Jonathan/1013 weekly assets

assets/adobe.yaml

@@ -0,0 +1,88 @@
+---
+- type: model
+  name: Firefly Image 2
+  organization: Adobe
+  description: Firefly Image 2 is the next generation of generative AI for imaging, bringing significant advancements to creative control and quality, including new Text to Image capabilities now available in the popular Firefly web app where 90% of users are new to Adobe products.
+  created_date: 2023-10-10
+  url: https://firefly.adobe.com/
+  model_card: none
+  modality: text; image
+  analysis: ''
+  size: unknown
+  dependencies: []
+  training_emissions: unknown
+  training_time: unknown
+  training_hardware: unknown
+  quality_control: ''
+  access: closed
+  license: unknown
+  intended_uses: creative generation of digital art and images
+  prohibited_uses: AI/ML training, attempting to create abusive, illegal, or confidential content.
+  monitoring: ''
+  feedback: ''
+
+- type: model
+  name: Firefly Vector
+  organization: Adobe
+  description: Firefly Vector is the world’s first generative AI focused on producing vector graphics, bringing Adobe's vector graphic and generative AI expertise directly into Adobe Illustrator workflows with Text to Vector Graphic.
+  created_date: 2023-10-10
+  url: https://firefly.adobe.com/
+  model_card: none
+  modality: text; vector graphic
+  analysis: ''
+  size: unknown
+  dependencies: []
+  training_emissions: unknown
+  training_time: unknown
+  training_hardware: unknown
+  quality_control: ''
+  access: closed
+  license: unknown
+  intended_uses: creative generation of digital art and images
+  prohibited_uses: AI/ML training, attempting to create abusive, illegal, or confidential content.
+  monitoring: ''
+  feedback: ''
+
+- type: model
+  name: Firefly Design
+  organization: Adobe
+  description: Firefly Design powers instant generation of amazing quality template designs in Adobe Express with the new Text to Template capability.
+  created_date: 2023-10-10
+  url: https://firefly.adobe.com/
+  model_card: none
+  modality: text; template design
+  analysis: ''
+  size: unknown
+  dependencies: []
+  training_emissions: unknown
+  training_time: unknown
+  training_hardware: unknown
+  quality_control: ''
+  access: closed
+  license: unknown
+  intended_uses: creative generation of digital art and images
+  prohibited_uses: AI/ML training, attempting to create abusive, illegal, or confidential content.
+  monitoring: ''
+  feedback: ''
+
+- type: application
+  name: Firefly
+  organization: Adobe
+  description: Adobe Firefly is a standalone web application. It offers new ways to ideate, create, and communicate while significantly improving creative workflows using generative AI.
+  created_date: 2023-03-21
+  url: https://firefly.adobe.com/
+  dependencies: [Firefly Image 2, Firefly Vector, Firefly Design]
+  adaptation: ''
+  output_space: AI-generated creations
+  quality_control: ''
+  access: limited
+  license: unknown
+  terms_of_service: https://www.adobe.com/legal/licenses-terms/adobe-gen-ai-user-guidelines.html
+  intended_uses: creative generation of digital art and images
+  prohibited_uses: AI/ML training, attempting to create abusive, illegal, or confidential content.
+  monitoring: ''
+  feedback: ''
+  monthly_active_users: unknown
+  user_distribution: unknown
+  failures: unknown
+

assets/amazon.yaml

@@ -29,3 +29,24 @@
   monthly_active_users: ''
   user_distribution: ''
   failures: ''
+- type: model
+  name: FalconLite2
+  organization: Amazon
+  description: FalconLite2 is a fine-tuned and quantized Falcon language model, capable of processing long (up to 24K tokens) input sequences.
+  created_date: 2023-08-08
+  url: https://huggingface.co/amazon/FalconLite2
+  model_card: https://huggingface.co/amazon/FalconLite2
+  modality: text; text
+  analysis: Evaluated against benchmarks that are specifically designed to assess the capabilities of LLMs in handling longer contexts.
+  size: 40B parameters (dense)
+  dependencies: [Falcon]
+  training_emissions: unknown
+  training_time: unknown
+  training_hardware: unknown
+  quality_control: ''
+  access: open
+  license: Apache 2.0
+  intended_uses: ''
+  prohibited_uses: ''
+  monitoring: ''
+  feedback: https://huggingface.co/amazon/FalconLite2/discussions

