A commercial-friendly small language model by NVIDIA optimized for roleplay, RAG QA, and function calling.
tools
4b
39.8K Pulls Updated 2 months ago
a83fbc63c044 · 8.4GB
-
general.architecturenemotron
-
general.base_model.0.nameMinitron 4B Base
-
general.base_model.0.organizationNvidia
-
general.base_model.0.repo_urlhttps://huggingface.co/nvidia/Minitron-4B-Base
-
general.base_model.count1
-
general.basenameMinitron
-
general.file_type1
-
general.finetuneInstruct
-
general.languages[en]
-
general.licenseother
-
general.license.linkhttps://huggingface.co/nvidia/Nemotron-Mini-4B-Instruct/blob/main/nvidia-community-model-license-aug2024.pdf
-
general.license.namenvidia-community-model-license
-
general.nameMinitron 4B Instruct
-
general.organizationNvidia
-
general.quantization_version2
-
general.size_label4B
-
general.typemodel
-
nemotron.attention.head_count24
-
nemotron.attention.head_count_kv8
-
nemotron.attention.layer_norm_epsilon1e-05
-
nemotron.block_count32
-
nemotron.context_length4096
-
nemotron.embedding_length3072
-
nemotron.feed_forward_length9216
-
nemotron.rope.dimension_count64
-
nemotron.rope.freq_base10000
-
nemotron.rope.scaling.typenone
-
nemotron.vocab_size256000
-
tokenizer.ggml.bos_token_id2
-
tokenizer.ggml.eos_token_id3
-
tokenizer.ggml.modelllama
-
tokenizer.ggml.padding_token_id0
-
tokenizer.ggml.predefault
-
tokenizer.ggml.scores[-1000, 0, -1000, -1000, -1000, ...]
-
tokenizer.ggml.token_type[3, 2, 3, 3, 4, ...]
-
tokenizer.ggml.tokens[<pad>, <unk>, <s>, </s>, <extra_id_0>, ...]
-
tokenizer.ggml.unknown_token_id1
-
NameTypeShape
-
token_embd.weightF16[3072, 256000]
-
blk.0.attn_k.weightF16[3072, 1024]
-
blk.0.attn_norm.biasF32[3072]
-
blk.0.attn_norm.weightF32[3072]
-
blk.0.attn_output.weightF16[3072, 3072]
-
blk.0.attn_q.weightF16[3072, 3072]
-
blk.0.attn_v.weightF16[3072, 1024]
-
blk.0.ffn_down.weightF16[9216, 3072]
-
blk.0.ffn_norm.biasF32[3072]
-
blk.0.ffn_norm.weightF32[3072]
-
blk.0.ffn_up.weightF16[3072, 9216]
-
blk.1.attn_k.weightF16[3072, 1024]
-
blk.1.attn_norm.biasF32[3072]
-
blk.1.attn_norm.weightF32[3072]
-
blk.1.attn_output.weightF16[3072, 3072]
-
blk.1.attn_q.weightF16[3072, 3072]
-
blk.1.attn_v.weightF16[3072, 1024]
-
blk.1.ffn_down.weightF16[9216, 3072]
-
blk.1.ffn_norm.biasF32[3072]
-
blk.1.ffn_norm.weightF32[3072]
-
blk.1.ffn_up.weightF16[3072, 9216]
-
blk.2.attn_k.weightF16[3072, 1024]
-
blk.2.attn_norm.biasF32[3072]
-
blk.2.attn_norm.weightF32[3072]
-
blk.2.attn_output.weightF16[3072, 3072]
-
blk.2.attn_q.weightF16[3072, 3072]
-
blk.2.attn_v.weightF16[3072, 1024]
-
blk.2.ffn_down.weightF16[9216, 3072]
-
blk.2.ffn_norm.biasF32[3072]
-
blk.2.ffn_norm.weightF32[3072]
-
blk.2.ffn_up.weightF16[3072, 9216]
