PyTorch Hub, a repository offering pretrained models such as ResNet, BERT and GPT, was announced on 2019-06-10.
Resolved true The record shows it happened or was so.
Factual: Statements about the present or the past, checked against the record: did they hold up?
Said for an organisation they lead or work for; it counts on both records.
What would count as true or false
True: PyTorch Hub was officially announced (e.g. by a PyTorch blog post) on or about 2019-06-10, as a repository of pretrained models that included models such as ResNet, BERT and GPT(-2) at or near launch. Mostly true: announced on that date, but one of the example models was not available at launch, or the date is off by a day or two. Mixed: the announcement was on a different date by weeks, or most of the named models were missing. Mostly false: PyTorch Hub launched but the announcement was months away from that date. False: no PyTorch Hub announcement took place around then.
Set before the research began.
Evidence (4)
-
E1 Towards Reproducible Research with PyTorch Hub - PyTorch PyTorch (official blog) · 10 June 2019
“We are excited to announce the availability of PyTorch Hub, a simple API and workflow that provides the basic building blocks for improving machine learning research reproducibility.”
-
E2 Towards Reproducible Research with PyTorch Hub - PyTorch PyTorch (official blog) · 10 June 2019
“It also has built-in support for Colab, integration with Papers With Code and currently contains a broad set of models that include Classification and Segmentation, Generative, Transformers, etc.”
-
E3 Towards Reproducible Research with PyTorch Hub - PyTorch PyTorch (official blog) · 10 June 2019
“from torchvision.models.resnet import resnet18, resnet34, resnet50, resnet101, resnet152,”
-
E4 Facebook launches PyTorch Hub for reproducing AI model results VentureBeat · 10 June 2019
“At launch, PyTorch Hub comes with access to roughly 20 pretrained versions of Google's BERT, WaveGlow, and Tacotron 2 from Nvidia, and the Generative Pre-Training (GPT) for language understanding from Hugging Face.”
Reasoning
The official PyTorch blog post dated 2019-06-10 announced the availability of PyTorch Hub, a hub of models in categories that included Transformers (E1, E2). The same post's torchvision example lists ResNet entrypoints (E3). On the same date, VentureBeat reported that at launch the hub gave access to pretrained BERT and to GPT from Hugging Face (E4). This matches the criteria for true: the announcement came on the stated date and included ResNet, BERT and GPT.
Cite this claim
Yann LeCun, “PyTorch Hub, a repository offering pretrained models such as ResNet, BERT and GPT, was announced on 2019-06-10.” (said 10 June 2019). Verdict: resolved true. Reliability API, claim c_aebef10a9755, method 0.1.0+c876030c, dataset release v1.1.2. https://epistacklabs.com/claims/c_aebef10a9755
This claim is in dataset release v1.1.2, with the same verdict.
Reply to this claim
More factual claims by Yann LeCun
-
Mostly true
“This white-noise sounding signal with a flat spectrum is the pseudo-random bit sequence used to pre-train the adaptive equalizer that cleans up the mushy analog signal into clean bits.”
-
Resolved true
“Opacus: a new PyTorch library for ML with differential privacy.”
Facebook AI released Opacus, a new PyTorch library for machine learning with differential privacy.
-
Mostly true
“We are using Torch7 for many projects (as does Deep Mind and several groups at Google) and will be contributing to the public version.”
Facebook AI Research uses Torch7 for many projects, as do Deep Mind and several groups at Google.
-
Resolved false
““Those agents are doing exactly what they’ve been asked to do,” LeCun said.”
-
Resolved true
““It’s AI for the physical world, so it’s not language-related,” LeCun said.”
AMI Labs' work is AI for the physical world and is not language-related.
Verdicts are made automatically; how verdicts are made.