assets/anthropic.yaml

@@ -524,3 +524,24 @@
     where an incorrect answer would cause harm.
   monitoring: ''
   feedback: ''
+- type: model
+  name: Claude 2.1
+  organization: Anthropic
+  description: Claude 2.1 is an updated version of Claude 2, with an increased context window, less hallucination and tool use.
+  created_date: 2023-11-21
+  url: https://www.anthropic.com/index/claude-2-1
+  model_card: none
+  modality: text; text
+  analysis: Evaluated on open-ended conversation accuracy and long context question answering. In evaluations, Claude 2.1 demonstrated a 30% reduction in incorrect answers and a 3-4x lower rate of mistakenly concluding a document supports a particular claim.
+  size: unknown
+  dependencies: []
+  training_emissions: unknown
+  training_time: unknown
+  training_hardware: unknown
+  quality_control: ''
+  access: open
+  license: unknown
+  intended_uses: ''
+  prohibited_uses: ''
+  monitoring: ''
+  feedback: none

assets/baichuan.yaml

@@ -0,0 +1,22 @@
+---
+- type: model
+  name: Baichuan 2
+  organization: Baichuan Inc.
+  description: Baichuan 2 is a series of large-scale multilingual language models containing 7 billion and 13 billion parameters, trained from scratch, on 2.6 trillion tokens.
+  created_date: 2023-09-20
+  url: https://arxiv.org/pdf/2309.10305.pdf
+  model_card: none
+  modality: text; text
+  analysis: Evaluated on public benchmarks like MMLU, CMMLU, GSM8K, and HumanEval.
+  size: 13B parameters (dense)
+  dependencies: []
+  training_emissions: unknown
+  training_time: unknown
+  training_hardware: 1024 NVIDIA A800 GPUs
+  quality_control: ''
+  access: open
+  license: unknown
+  intended_uses: ''
+  prohibited_uses: ''
+  monitoring: none
+  feedback: https://huggingface.co/Xwin-LM/Xwin-LM-70B-V0.1/discussions

assets/character.yaml

@@ -0,0 +1 @@
+

assets/columbia.yaml

@@ -20,3 +20,24 @@
   prohibited_uses: ''
   monitoring: ''
   feedback: ''
+- type: model
+  name: Ferret
+  organization: Columbia
+  description: Ferret is a Multimodal Large Language Model (MLLM) capable of understanding spatial referring of any shape or granularity within an image and accurately grounding open-vocabulary descriptions.
+  created_date: 2023-10-11
+  url: https://arxiv.org/pdf/2310.07704.pdf
+  model_card: none
+  modality: image, text; image, text
+  analysis: Evaluated on the object hallucination benchmark and compared to GPT-4V.
+  size: 13B parameters
+  dependencies: [CLIP, Vicuna]
+  training_emissions: unknown
+  training_time: 2.5 to 5 days
+  training_hardware: 8 A100 GPUs
+  quality_control: ''
+  access: open
+  license: unknown
+  intended_uses: ''
+  prohibited_uses: ''
+  monitoring: none
+  feedback: none

