From DeepSpeed to FSDP and Back Again with Hugging Face Accelerate
•
58
ContextualAI
company
Verified
AI & ML interests
Best-in-class enterprise-grade components for context engineering.
Recent Activity
View all activity
Articles
Family of instruction-following multilingual rerankers on the cost/performance Pareto frontier across public and customer benchmarks
-
ContextualAI/ctxl-rerank-v2-instruct-multilingual-6b
Text Ranking • 7B • Updated • 99 • 13 -
ContextualAI/ctxl-rerank-v2-instruct-multilingual-6b-nvfp4
Text Ranking • 4B • Updated • 76 • 2 -
ContextualAI/ctxl-rerank-v2-instruct-multilingual-2b
Text Ranking • 3B • Updated • 296 • 1 -
ContextualAI/ctxl-rerank-v2-instruct-multilingual-2b-nvfp4
Text Ranking • 2B • Updated • 17
Models & Datasets for LMunit evaluation
Archangel is a suite of human feedback-aligned LLMs, released as part of the Human-Aware Loss Functions (HALOs) project by Ethayarajh et al. (2024).
-
ContextualAI/archangel_sft_llama7b
Text Generation • 7B • Updated • 10 • 1 -
ContextualAI/archangel_kto_llama30b
Text Generation • 33B • Updated • 2 • 2 -
ContextualAI/archangel_sft-kto_llama13b
Text Generation • 13B • Updated • 1.16k • 3 -
ContextualAI/archangel_sft-kto_llama30b
Text Generation • 33B • Updated • 4 • 2
Data and Models for the paper "Anchored Preference Optimization and Contrastive Revisions: Addressing Underspecification in Alignment"
-
ContextualAI/ultrafeedback_clair_32k
Viewer • Updated • 29.1k • 137 • 5 -
ContextualAI/ultrafeedback_rlaif_32k
Viewer • Updated • 31.8k • 18 • 1 -
ContextualAI/ultrafeedback_rlaif-offpolicy_32k
Viewer • Updated • 29.1k • 14 • 1 -
ContextualAI/ultrafeedback_stronger-preferred_32k
Viewer • Updated • 29.1k • 14 • 1
Family of instruction-following multilingual rerankers on the cost/performance Pareto frontier across public and customer benchmarks
-
ContextualAI/ctxl-rerank-v2-instruct-multilingual-6b
Text Ranking • 7B • Updated • 99 • 13 -
ContextualAI/ctxl-rerank-v2-instruct-multilingual-6b-nvfp4
Text Ranking • 4B • Updated • 76 • 2 -
ContextualAI/ctxl-rerank-v2-instruct-multilingual-2b
Text Ranking • 3B • Updated • 296 • 1 -
ContextualAI/ctxl-rerank-v2-instruct-multilingual-2b-nvfp4
Text Ranking • 2B • Updated • 17
Instruction-following retrieval evaluation datasets
Models & Datasets for LMunit evaluation
Text-to-SQL models
Archangel is a suite of human feedback-aligned LLMs, released as part of the Human-Aware Loss Functions (HALOs) project by Ethayarajh et al. (2024).
-
ContextualAI/archangel_sft_llama7b
Text Generation • 7B • Updated • 10 • 1 -
ContextualAI/archangel_kto_llama30b
Text Generation • 33B • Updated • 2 • 2 -
ContextualAI/archangel_sft-kto_llama13b
Text Generation • 13B • Updated • 1.16k • 3 -
ContextualAI/archangel_sft-kto_llama30b
Text Generation • 33B • Updated • 4 • 2
Aligned models based on Zephyr-SFT from Table 2 and 3 in the KTO paper by Ethayarajh et al. (2024) (https://arxiv.org/pdf/2402.01306).
Data and Models for the paper "Anchored Preference Optimization and Contrastive Revisions: Addressing Underspecification in Alignment"
-
ContextualAI/ultrafeedback_clair_32k
Viewer • Updated • 29.1k • 137 • 5 -
ContextualAI/ultrafeedback_rlaif_32k
Viewer • Updated • 31.8k • 18 • 1 -
ContextualAI/ultrafeedback_rlaif-offpolicy_32k
Viewer • Updated • 29.1k • 14 • 1 -
ContextualAI/ultrafeedback_stronger-preferred_32k
Viewer • Updated • 29.1k • 14 • 1