NuExtract-tiny-v1.5 is a fine-tuning of Qwen/Qwen2.5-0.5B, trained on a private high-quality dataset for structured information extraction. It supports long documents and several languages (English, French, Spanish, German, Portuguese, and Italian).
12 Pulls Updated 4 days ago
ebee6e5e0f8a · 994MB
-
general.architectureqwen2
-
general.base_model.0.nameQwen2.5 0.5B
-
general.base_model.0.organizationQwen
-
general.base_model.0.repo_urlhttps://huggingface.co/Qwen/Qwen2.5-0.5B
-
general.base_model.count1
-
general.basenameNuExtract-tiny
-
general.file_type1
-
general.languages[multilingual]
-
general.licensemit
-
general.nameNuExtract Tiny v1.5
-
general.organizationNumind
-
general.quantization_version2
-
general.size_label494M
-
general.tags[nlp, text-generation]
-
general.typemodel
-
general.versionv1.5
-
qwen2.attention.head_count14
-
qwen2.attention.head_count_kv2
-
qwen2.attention.layer_norm_rms_epsilon1e-06
-
qwen2.block_count24
-
qwen2.context_length32768
-
qwen2.embedding_length896
-
qwen2.feed_forward_length4864
-
qwen2.rope.freq_base1e+06
-
tokenizer.ggml.add_bos_tokenfalse
-
tokenizer.ggml.bos_token_id151643
-
tokenizer.ggml.eos_token_id151643
-
tokenizer.ggml.merges[Ġ Ġ, ĠĠ ĠĠ, i n, Ġ t, ĠĠĠĠ ĠĠĠĠ, ...]
-
tokenizer.ggml.modelgpt2
-
tokenizer.ggml.padding_token_id151643
-
tokenizer.ggml.preqwen2
-
tokenizer.ggml.token_type[1, 1, 1, 1, 1, ...]
-
tokenizer.ggml.tokens[!, ", #, $, %, ...]
-
NameTypeShape
-
token_embd.weightF16[896, 151936]
-
blk.0.attn_k.biasF32[128]
-
blk.0.attn_k.weightF16[896, 128]
-
blk.0.attn_norm.weightF32[896]
-
blk.0.attn_output.weightF16[896, 896]
-
blk.0.attn_q.biasF32[896]
-
blk.0.attn_q.weightF16[896, 896]
-
blk.0.attn_v.biasF32[128]
-
blk.0.attn_v.weightF16[896, 128]
-
blk.0.ffn_down.weightF16[4864, 896]
-
blk.0.ffn_gate.weightF16[896, 4864]
-
blk.0.ffn_norm.weightF32[896]
-
blk.0.ffn_up.weightF16[896, 4864]
-
blk.1.attn_k.biasF32[128]
-
blk.1.attn_k.weightF16[896, 128]
-
blk.1.attn_norm.weightF32[896]
-
blk.1.attn_output.weightF16[896, 896]
-
blk.1.attn_q.biasF32[896]
-
blk.1.attn_q.weightF16[896, 896]
-
blk.1.attn_v.biasF32[128]
-
blk.1.attn_v.weightF16[896, 128]
-
blk.1.ffn_down.weightF16[4864, 896]
-
blk.1.ffn_gate.weightF16[896, 4864]
-
blk.1.ffn_norm.weightF32[896]
-
blk.1.ffn_up.weightF16[896, 4864]
-
blk.2.attn_k.biasF32[128]
-
blk.2.attn_k.weightF16[896, 128]
-
blk.2.attn_norm.weightF32[896]
-
blk.2.attn_output.weightF16[896, 896]
-
blk.2.attn_q.biasF32[896]
-
blk.2.attn_q.weightF16[896, 896]
-
blk.2.attn_v.biasF32[128]
-
blk.2.attn_v.weightF16[896, 128]
-
blk.2.ffn_down.weightF16[4864, 896]
-
blk.2.ffn_gate.weightF16[896, 4864]
-
blk.2.ffn_norm.weightF32[896]
-
blk.2.ffn_up.weightF16[896, 4864]
-
blk.3.attn_k.biasF32[128]
-
blk.3.attn_k.weightF16[896, 128]
-
blk.3.attn_norm.weightF32[896]
-
blk.3.attn_output.weightF16[896, 896]
-
blk.3.attn_q.biasF32[896]
-
