4_K
    [5120, 3136]
  • blk.24.attn_q.weight
    Q5_0
    [3136, 5120]
  • blk.24.attn_v.weight
    Q8_0
    [3136, 1024]
  • blk.25
  • blk.25.attn_norm.weight
    F32
    [3136]
  • blk.25.ffn_down.weight
    Q6_K
    [12544, 3136]
  • blk.25.ffn_up.weight
    Q5_0
    [3136, 12544]
  • blk.26
  • blk.26.attn_norm.weight
    F32
    [3136]
  • blk.26.ssm_a
    F32
    [1, 96]
  • blk.26.ssm_conv1d.bias
    F32
    [9728]
  • blk.26.ssm_conv1d.weight
    F32
    [4, 9728]
  • blk.26.ssm_d
    F32
    [1, 96]
  • blk.26.ssm_dt.bias
    F32
    [96]
  • blk.26.ssm_in.weight
    Q5_0
    [3136, 17504]
  • blk.26.ssm_norm.weight
    F32
    [960, 8]
  • blk.26.ssm_out.weight
    Q4_K
    [7680, 3136]
  • blk.27
  • blk.27.attn_norm.weight
    F32
    [3136]
  • blk.27.ffn_down.weight
    Q4_K
    [12544, 3136]
  • blk.27.ffn_up.weight
    Q5_0
    [3136, 12544]
  • blk.28
  • blk.28.attn_norm.weight
    F32
    [3136]
  • blk.28.ssm_a
    F32
    [1, 96]
  • blk.28.ssm_conv1d.bias
    F32
    [9728]
  • blk.28.ssm_conv1d.weight
    F32
    [4, 9728]
  • blk.28.ssm_d
    F32
    [1, 96]
  • blk.28.ssm_dt.bias
    F32
    [96]
  • blk.28.ssm_in.weight
    Q5_0
    [3136, 17504]
  • blk.28.ssm_norm.weight
    F32
    [960, 8]
  • blk.28.ssm_out.weight
    Q4_K
    [7680, 3136]
  • blk.29
  • blk.29.attn_norm.weight
    F32
    [3136]
  • blk.29.ffn_down.weight
    Q4_K
    [12544, 3136]
  • blk.29.ffn_up.weight
    Q5_0
    [3136, 12544]
  • blk.30
  • blk.30.attn_norm.weight
    F32
    [3136]
  • blk.30.ssm_a
    F32
    [1, 96]
  • blk.30.ssm_conv1d.bias
    F32
    [9728]
  • blk.30.ssm_conv1d.weight
    F32
    [4, 9728]
  • blk.30.ssm_d
    F32
    [1, 96]
  • blk.30.ssm_dt.bias
    F32
    [96]
  • blk.30.ssm_in.weight
    Q5_0
    [3136, 17504]
  • blk.30.ssm_norm.weight
    F32
    [960, 8]
  • blk.30.ssm_out.weight
    Q4_K
    [7680, 3136]
  • blk.31
  • blk.31.attn_norm.weight
    F32
    [3136]
  • blk.31.ssm_a
    F32
    [1, 96]
  • blk.31.ssm_conv1d.bias
    F32
    [9728]
  • blk.31.ssm_conv1d.weight
    F32
    [4, 9728]
  • blk.31.ssm_d
    F32
    [1, 96]
  • blk.31.ssm_dt.bias
    F32
    [96]
  • blk.31.ssm_in.weight
    Q5_0
    [3136, 17504]
  • blk.31.ssm_norm.weight
    F32
    [960, 8]
  • blk.31.ssm_out.weight
    Q4_K
    [7680, 3136]
  • blk.32
  • blk.32.attn_k.weight
    Q5_0
    [3136, 1024]
  • blk.32.attn_norm.weight
    F32
    [3136]
  • blk.32.attn_output.weight
    Q4_K
    [5120, 3136]
  • blk.32.attn_q.weight
    Q5_0
    [3136, 5120]
  • blk.32.attn_v.weight
    Q8_0
    [3136, 1024]
  • blk.33
  • blk.33.attn_norm.weight
    F32
    [3136]
  • blk.33.ffn_down.weight
    Q6_K
    [12544, 3136]
  • blk.33.ffn_up.weight
