Inference_HUAWEI is an open-weight model from Linfeng Wang. Its published files total 3.1 GB.
本包提供内源多模态数据生成、内源推理、外源 bank 构建、离线权重校准和内源＋外源推理。 先按第 1–3 节推理一个内源例子；需要外源检索时再看第 5 节。真实 M4 yearly 的切分、推理和 MASE 计算见完整评测例子。 - 原始历史可以是 NPZ/Zarr、CSV、Arrow 或 Parquet；不能包含可被编码的未来标签。 - query.npz 除数值历史外，还包含每个 32 点 patch 的图像、文本特征。 - runendogenous.py 读取已经生成的特征，不直接把原始…
Model Card
本包提供内源多模态数据生成、内源推理、外源 bank 构建、离线权重校准和内源＋外源推理。 先按第 1–3 节推理一个内源例子；需要外源检索时再看第 5 节。真实 M4 yearly 的切分、推理和 MASE 计算见完整评测例子。 - 原始历史可以是 NPZ/Zarr、CSV、Arrow 或 Parquet；不能包含可被编码的未来标签。 - query.npz 除数值历史外，还包含每个 32 点 patch 的图像、文本特征。 - runendogenous.py 读取已经生成的特征，不直接把原始 CSV/Arrow 当模型输入。 建议 Python 3.11，支持 3.10–3.12。依赖 requirements.txt。 把下面示例路径换成你自己的路径，文件名保持 checkpoint 原来的写法： 只做内源推理需要内源模型和 CLIP。若测最终联合训练模型的内源结果，还必须加载最终 sharedheadstate.pt；不传 --shared-head 使用的是内源目录里的旧 head。 外源推理另需外源模型和 TS 检索编码器。TS encoder 是冻结的数值骨干，可选用任意基模（TimesFM3等）。 两份文件的变量 ID 都是 series-A、series-B。这个例子的查询取自合成训练序列末尾，仅用于检查接口；它不是独立的 benchmark 测试集。 此时 demo/query.npz 包含： 256 个点对应 8 个 32 点 patch；每个 patch 的图像和文本都编码成 512 维特征。不调用大语言模型，描述由统计量模板生成。…
Excerpt from the card by Linfeng Wang.
Identity and Version
- Repository
- linfengwangpower/Inference_HUAWEI
- Publisher
- Linfeng Wang
- Task
- Not stated by the source
- Modality
- Other
- Library
- Not stated by the source
- Parameters
- Not stated by the source
- Languages
- Not stated by the source
- Revision
- 3be8d7f119cb3f68c969bc208fbe7c8da2a45dc4
- First published
- 2026-10-14
- Last updated
- 2026-10-14
Files and Weights
52 files, 3.1 GB in total. The weights are 6 files totalling 3.1 GB in pt, safetensors.
Every file
| File | Type | Size | SHA-256 |
|---|---|---|---|
| checkpoints/endogenous/layer16_concat_additive_state.pt | Weights | 146.9 MB | ebe0ca3e2bfa |
| checkpoints/endogenous/model.safetensors | Weights | 1.3 GB | c0114a2a90fa |
| checkpoints/external/external_rag_state.pt | Weights | 104.9 MB | b1d5426bafbe |
| checkpoints/external/shared_head_state.pt | Weights | 22.7 MB | 1050fa6b72ec |
| checkpoints/ts_encoder/model.safetensors | Weights | 1.3 GB | 9e85c449bc2b |
| weights/ViT-B-32.pt | Weights | 354.0 MB | 40d365715913 |
| calibrate_weights.py | Configuration | 2.1 KB | — |
| checkpoints/endogenous/config.json | Configuration | 598 B | — |
| checkpoints/endogenous/layer16_concat_additive_config.json | Configuration | 149 B | — |
| checkpoints/external/external_rag_config.json | Configuration | 3.4 KB | — |
| checkpoints/ts_encoder/config.json | Configuration | 598 B | — |
| dd_unit_scaling/__init__.py | Configuration | 1.9 KB | — |
| dd_unit_scaling/_modules.py | Configuration | 6.0 KB | — |
| dd_unit_scaling/functional.py | Configuration | 9.1 KB | — |
| dd_unit_scaling/optim.py | Configuration | 11.7 KB | — |
| dd_unit_scaling/scale.py | Configuration | 1.8 KB | — |
| examples/make_demo.py | Configuration | 831 B | — |
| generate_endogenous.py | Configuration | 3.1 KB | — |
| generate_external.py | Configuration | 5.1 KB | — |
| qiyaom/__init__.py | Configuration | 573 B | — |
| qiyaom/backbone_qiyaom.py | Configuration | 62.5 KB | — |
| qiyaom/configuration_backbone.py | Configuration | 3.9 KB | — |
| qiyaom/configuration_qiyaom.py | Configuration | 1.3 KB | — |
| qiyaom/modeling_qiyaom.py | Configuration | 20.1 KB | — |
| qiyaom/pipeline.py | Configuration | 7.3 KB | — |
| qiyaom/retrieval_qiyaom.py | Configuration | 5.4 KB | — |
| qiyaom/utils.py | Configuration | 1.9 KB | — |
| qiyaom_tools/__init__.py | Configuration | 82 B | — |
| qiyaom_tools/encoding.py | Configuration | 8.4 KB | — |
| qiyaom_tools/io.py | Configuration | 8.7 KB | — |
| qiyaom_tools/render.py | Configuration | 5.0 KB | — |
| qiyaom_tools/retrieval.py | Configuration | 8.3 KB | — |
| qiyaom_tools/runtime.py | Configuration | 1.9 KB | — |
| qiyaom_tools/sequence_text_template.py | Configuration | 16.3 KB | — |
| qiyaom_tools/tabular.py | Configuration | 9.6 KB | — |
| qiyaom_tools/text_template.py | Configuration | 14.1 KB | — |
| run_endogenous.py | Configuration | 3.1 KB | — |
| run_external.py | Configuration | 7.6 KB | — |
| LICENSE | Documentation | 11.4 KB | — |
| README.md | Documentation | 25.5 KB | — |
| docs/DATA_FORMAT.md | Documentation | 7.8 KB | — |
| docs/EXAMPLE_M4_YEARLY.md | Documentation | 8.6 KB | — |
| docs/RETRIEVAL_OPTIONS.md | Documentation | 6.3 KB | — |
| calibrate_weights.sh | Other | 318 B | — |
| generate_endogenous.sh | Other | 611 B | — |
| generate_external.sh | Other | 454 B | — |
| licenses/Apache-2.0.txt | Other | 11.4 KB | — |
| licenses/CLIP-MIT.txt | Other | 1.1 KB | — |
| requirements.txt | Other | 541 B | — |
| run_endogenous.sh | Other | 361 B | — |
| run_external.sh | Other | 508 B | — |
| .gitattributes | Repository | 1.5 KB | — |
License and Download
- License
- Not stated by the source
- Access
- Open weights, no gate
- Download size
- 3.1 GB
Released by Linfeng Wang through its official repository on Hugging Face.
Memory Requirements
| Precision | Weights in memory |
|---|---|
| As published | 3.1 GB |
Weights only, from the published parameter count; the key-value cache and runtime add to this.