What is the best open-source implementation of "FiLM: Frequency improved Legendre Memory Model for Long-term Time Series Forecasting"?

The best maintained implementation is WenjieDu/PyPOTS with 1,960 stars on GitHub. Confidence: high. Reproducibility: Strong.

Are there pretrained models available for "FiLM: Frequency improved Legendre Memory Model for Long-term Time Series Forecasting"?

Yes, 1 Hugging Face model found. The top result is Testament200156/MakeGemma3-abliterated-Improved with 45 downloads.

FiLM: Frequency improved Legendre Memory Model for Long-term Time Series Forecasting

Q: How reproducible is "FiLM: Frequency improved Legendre Memory Model for Long-term Time Series Forecasting"?

Estimated time to first reproduction: a few hours. No risk flags identified. Start with WenjieDu/PyPOTS and validate setup instructions in README.

Q: What framework is used to implement "FiLM: Frequency improved Legendre Memory Model for Long-term Time Series Forecasting"?

The primary implementation uses pytorch.

Published: May 1, 2022

Best maintained implementation now

Evidence: Direct

Domain fit: AI-adjacent

Verified repos: 2

Top repo stars: 1,960

Paper appears method- or tooling-adjacent to AI workflows with partial ecosystem coverage.

Framework: pytorch

Time to first repro: a few hours

No risk flags

arXiv PDF

Technical details

Canonical key: arxiv-2205.08897

Cache status: Stale (SWR served)

Generated at: Mar 7, 2026, 3:39 AM

Artifact coverage: direct

HF provider: ok (public)

PWC source used: Yes

LLM status: ready

LLM model: openai/gpt-5.1-20251113

LLM generated: Mar 6, 2026, 11:13 AM

LLM content type: researcher_benchmark_brief

HF policy: hf-relevance-v27

LLM evidence refs: evidencePack.paperSections[id=paper_11], evidencePack.paperSections[id=paper_10], evidencePack.paperSections[id=paper_1], evidencePack.paperSections[id=paper_2], evidencePack.paperSections[id=paper_18], evidencePack.paperSections[id=paper_20], evidencePack.paperSections[id=paper_21], evidencePack.paperSections[id=paper_22], evidencePack.paperSections[id=paper_25], researcherSummary.reproductionRisks[0], researcherSummary.benchmarkSnapshot[0], paper.title, summary.hasReliableImplementation

Researcher verdict

Recommended implementation path available

implementation baseline

Benchmark trust: grounded evidence

Quality tier: researcher ready

This page has evidence-backed benchmark findings and a concrete implementation recommendation anchored on WenjieDu/PyPOTS. Use it as an implementation baseline, then validate benchmark parity before adapting it.

Why this page is still worth reading

Benchmark findings give you an audit trail for validation before picking an implementation path.
A concrete repository path exists via WenjieDu/PyPOTS, so this page can act as a practical starting point.
Reproduction risks are surfaced explicitly, which helps decide whether the paper is worth immediate prototyping.

Benchmark trust

Concrete benchmark findings are present and can be audited against the extracted evidence.

Use this page as

Start here when you need the most practical implementation path quickly.

Results & Benchmarks

Freshness tier: cold

Direct + Inferred Evidence

Long-term univariate time series forecasting

Weather and Electricity

Reported value

Source: llm grounded

LPU boosting across backbones

ETT full

average performance gain

MLP 119.4%, LSTM 97.0%, CNN 13.8%, Attn

Source: llm grounded

Benchmark evidence drill-down

2 findings

Audit each benchmark finding before selecting an implementation path. Evidence refs map to the disclosure section below.

Task	Dataset	Metric	Value	Source	Evidence refs
Long-term univariate time series forecasting	Weather and Electricity	Reported value	17	llm-grounded	evidencePack.paperSections[id=paper_21]
LPU boosting across backbones	ETT full	average performance gain	MLP 119.4%, LSTM 97.0%, CNN 13.8%…	llm-grounded	evidencePack.paperSections[id=paper_22]

FiLM: Frequency improved Legendre Memory Model for Long-term Time Series Forecasting is the primary contribution described in this paper.

Use This Implementation Because…

Confidence: high

WenjieDu/PyPOTS is the strongest maintained implementation based on ranking signals. CI workflows are present. License is declared (BSD-3-Clause).

Open WenjieDu/PyPOTS

Reproduction Risks

No repository-level red flags were detected, but paper-specific preprocessing and hyperparameter details may still be under-specified.