-
blk.3.attn_k.weightF16[3072, 1024]
-
blk.3.attn_norm.biasF32[3072]
-
blk.3.attn_norm.weightF32[3072]
-
blk.3.attn_output.weightF16[3072, 3072]
-
blk.3.attn_q.weightF16[3072, 3072]
-
blk.3.attn_v.weightF16[3072, 1024]
-
blk.3.ffn_down.weightF16[9216, 3072]
-
blk.3.ffn_norm.biasF32[3072]
-
blk.3.ffn_norm.weightF32[3072]
-
blk.3.ffn_up.weightF16[3072, 9216]
-
blk.4.attn_k.weightF16[3072, 1024]
-
blk.4.attn_norm.biasF32[3072]
-
blk.4.attn_norm.weightF32[3072]
-
blk.4.attn_output.weightF16[3072, 3072]
-
blk.4.attn_q.weightF16[3072, 3072]
-
blk.4.attn_v.weightF16[3072, 1024]
-
blk.4.ffn_down.weightF16[9216, 3072]
-
blk.4.ffn_norm.biasF32[3072]
-
blk.4.ffn_norm.weightF32[3072]
-
blk.4.ffn_up.weightF16[3072, 9216]
-
blk.5.attn_k.weightF16[3072, 1024]
-
blk.5.attn_norm.biasF32[3072]
-
blk.5.attn_norm.weightF32[3072]
-
blk.5.attn_output.weightF16[3072, 3072]
-
blk.5.attn_q.weightF16[3072, 3072]
-
blk.5.attn_v.weightF16[3072, 1024]
-
blk.5.ffn_down.weightF16[9216, 3072]
-
blk.5.ffn_norm.biasF32[3072]
-
blk.5.ffn_norm.weightF32[3072]
-
blk.5.ffn_up.weightF16[3072, 9216]
-
blk.6.attn_k.weightF16[3072, 1024]
-
blk.6.attn_norm.biasF32[3072]
-
blk.6.attn_norm.weightF32[3072]
-
blk.6.attn_output.weightF16[3072, 3072]
-
blk.6.attn_q.weightF16[3072, 3072]
-
blk.6.attn_v.weightF16[3072, 1024]
-
blk.6.ffn_down.weightF16[9216, 3072]
-
blk.6.ffn_norm.biasF32[3072]
-
blk.6.ffn_norm.weightF32[3072]
-
blk.6.ffn_up.weightF16[3072, 9216]
-
blk.7.attn_k.weightF16[3072, 1024]
-
blk.7.attn_norm.biasF32[3072]
-
blk.7.attn_norm.weightF32[3072]
-
blk.7.attn_output.weightF16[3072, 3072]
-
blk.7.attn_q.weightF16[3072, 3072]
-
blk.7.attn_v.weightF16[3072, 1024]
-
blk.7.ffn_down.weightF16[9216, 3072]
-
blk.7.ffn_norm.biasF32[3072]
-
blk.7.ffn_norm.weightF32[3072]
-
blk.7.ffn_up.weightF16[3072, 9216]
-
blk.8.attn_k.weightF16[3072, 1024]
-
blk.8.attn_norm.biasF32[3072]
-
blk.8.attn_norm.weightF32[3072]
-
blk.8.attn_output.weightF16[3072, 3072]
-
blk.8.attn_q.weightF16[3072, 3072]
-
blk.8.attn_v.weightF16[3072, 1024]
-
blk.8.ffn_down.weightF16[9216, 3072]
-
blk.8.ffn_norm.biasF32[3072]
-
blk.8.ffn_norm.weightF32[3072]
-
blk.8.ffn_up.weightF16[3072, 9216]
-
blk.9.attn_k.weightF16[3072, 1024]
-
blk.9.attn_norm.biasF32[3072]
-
blk.9.attn_norm.weightF32[3072]
-
blk.9.attn_output.weightF16[3072, 3072]
-
blk.9.attn_q.weightF16[3072, 3072]
-
blk.9.attn_v.weightF16[3072, 1024]
-
blk.9.ffn_down.weightF16[9216, 3072]
-
blk.9.ffn_norm.biasF32[3072]
-
blk.9.ffn_norm.weightF32[3072]
-
blk.9.ffn_up.weightF16[3072, 9216]
-
blk.10.attn_k.weightF16[3072, 1024]
-
blk.10.attn_norm.biasF32[3072]
-
blk.10.attn_norm.weightF32[3072]
-
blk.10.attn_output.weightF16[3072, 3072]
-