assets/huggingface.yaml

@@ -24,3 +24,68 @@
   prohibited_uses: none
   monitoring: none
   feedback: none
+- type: model
+  name: Zephyr
+  organization: HuggingFace
+  description: Zephyr is a series of language models that are trained to act as helpful assistants.
+  created_date: 2023-10-11
+  url: https://huggingface.co/HuggingFaceH4/zephyr-7b-alpha
+  model_card: https://huggingface.co/HuggingFaceH4/zephyr-7b-alpha
+  modality: text; text
+  analysis: Evaluated on loss, rewards, logps, and logits rejected and chosen.
+  size: 7B parameters (dense)
+  dependencies: [Mistral]
+  training_emissions: unknown
+  training_time: unknown
+  training_hardware: unknown
+  quality_control: none
+  access: open
+  license: MIT
+  intended_uses: Educational and research purposes
+  prohibited_uses: none
+  monitoring: none
+  feedback: https://huggingface.co/HuggingFaceH4/zephyr-7b-alpha/discussions
+- type: model
+  name: IDEFICS
+  organization: HuggingFace
+  description: IDEFICS is an open-access visual language model, based on Flamingo.
+  created_date: 2023-08-22
+  url: https://huggingface.co/blog/idefics
+  model_card: https://huggingface.co/HuggingFaceM4/idefics-80b-instruct
+  modality: image, text; text
+  analysis: Evaluated in comparison to Flamingo and OpenFlamingo on standard benchmarks.
+  size: 80B parameters (dense)
+  dependencies: [OBELICS, Wikipedia, LAION-5B, PMD]
+  training_emissions: unknown
+  training_time: unknown
+  training_hardware: unknown
+  quality_control: none
+  access: open
+  license:
+    explanation: Can be found at https://huggingface.co/HuggingFaceM4/idefics-80b-instruct#license
+    value: custom
+  intended_uses: Educational and research purposes
+  prohibited_uses: none
+  monitoring: none
+  feedback: https://huggingface.co/HuggingFaceM4/idefics-80b-instruct/discussions
+- type: dataset
+  name: OBELICS
+  organization: HuggingFace
+  description: OBELICS is a dataset consisting of 141 million interleaved image-text documents scraped from the web and contains 353 million images.
+  created_date: 2023-08-22
+  url: https://huggingface.co/blog/idefics
+  datasheet: https://huggingface.co/datasets/HuggingFaceM4/OBELICS
+  modality: image, text
+  size: 115B tokens
+  sample: []
+  analysis: Subset of training dataset evaluated for bias using Data Measurements Tool.
+  dependencies: []
+  included: ''
+  excluded: All images for which creators explicitly requested opt-out of AI training.
+  quality_control: Sexual and violent content still present in OBELICS even after filtering.
+  access: open
+  license: CC-BY-4.0
+  intended_uses: ''
+  prohibited_uses: ''
+  monitoring: ''
+  feedback: https://huggingface.co/datasets/HuggingFaceM4/OBELICS/discussions

assets/microsoft.yaml

@@ -763,3 +763,68 @@
   prohibited_uses: ''
   monitoring: ''
   feedback: https://huggingface.co/WizardLM/WizardCoder-Python-34B-V1.0/discussions
+- type: dataset
+  name: OpenOrca
+  organization: Microsoft
+  description: The OpenOrca dataset is a collection of augmented FLAN Collection data. Currently ~1M GPT-4 completions, and ~3.2M GPT-3.5 completions. It is tabularized in alignment with the distributions presented in the ORCA paper and currently represents a partial completion of the full intended dataset, with ongoing generation to expand its scope.
+  created_date: 2023-06-05
+  url: https://huggingface.co/datasets/Open-Orca/OpenOrca
+  datasheet: https://huggingface.co/datasets/Open-Orca/OpenOrca
+  modality: text
+  size: 4.5M text queries
+  sample: []
+  analysis: Models trained on OpenOrca compared to GPT-series on language benchmarks.
+  dependencies: [GPT-3.5, GPT-4, Flan Collection]
+  included: ''
+  excluded: ''
+  quality_control: ''
+  access: open
+  license: MIT
+  intended_uses: training and evaluation in the field of natural language processing.
+  prohibited_uses: none
+  monitoring: ''
+  feedback: none
+- type: model
+  name: LlongOrca
+  organization: Microsoft
+  description: LlongOrca is an attempt to make OpenOrca able to function in a Llong context.
+  created_date: 2023-08-01
+  url: https://huggingface.co/Open-Orca/LlongOrca-7B-16k
+  model_card: https://huggingface.co/Open-Orca/LlongOrca-7B-16k
+  modality: text; text
+  analysis: LlongOrca evaluated on BigBench-Hard and AGIEval results.
+  size: 7B parameters (dense)
+  dependencies: [OpenOrca, LLongMA-2]
+  training_emissions: unknown
+  training_time: 37 hours
+  training_hardware: 8x A6000-48GB (first-gen) GPUs 
+  quality_control: ''
+  access: open
+  license: LLaMA2
+  intended_uses: training and evaluation in the field of natural language processing.
+  prohibited_uses: none
+  monitoring: ''
+  feedback: https://huggingface.co/Open-Orca/LlongOrca-7B-16k/discussions
+- type: model
+  name: Phi-1.5
+  organization: Microsoft
+  description: Phi-1.5 is a large language transformer model.
+  created_date: 2023-09-11
+  url: https://arxiv.org/pdf/2309.05463.pdf
+  model_card: https://huggingface.co/microsoft/phi-1_5
+  modality: text; text
+  analysis: Evaluated on common sense reasoning, language understanding, and multi-step reasoning compared to other SOTA language models.
+  size: 1.3B parameters (dense)
+  dependencies: [phi-1]
+  training_emissions: unknown
+  training_time: 8 days
+  training_hardware: 32 A100-40G GPUs
+  quality_control: generic web-crawl data is removed from dataset.
+  access: open
+  license: 
+    explanation: can be found via the license tab at top of https://huggingface.co/microsoft/phi-1_5
+    value: microsoft research license
+  intended_uses: Phi-1.5 is best suited for answering prompts using the QA format, the chat format, and the code format.
+  prohibited_uses: ''
+  monitoring: none
+  feedback: https://huggingface.co/microsoft/phi-1_5/discussions