blk.3.attn_q.weightF16[896, 896]
-
blk.3.attn_v.biasF32[128]
-
blk.3.attn_v.weightF16[896, 128]
-
blk.3.ffn_down.weightF16[4864, 896]
-
blk.3.ffn_gate.weightF16[896, 4864]
-
blk.3.ffn_norm.weightF32[896]
-
blk.3.ffn_up.weightF16[896, 4864]
-
blk.4.attn_k.biasF32[128]
-
blk.4.attn_k.weightF16[896, 128]
-
blk.4.attn_norm.weightF32[896]
-
blk.4.attn_output.weightF16[896, 896]
-
blk.4.attn_q.biasF32[896]
-
blk.4.attn_q.weightF16[896, 896]
-
blk.4.attn_v.biasF32[128]
-
blk.4.attn_v.weightF16[896, 128]
-
blk.4.ffn_down.weightF16[4864, 896]
-
blk.4.ffn_gate.weightF16[896, 4864]
-
blk.4.ffn_norm.weightF32[896]
-
blk.4.ffn_up.weightF16[896, 4864]
-
blk.5.attn_k.biasF32[128]
-
blk.5.attn_k.weightF16[896, 128]
-
blk.5.attn_norm.weightF32[896]
-
blk.5.attn_output.weightF16[896, 896]
-
blk.5.attn_q.biasF32[896]
-
blk.5.attn_q.weightF16[896, 896]
-
blk.5.attn_v.biasF32[128]
-
blk.5.attn_v.weightF16[896, 128]
-
blk.5.ffn_down.weightF16[4864, 896]
-
blk.5.ffn_gate.weightF16[896, 4864]
-
blk.5.ffn_norm.weightF32[896]
-
blk.5.ffn_up.weightF16[896, 4864]
-
blk.6.attn_k.biasF32[128]
-
blk.6.attn_k.weightF16[896, 128]
-
blk.6.attn_norm.weightF32[896]
-
blk.6.attn_output.weightF16[896, 896]
-
blk.6.attn_q.biasF32[896]
-
blk.6.attn_q.weightF16[896, 896]
-
blk.6.attn_v.biasF32[128]
-
blk.6.attn_v.weightF16[896, 128]
-
blk.6.ffn_down.weightF16[4864, 896]
-
blk.6.ffn_gate.weightF16[896, 4864]
-
blk.6.ffn_norm.weightF32[896]
-
blk.6.ffn_up.weightF16[896, 4864]
-
blk.7.attn_k.biasF32[128]
-
blk.7.attn_k.weightF16[896, 128]
-
blk.7.attn_norm.weightF32[896]
-
blk.7.attn_output.weightF16[896, 896]
-
blk.7.attn_q.biasF32[896]
-
blk.7.attn_q.weightF16[896, 896]
-
blk.7.attn_v.biasF32[128]
-
blk.7.attn_v.weightF16[896, 128]
-
blk.7.ffn_down.weightF16[4864, 896]
-
blk.7.ffn_gate.weightF16[896, 4864]
-
blk.7.ffn_norm.weightF32[896]
-
blk.7.ffn_up.weightF16[896, 4864]
-
blk.8.attn_k.biasF32[128]
-
blk.8.attn_k.weightF16[896, 128]
-
blk.8.attn_norm.weightF32[896]
-
blk.8.attn_output.weightF16[896, 896]
-
blk.8.attn_q.biasF32[896]
-
blk.8.attn_q.weightF16[896, 896]
-
blk.8.attn_v.biasF32[128]
-
blk.8.attn_v.weightF16[896, 128]
-
blk.8.ffn_down.weightF16[4864, 896]
-
blk.8.ffn_gate.weightF16[896, 4864]
-
blk.8.ffn_norm.weightF32[896]
-
blk.8.ffn_up.weightF16[896, 4864]
-
blk.9.attn_k.biasF32[128]
-
blk.9.attn_k.weightF16[896, 128]
-
blk.9.attn_norm.weightF32[896]
-
blk.9.attn_output.weightF16[896, 896]
-
blk.9.attn_q.biasF32[896]
-
blk.9.attn_q.weightF16[896, 896]
-
blk.9.attn_v.biasF32[128]
-
blk.9.attn_v.weightF16[896, 128]
-
blk.9.ffn_down.weightF16[4864, 896]
-
blk.9.ffn_gate.weightF16[896, 4864]
-
blk.9.ffn_norm.weightF32[896]
-
blk.9.ffn_up.weightF16[896, 4864]
-
blk.10.attn_k.biasF32[128]
-
blk.10.attn_k.weightF16[896, 128]
-
blk.10.attn_norm.weightF32[896]
-
blk.10.attn_output.weightF16[896, 896]
-
blk.10.attn_q.biasF32[896]
-
blk.10.attn_q.weightF16[896, 896]