    Q5_0
    [3136, 12544]
  • blk.34
  • blk.34.attn_norm.weight
    F32
    [3136]
  • blk.34.ssm_a
    F32
    [1, 96]
  • blk.34.ssm_conv1d.bias
    F32
    [9728]
  • blk.34.ssm_conv1d.weight
    F32
    [4, 9728]
  • blk.34.ssm_d
    F32
    [1, 96]
  • blk.34.ssm_dt.bias
    F32
    [96]
  • blk.34.ssm_in.weight
    Q5_0
    [3136, 17504]
  • blk.34.ssm_norm.weight
    F32
    [960, 8]
  • blk.34.ssm_out.weight
    Q4_K
    [7680, 3136]
  • blk.35
  • blk.35.attn_norm.weight
    F32
    [3136]
  • blk.35.ssm_a
    F32
    [1, 96]
  • blk.35.ssm_conv1d.bias
    F32
    [9728]
  • blk.35.ssm_conv1d.weight
    F32
    [4, 9728]
  • blk.35.ssm_d
    F32
    [1, 96]
  • blk.35.ssm_dt.bias
    F32
    [96]
  • blk.35.ssm_in.weight
    Q5_0
    [3136, 17504]
  • blk.35.ssm_norm.weight
    F32
    [960, 8]
  • blk.35.ssm_out.weight
    Q4_K
    [7680, 3136]
  • blk.36
  • blk.36.attn_norm.weight
    F32
    [3136]
  • blk.36.ssm_a
    F32
    [1, 96]
  • blk.36.ssm_conv1d.bias
    F32
    [9728]
  • blk.36.ssm_conv1d.weight
    F32
    [4, 9728]
  • blk.36.ssm_d
    F32
    [1, 96]
  • blk.36.ssm_dt.bias
    F32
    [96]
  • blk.36.ssm_in.weight
    Q5_0
    [3136, 17504]
  • blk.36.ssm_norm.weight
    F32
    [960, 8]
  • blk.36.ssm_out.weight
    Q4_K
    [7680, 3136]
  • blk.37
  • blk.37.attn_norm.weight
    F32
    [3136]
  • blk.37.ffn_down.weight
    Q4_K
    [12544, 3136]
  • blk.37.ffn_up.weight
    Q5_0
    [3136, 12544]
  • blk.38
  • blk.38.attn_norm.weight
    F32
    [3136]
  • blk.38.ssm_a
    F32
    [1, 96]
  • blk.38.ssm_conv1d.bias
    F32
    [9728]
  • blk.38.ssm_conv1d.weight
    F32
    [4, 9728]
  • blk.38.ssm_d
    F32
    [1, 96]
  • blk.38.ssm_dt.bias
    F32
    [96]
  • blk.38.ssm_in.weight
    Q5_0
    [3136, 17504]
  • blk.38.ssm_norm.weight
    F32
    [960, 8]
  • blk.38.ssm_out.weight
    Q4_K
    [7680, 3136]
  • blk.39
  • blk.39.attn_norm.weight
    F32
    [3136]
  • blk.39.ffn_down.weight
    Q4_K
    [12544, 3136]
  • blk.39.ffn_up.weight
    Q5_0
    [3136, 12544]
  • blk.40
  • blk.40.attn_norm.weight
    F32
    [3136]
  • blk.40.ssm_a
    F32
    [1, 96]
  • blk.40.ssm_conv1d.bias
    F32
    [9728]
  • blk.40.ssm_conv1d.weight
    F32
    [4, 9728]
  • blk.40.ssm_d
    F32
    [1, 96]
  • blk.40.ssm_dt.bias
    F32
    [96]
  • blk.40.ssm_in.weight
    Q5_0
    [3136, 17504]
  • blk.40.ssm_norm.weight
    F32
    [960, 8]
  • blk.40.ssm_out.weight
    Q4_K
    [7680, 3136]
  • blk.41
  • blk.41.attn_norm.weight
    F32
    [3136]
  • blk.41.ffn_down.weight
    Q6_K
    [12544, 3136]
  • blk.41.ffn_up.weight
    Q5_0
    [3136, 12544]
  • output.weight
    Q8_0
    [3136, 131072]
  • output_norm.weight
    F32
    [3136]