assets/mila.yaml

@@ -0,0 +1,64 @@
+---
+- type: dataset
+  name: ToyMix
+  organization: Mila - Quebec AI Institute
+  description: ToyMix is the smallest dataset of three extensive and meticulously curated multi-label datasets that cover nearly 100 million molecules and over 3000 sparsely defined tasks. 
+  created_date: 2023-10-09
+  url: https://arxiv.org/pdf/2310.04292.pdf
+  datasheet: none
+  modality: molecules, tasks
+  size: 13B labels of quantum and biological nature.
+  sample: []
+  analysis: Models of size 150k parameters trained on ToyMix and compared to models trained on its dependencies across GNN baselines.
+  dependencies: [QM9, TOX21, ZINC12K]
+  included: ''
+  excluded: ''
+  quality_control: ''
+  access: open
+  license: CC BY-NC-SA 4.0
+  intended_uses: The datasets are intended to be used in an academic setting for training molecular GNNs with orders of magnitude more parameters than current large models. Further, the ToyMix dataset is intended to be used in a multi-task setting, meaning that a single model should be trained to predict them simultaneously.
+  prohibited_uses: none
+  monitoring: none
+  feedback: none
+- type: dataset
+  name: LargeMix
+  organization: Mila - Quebec AI Institute
+  description: LargeMix is the middle-sized dataset of three extensive and meticulously curated multi-label datasets that cover nearly 100 million molecules and over 3000 sparsely defined tasks. 
+  created_date: 2023-10-09
+  url: https://arxiv.org/pdf/2310.04292.pdf
+  datasheet: none
+  modality: molecules, tasks
+  size: 13B labels of quantum and biological nature.
+  sample: []
+  analysis: Models of size between 4M and 6M parameters trained for 200 epochs on LargeMix and compared to models trained on its dependencies across GNN baselines.
+  dependencies: [L1000 VCAP, L1000 MCF7, PCBA1328, PCQM4M_G25_N4]
+  included: ''
+  excluded: ''
+  quality_control: ''
+  access: open
+  license: CC BY-NC-SA 4.0
+  intended_uses: The datasets are intended to be used in an academic setting for training molecular GNNs with orders of magnitude more parameters than current large models. Further, the LargeMix dataset is intended to be used in a multi-task setting, meaning that a single model should be trained to predict them simultaneously.
+  prohibited_uses: none
+  monitoring: none
+  feedback: none
+- type: dataset
+  name: UltraLarge
+  organization: Mila - Quebec AI Institute
+  description: UltraLarge is the largest dataset of three extensive and meticulously curated multi-label datasets that cover nearly 100 million molecules and over 3000 sparsely defined tasks. 
+  created_date: 2023-10-09
+  url: https://arxiv.org/pdf/2310.04292.pdf
+  datasheet: none
+  modality: molecules, tasks
+  size: 13B labels of quantum and biological nature.
+  sample: []
+  analysis: Models of size between 4M and 6M parameters trained for 50 epochs on UltraLarge and compared to models trained on its dependencies across GNN baselines.
+  dependencies: [PM6_83M]
+  included: ''
+  excluded: ''
+  quality_control: ''
+  access: open
+  license: CC BY-NC-SA 4.0
+  intended_uses: The datasets are intended to be used in an academic setting for training molecular GNNs with orders of magnitude more parameters than current large models. 
+  prohibited_uses: none
+  monitoring: none
+  feedback: none