blk.10.attn_q.weightF16[3072, 3072]
-
blk.10.attn_v.weightF16[3072, 1024]
-
blk.10.ffn_down.weightF16[9216, 3072]
-
blk.10.ffn_norm.biasF32[3072]
-
blk.10.ffn_norm.weightF32[3072]
-
blk.10.ffn_up.weightF16[3072, 9216]
-
blk.11.attn_k.weightF16[3072, 1024]
-
blk.11.attn_norm.biasF32[3072]
-
blk.11.attn_norm.weightF32[3072]
-
blk.11.attn_output.weightF16[3072, 3072]
-
blk.11.attn_q.weightF16[3072, 3072]
-
blk.11.attn_v.weightF16[3072, 1024]
-
blk.11.ffn_down.weightF16[9216, 3072]
-
blk.11.ffn_norm.biasF32[3072]
-
blk.11.ffn_norm.weightF32[3072]
-
blk.11.ffn_up.weightF16[3072, 9216]
-
blk.12.attn_k.weightF16[3072, 1024]
-
blk.12.attn_norm.biasF32[3072]
-
blk.12.attn_norm.weightF32[3072]
-
blk.12.attn_output.weightF16[3072, 3072]
-
blk.12.attn_q.weightF16[3072, 3072]
-
blk.12.attn_v.weightF16[3072, 1024]
-
blk.12.ffn_down.weightF16[9216, 3072]
-
blk.12.ffn_norm.biasF32[3072]
-
blk.12.ffn_norm.weightF32[3072]
-
blk.12.ffn_up.weightF16[3072, 9216]
-
blk.13.attn_k.weightF16[3072, 1024]
-
blk.13.attn_norm.biasF32[3072]
-
blk.13.attn_norm.weightF32[3072]
-
blk.13.attn_output.weightF16[3072, 3072]
-
blk.13.attn_q.weightF16[3072, 3072]
-
blk.13.attn_v.weightF16[3072, 1024]
-
blk.13.ffn_down.weightF16[9216, 3072]
-
blk.13.ffn_norm.biasF32[3072]
-
blk.13.ffn_norm.weightF32[3072]
-
blk.13.ffn_up.weightF16[3072, 9216]
-
blk.14.attn_k.weightF16[3072, 1024]
-
blk.14.attn_norm.biasF32[3072]
-
blk.14.attn_norm.weightF32[3072]
-
blk.14.attn_output.weightF16[3072, 3072]
-
blk.14.attn_q.weightF16[3072, 3072]
-
blk.14.attn_v.weightF16[3072, 1024]
-
blk.14.ffn_down.weightF16[9216, 3072]
-
blk.14.ffn_norm.biasF32[3072]
-
blk.14.ffn_norm.weightF32[3072]
-
blk.14.ffn_up.weightF16[3072, 9216]
-
blk.15.attn_k.weightF16[3072, 1024]
-
blk.15.attn_norm.biasF32[3072]
-
blk.15.attn_norm.weightF32[3072]
-
blk.15.attn_output.weightF16[3072, 3072]
-
blk.15.attn_q.weightF16[3072, 3072]
-
blk.15.attn_v.weightF16[3072, 1024]
-
blk.15.ffn_down.weightF16[9216, 3072]
-
blk.15.ffn_norm.biasF32[3072]
-
blk.15.ffn_norm.weightF32[3072]
-
blk.15.ffn_up.weightF16[3072, 9216]
-
blk.16.attn_k.weightF16[3072, 1024]
-
blk.16.attn_norm.biasF32[3072]
-
blk.16.attn_norm.weightF32[3072]
-
blk.16.attn_output.weightF16[3072, 3072]
-
blk.16.attn_q.weightF16[3072, 3072]
-
blk.16.attn_v.weightF16[3072, 1024]
-
blk.16.ffn_down.weightF16[9216, 3072]
-
blk.16.ffn_norm.biasF32[3072]
-
blk.16.ffn_norm.weightF32[3072]
-
blk.16.ffn_up.weightF16[3072, 9216]
-
blk.17.attn_k.weightF16[3072, 1024]
-
blk.17.attn_norm.biasF32[3072]
-
blk.17.attn_norm.weightF32[3072]
-
blk.17.attn_output.weightF16[3072, 3072]
-
blk.17.attn_q.weightF16[3072, 3072]
-
blk.17.attn_v.weightF16[3072, 1024]
-
blk.17.ffn_down.weightF16[9216, 3072]
-