Evidence disclosure

Evidence graph: 4 refs, 4 links.

Utility signals: depth 90/100, grounding 95/100, status high.

Implementation Comparison

Top 3 paths

Compare maintenance quality, reproducibility coverage, and evidence confidence before choosing a reproduction baseline.

WenjieDu/PyPOTS

best maintained

Maintenance: Active

Confidence: High

Reproducibility: Strong

Official implementation from Papers with Code · Strong overlap with paper title keywords

Stars: 1,960
Last push: Mar 6, 2026 (1d ago)

CIReleasesDependencies

Risk flags

No Docker setup

tianzhou2011/FiLM

alternative

Maintenance: Stale

Confidence: Medium

Reproducibility: Moderate

Repository link is mentioned in the paper metadata · Matched via arXiv identifier search

Stars: 97
Last push: Oct 31, 2022 (1223d ago)

Dependencies

Risk flags

No push in 12+ months
No CI pipeline detected
No tagged releases

damo-di-ml/neurips2022-film

alternative

Maintenance: Stale

Confidence: Low

Reproducibility: Moderate

Strong overlap with paper title keywords · Community adoption signal (33 stars)

Stars: 33
Last push: Oct 31, 2022 (1223d ago)

Dependencies

Risk flags

No push in 12+ months
No CI pipeline detected
No tagged releases

Paper summary

AI-generated

AI-generated summary grounded in paper metadata and artifact signals.

FiLM (Frequency improved Legendre Memory Model) maps an input time-series sequence to an output sequence using two main sub-layers, the Legendre Projection Unit and the Fourier Enhanced Layer, together with a data normalization block RevIN. This page includes benchmark evidence for Long-term multivariate time series forecasting on six real-world benchmark datasets. Reproduction guidance focuses on implementation viability and concrete risk controls.

Key contributions

FiLM (Frequency improved Legendre Memory Model) maps an input time-series sequence to an output sequence using two main sub-layers, the Legendre Projection Unit and the Fourier Enhanced Layer, together with a data.
FiLM is proposed as a deep learning architecture specifically designed for long-term time series forecasting, targeting improved preservation and utilization of historical information compared to prior RNN.
On multivariate long-term forecasting tasks, FiLM achieves the best performance across all six benchmark datasets and all forecast horizons compared with baselines such as FEDformer.
For univariate long-term time series forecasting, FiLM shows substantial gains, with improvements exceeding 40 percent on some datasets such as Weather and Electricity over strong baselines.
The Legendre Projection Unit acts as a pluggable memory module that significantly boosts long-term forecasting performance of various backbone models, improving MLP by 119.4%, LSTM by 97.0%, CNN by 13.8%.

Implementation guidance

Use WenjieDu/PyPOTS first because deterministic ranking and extracted evidence align on implementation viability. Start with the repo setup path, then validate benchmark reproduction before adaptation.

Reproducibility notes

Reproduction results may deviate from the paper if dataset preprocessing steps are not matched to the original FiLM setup.
Insufficiently documented hyperparameters or training schedules can lead to significantly worse performance than the reported FiLM benchmarks.
Differences in software stack or library versions relative to the original implementation may subtly affect long-horizon forecasting accuracy.

Best implementation now

WenjieDu/PyPOTS

Confidence: High

Reproducibility: Strong

A Python toolkit/library for reality-centric machine/deep learning and data mining on partially-observed time series, including SOTA neural network models for scientific analysis tasks of imputation/classification/clustering/forecasting/anomaly detection/cleaning on incomplete industrial (irregularly-sampled) multivariate TS with NaN missing values

Stars: 1,960

Forks: 179

Last push: Mar 6, 2026

License: BSD-3-Clause

Official implementation from Papers with Code

Strong overlap with paper title keywords

Community adoption signal (1960 stars)

License ✓

CI ✓

Deps ✓

Docker –

Selected WenjieDu/PyPOTS as the strongest maintained implementation for new work.
Includes CI workflow signals.
Includes dependency/environment manifest signals.
Repository activity is within the last 24 months.

Reproduction path

Direct

Follow the direct implementation path

1

Start with WenjieDu/PyPOTS and validate setup instructions in README.
2

Reproduce the baseline result with the provided defaults before modifying hyperparameters.
3

Log exact dependency versions and runtime environment for reproducibility.

Time to first repro: a few hours