-
blk.10.attn_v.biasF32[128]
-
blk.10.attn_v.weightF16[896, 128]
-
blk.10.ffn_down.weightF16[4864, 896]
-
blk.10.ffn_gate.weightF16[896, 4864]
-
blk.10.ffn_norm.weightF32[896]
-
blk.10.ffn_up.weightF16[896, 4864]
-
blk.11.attn_k.biasF32[128]
-
blk.11.attn_k.weightF16[896, 128]
-
blk.11.attn_norm.weightF32[896]
-
blk.11.attn_output.weightF16[896, 896]
-
blk.11.attn_q.biasF32[896]
-
blk.11.attn_q.weightF16[896, 896]
-
blk.11.attn_v.biasF32[128]
-
blk.11.attn_v.weightF16[896, 128]
-
blk.11.ffn_down.weightF16[4864, 896]
-
blk.11.ffn_gate.weightF16[896, 4864]
-
blk.11.ffn_norm.weightF32[896]
-
blk.11.ffn_up.weightF16[896, 4864]
-
blk.12.attn_k.biasF32[128]
-
blk.12.attn_k.weightF16[896, 128]
-
blk.12.attn_norm.weightF32[896]
-
blk.12.attn_output.weightF16[896, 896]
-
blk.12.attn_q.biasF32[896]
-
blk.12.attn_q.weightF16[896, 896]
-
blk.12.attn_v.biasF32[128]
-
blk.12.attn_v.weightF16[896, 128]
-
blk.12.ffn_down.weightF16[4864, 896]
-
blk.12.ffn_gate.weightF16[896, 4864]
-
blk.12.ffn_norm.weightF32[896]
-
blk.12.ffn_up.weightF16[896, 4864]
-
blk.13.attn_k.biasF32[128]
-
blk.13.attn_k.weightF16[896, 128]
-
blk.13.attn_norm.weightF32[896]
-
blk.13.attn_output.weightF16[896, 896]
-
blk.13.attn_q.biasF32[896]
-
blk.13.attn_q.weightF16[896, 896]
-
blk.13.attn_v.biasF32[128]
-
blk.13.attn_v.weightF16[896, 128]
-
blk.13.ffn_down.weightF16[4864, 896]
-
blk.13.ffn_gate.weightF16[896, 4864]
-
blk.13.ffn_norm.weightF32[896]
-
blk.13.ffn_up.weightF16[896, 4864]
-
blk.14.attn_k.biasF32[128]
-
blk.14.attn_k.weightF16[896, 128]
-
blk.14.attn_norm.weightF32[896]
-
blk.14.attn_output.weightF16[896, 896]
-
blk.14.attn_q.biasF32[896]
-
blk.14.attn_q.weightF16[896, 896]
-
blk.14.attn_v.biasF32[128]
-
blk.14.attn_v.weightF16[896, 128]
-
blk.14.ffn_down.weightF16[4864, 896]
-
blk.14.ffn_gate.weightF16[896, 4864]
-
blk.14.ffn_norm.weightF32[896]
-
blk.14.ffn_up.weightF16[896, 4864]
-
blk.15.attn_k.biasF32[128]
-
blk.15.attn_k.weightF16[896, 128]
-
blk.15.attn_norm.weightF32[896]
-
blk.15.attn_output.weightF16[896, 896]
-
blk.15.attn_q.biasF32[896]
-
blk.15.attn_q.weightF16[896, 896]
-
blk.15.attn_v.biasF32[128]
-
blk.15.attn_v.weightF16[896, 128]
-
blk.15.ffn_down.weightF16[4864, 896]
-
blk.15.ffn_gate.weightF16[896, 4864]
-
blk.15.ffn_norm.weightF32[896]
-
blk.15.ffn_up.weightF16[896, 4864]
-
blk.16.attn_k.biasF32[128]
-
blk.16.attn_k.weightF16[896, 128]
-
blk.16.attn_norm.weightF32[896]
-
blk.16.attn_output.weightF16[896, 896]
-
blk.16.attn_q.biasF32[896]
-
blk.16.attn_q.weightF16[896, 896]
-
blk.16.attn_v.biasF32[128]
-
blk.16.attn_v.weightF16[896, 128]
-
blk.16.ffn_down.weightF16[4864, 896]
-
blk.16.ffn_gate.weightF16[896, 4864]
-
blk.16.ffn_norm.weightF32[896]
-
blk.16.ffn_up.weightF16[896, 4864]
-
blk.17.attn_k.biasF32[128]
-
blk.17.attn_k.weightF16[896, 128]
-
blk.17.attn_norm.weightF32[896]
-
blk.17.attn_output.weightF16[896, 896]
-