blk.17.ffn_norm.biasF32[3072]
-
blk.17.ffn_norm.weightF32[3072]
-
blk.17.ffn_up.weightF16[3072, 9216]
-
blk.18.attn_k.weightF16[3072, 1024]
-
blk.18.attn_norm.biasF32[3072]
-
blk.18.attn_norm.weightF32[3072]
-
blk.18.attn_output.weightF16[3072, 3072]
-
blk.18.attn_q.weightF16[3072, 3072]
-
blk.18.attn_v.weightF16[3072, 1024]
-
blk.18.ffn_down.weightF16[9216, 3072]
-
blk.18.ffn_norm.biasF32[3072]
-
blk.18.ffn_norm.weightF32[3072]
-
blk.18.ffn_up.weightF16[3072, 9216]
-
blk.19.attn_k.weightF16[3072, 1024]
-
blk.19.attn_norm.biasF32[3072]
-
blk.19.attn_norm.weightF32[3072]
-
blk.19.attn_output.weightF16[3072, 3072]
-
blk.19.attn_q.weightF16[3072, 3072]
-
blk.19.attn_v.weightF16[3072, 1024]
-
blk.19.ffn_down.weightF16[9216, 3072]
-
blk.19.ffn_norm.biasF32[3072]
-
blk.19.ffn_norm.weightF32[3072]
-
blk.19.ffn_up.weightF16[3072, 9216]
-
blk.20.attn_k.weightF16[3072, 1024]
-
blk.20.attn_norm.biasF32[3072]
-
blk.20.attn_norm.weightF32[3072]
-
blk.20.attn_output.weightF16[3072, 3072]
-
blk.20.attn_q.weightF16[3072, 3072]
-
blk.20.attn_v.weightF16[3072, 1024]
-
blk.20.ffn_down.weightF16[9216, 3072]
-
blk.20.ffn_norm.biasF32[3072]
-
blk.20.ffn_norm.weightF32[3072]
-
blk.20.ffn_up.weightF16[3072, 9216]
-
blk.21.attn_k.weightF16[3072, 1024]
-
blk.21.attn_norm.biasF32[3072]
-
blk.21.attn_norm.weightF32[3072]
-
blk.21.attn_output.weightF16[3072, 3072]
-
blk.21.attn_q.weightF16[3072, 3072]
-
blk.21.attn_v.weightF16[3072, 1024]
-
blk.21.ffn_down.weightF16[9216, 3072]
-
blk.21.ffn_norm.biasF32[3072]
-
blk.21.ffn_norm.weightF32[3072]
-
blk.21.ffn_up.weightF16[3072, 9216]
-
blk.22.attn_k.weightF16[3072, 1024]
-
blk.22.attn_norm.biasF32[3072]
-
blk.22.attn_norm.weightF32[3072]
-
blk.22.attn_output.weightF16[3072, 3072]
-
blk.22.attn_q.weightF16[3072, 3072]
-
blk.22.attn_v.weightF16[3072, 1024]
-
blk.22.ffn_down.weightF16[9216, 3072]
-
blk.22.ffn_norm.biasF32[3072]
-
blk.22.ffn_norm.weightF32[3072]
-
blk.22.ffn_up.weightF16[3072, 9216]
-
blk.23.attn_k.weightF16[3072, 1024]
-
blk.23.attn_norm.biasF32[3072]
-
blk.23.attn_norm.weightF32[3072]
-
blk.23.attn_output.weightF16[3072, 3072]
-
blk.23.attn_q.weightF16[3072, 3072]
-
blk.23.attn_v.weightF16[3072, 1024]
-
blk.23.ffn_down.weightF16[9216, 3072]
-
blk.23.ffn_norm.biasF32[3072]
-
blk.23.ffn_norm.weightF32[3072]
-
blk.23.ffn_up.weightF16[3072, 9216]
-
blk.24.attn_k.weightF16[3072, 1024]
-
blk.24.attn_norm.biasF32[3072]
-
blk.24.attn_norm.weightF32[3072]
-
blk.24.attn_output.weightF16[3072, 3072]
-
blk.24.attn_q.weightF16[3072, 3072]
-
blk.24.attn_v.weightF16[3072, 1024]
-
blk.24.ffn_down.weightF16[9216, 3072]
-
blk.24.ffn_norm.biasF32[3072]
-
blk.24.ffn_norm.weightF32[3072]
-
blk.24.ffn_up.weightF16[3072, 9216]
-
blk.25.attn_k.weightF16[3072, 1024]
-
blk.25.attn_norm.biasF32[3072]
-
blk.25.attn_norm.weightF32[3072]
-
blk.25.attn_output.weightF16[3072, 3072]
-
blk.25.attn_q.weightF16[3072, 3072]
-
blk.25.attn_v.weightF16[3072, 1024]
-
blk.25.ffn_down.weightF16[9216, 3072]
-
blk.25.ffn_norm.biasF32[3072]
-
blk.25.ffn_norm.weightF32[3072]
-
blk.25.ffn_up.weightF16[3072, 9216]
-
blk.26.attn_k.weightF16[3072, 1024]
-
blk.26.attn_norm.biasF32[3072]
-
blk.26.attn_norm.weightF32[3072]
-
blk.26.attn_output.weightF16[3072, 3072]
-
blk.26.attn_q.weightF16[3072, 3072]
-
blk.26.attn_v.weightF16[3072, 1024]
-
blk.26.ffn_down.weightF16[9216, 3072]
-
blk.26.ffn_norm.biasF32[3072]
-
blk.26.ffn_norm.weightF32[3072]
-
blk.26.ffn_up.weightF16[3072, 9216]
-
blk.27.attn_k.weightF16[3072, 1024]
-
blk.27.attn_norm.biasF32[3072]
-
blk.27.attn_norm.weightF32[3072]
-
blk.27.attn_output.weightF16[3072, 3072]
-
blk.27.attn_q.weightF16[3072, 3072]
-
blk.27.attn_v.weightF16[3072, 1024]
-
blk.27.ffn_down.weightF16[9216, 3072]
-
blk.27.ffn_norm.biasF32[3072]
-
blk.27.ffn_norm.weightF32[3072]
-
blk.27.ffn_up.weightF16[3072, 9216]
-
blk.28.attn_k.weightF16[3072, 1024]
-
blk.28.attn_norm.biasF32[3072]
-
blk.28.attn_norm.weightF32[3072]
-
blk.28.attn_output.weightF16[3072, 3072]
-
blk.28.attn_q.weightF16[3072, 3072]
-
blk.28.attn_v.weightF16[3072, 1024]
-
blk.28.ffn_down.weightF16[9216, 3072]
-
blk.28.ffn_norm.biasF32[3072]
-
blk.28.ffn_norm.weightF32[3072]
-
blk.28.ffn_up.weightF16[3072, 9216]
-
blk.29.attn_k.weightF16[3072, 1024]
-
blk.29.attn_norm.biasF32[3072]
-
blk.29.attn_norm.weightF32[3072]
-
blk.29.attn_output.weightF16[3072, 3072]
-
blk.29.attn_q.weightF16[3072, 3072]
-
blk.29.attn_v.weightF16[3072, 1024]
-
blk.29.ffn_down.weightF16[9216, 3072]
-
blk.29.ffn_norm.biasF32[3072]
-
blk.29.ffn_norm.weightF32[3072]
-
blk.29.ffn_up.weightF16[3072, 9216]
-
blk.30.attn_k.weightF16[3072, 1024]
-
blk.30.attn_norm.biasF32[3072]
-
blk.30.attn_norm.weightF32[3072]
-
blk.30.attn_output.weightF16[3072, 3072]
-
blk.30.attn_q.weightF16[3072, 3072]
-
blk.30.attn_v.weightF16[3072, 1024]
-
blk.30.ffn_down.weightF16[9216, 3072]
-
blk.30.ffn_norm.biasF32[3072]
-
blk.30.ffn_norm.weightF32[3072]
-
blk.30.ffn_up.weightF16[3072, 9216]
-
blk.31.attn_k.weightF16[3072, 1024]
-
blk.31.attn_norm.biasF32[3072]
-
blk.31.attn_norm.weightF32[3072]
-
blk.31.attn_output.weightF16[3072, 3072]
-
blk.31.attn_q.weightF16[3072, 3072]
-
blk.31.attn_v.weightF16[3072, 1024]
-
blk.31.ffn_down.weightF16[9216, 3072]
-
blk.31.ffn_norm.biasF32[3072]
-
blk.31.ffn_norm.weightF32[3072]
-
blk.31.ffn_up.weightF16[3072, 9216]
-
output.weightF16[3072, 256000]
-
output_norm.biasF32[3072]
-
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