blk.17.attn_q.biasF32[896]
-
blk.17.attn_q.weightF16[896, 896]
-
blk.17.attn_v.biasF32[128]
-
blk.17.attn_v.weightF16[896, 128]
-
blk.17.ffn_down.weightF16[4864, 896]
-
blk.17.ffn_gate.weightF16[896, 4864]
-
blk.17.ffn_norm.weightF32[896]
-
blk.17.ffn_up.weightF16[896, 4864]
-
blk.18.attn_k.biasF32[128]
-
blk.18.attn_k.weightF16[896, 128]
-
blk.18.attn_norm.weightF32[896]
-
blk.18.attn_output.weightF16[896, 896]
-
blk.18.attn_q.biasF32[896]
-
blk.18.attn_q.weightF16[896, 896]
-
blk.18.attn_v.biasF32[128]
-
blk.18.attn_v.weightF16[896, 128]
-
blk.18.ffn_down.weightF16[4864, 896]
-
blk.18.ffn_gate.weightF16[896, 4864]
-
blk.18.ffn_norm.weightF32[896]
-
blk.18.ffn_up.weightF16[896, 4864]
-
blk.19.attn_k.biasF32[128]
-
blk.19.attn_k.weightF16[896, 128]
-
blk.19.attn_norm.weightF32[896]
-
blk.19.attn_output.weightF16[896, 896]
-
blk.19.attn_q.biasF32[896]
-
blk.19.attn_q.weightF16[896, 896]
-
blk.19.attn_v.biasF32[128]
-
blk.19.attn_v.weightF16[896, 128]
-
blk.19.ffn_down.weightF16[4864, 896]
-
blk.19.ffn_gate.weightF16[896, 4864]
-
blk.19.ffn_norm.weightF32[896]
-
blk.19.ffn_up.weightF16[896, 4864]
-
blk.20.attn_k.biasF32[128]
-
blk.20.attn_k.weightF16[896, 128]
-
blk.20.attn_norm.weightF32[896]
-
blk.20.attn_output.weightF16[896, 896]
-
blk.20.attn_q.biasF32[896]
-
blk.20.attn_q.weightF16[896, 896]
-
blk.20.attn_v.biasF32[128]
-
blk.20.attn_v.weightF16[896, 128]
-
blk.20.ffn_down.weightF16[4864, 896]
-
blk.20.ffn_gate.weightF16[896, 4864]
-
blk.20.ffn_norm.weightF32[896]
-
blk.20.ffn_up.weightF16[896, 4864]
-
blk.21.attn_k.biasF32[128]
-
blk.21.attn_k.weightF16[896, 128]
-
blk.21.attn_norm.weightF32[896]
-
blk.21.attn_output.weightF16[896, 896]
-
blk.21.attn_q.biasF32[896]
-
blk.21.attn_q.weightF16[896, 896]
-
blk.21.attn_v.biasF32[128]
-
blk.21.attn_v.weightF16[896, 128]
-
blk.21.ffn_down.weightF16[4864, 896]
-
blk.21.ffn_gate.weightF16[896, 4864]
-
blk.21.ffn_norm.weightF32[896]
-
blk.21.ffn_up.weightF16[896, 4864]
-
blk.22.attn_k.biasF32[128]
-
blk.22.attn_k.weightF16[896, 128]
-
blk.22.attn_norm.weightF32[896]
-
blk.22.attn_output.weightF16[896, 896]
-
blk.22.attn_q.biasF32[896]
-
blk.22.attn_q.weightF16[896, 896]
-
blk.22.attn_v.biasF32[128]
-
blk.22.attn_v.weightF16[896, 128]
-
blk.22.ffn_down.weightF16[4864, 896]
-
blk.22.ffn_gate.weightF16[896, 4864]
-
blk.22.ffn_norm.weightF32[896]
-
blk.22.ffn_up.weightF16[896, 4864]
-
blk.23.attn_k.biasF32[128]
-
blk.23.attn_k.weightF16[896, 128]
-
blk.23.attn_norm.weightF32[896]
-
blk.23.attn_output.weightF16[896, 896]
-
blk.23.attn_q.biasF32[896]
-
blk.23.attn_q.weightF16[896, 896]
-
blk.23.attn_v.biasF32[128]
-
blk.23.attn_v.weightF16[896, 128]
-
blk.23.ffn_down.weightF16[4864, 896]
-
blk.23.ffn_gate.weightF16[896, 4864]
-
blk.23.ffn_norm.weightF32[896]
-
blk.23.ffn_up.weightF16[896, 4864]
-
output_norm.weightF32[896]
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