# roberta-large-mnli by Facebook AI community: Open Model Source: https://savrn.com/models/roberta-large-mnli Markdown alternate of the page above; the site index is https://savrn.com/llms.txt --- ## Runs On What it takes to serve roberta-large-mnli (356M parameters): the memory its weights need at each precision, and the cheapest way to rent enough data-center GPUs to hold them. | Precision | Weights | Memory needed | Cheapest setup | Per hour | Also fits | | --- | --- | --- | --- | --- | --- | | 16-bit | 0.7 GB | 0.9 GB | 1x [MI300X](https://savrn.com/ai-index/pricing/gpus/mi300x) (192 GB) Vultr | $1.85 | [1x H100](https://savrn.com/ai-index/pricing/gpus/h100) $1.99 · [1x MI325X](https://savrn.com/ai-index/pricing/gpus/mi325x) $2.00 | | 8-bit | 0.4 GB | 0.4 GB | 1x [MI300X](https://savrn.com/ai-index/pricing/gpus/mi300x) (192 GB) Vultr | $1.85 | [1x H100](https://savrn.com/ai-index/pricing/gpus/h100) $1.99 · [1x MI325X](https://savrn.com/ai-index/pricing/gpus/mi325x) $2.00 | | 4-bit | 0.2 GB | 0.2 GB | 1x [MI300X](https://savrn.com/ai-index/pricing/gpus/mi300x) (192 GB) Vultr | $1.85 | [1x H100](https://savrn.com/ai-index/pricing/gpus/h100) $1.99 · [1x MI325X](https://savrn.com/ai-index/pricing/gpus/mi325x) $2.00 | Memory is the weights at that precision plus 20% for the runtime and a short context; a long context needs more. Prices are the lowest on-demand hourly rates in the [SAVRN Index](https://savrn.com/ai-index/pricing/gpus), read Oct 25, 2026. [roberta-large-mnli on every accelerator the SAVRN Index prices, at every precision](https://savrn.com/models/roberta-large-mnli/gpus) ## SAVRN's Notes on roberta-large-mnli This is RoBERTa large fine-tuned on MultiNLI: it decides whether one sentence entails, contradicts or is neutral toward another. For years it was the standard engine behind zero-shot classification pipelines, and it has 356M parameters. The MIT license allows commercial use. Newer NLI models in our catalog, especially the DeBERTa-v3 ones trained on several NLI datasets, are generally more accurate. Keep this one where an existing system depends on its behavior. SAVRN Research, 2026-09-18 ## Model Card By Facebook AI community, published under mit, revision 2a8f12d27941. ### Table of Contents - Model Details - How To Get Started With the Model - Uses - Risks, Limitations and Biases - Training - Evaluation - Environmental Impact - Technical Specifications - Citation Information - Model Card Authors ### Model Details Model Description: roberta-large-mnli is the [RoBERTa large model](https://huggingface.co/roberta-large) fine-tuned on the [Multi-Genre Natural Language Inference (MNLI)](https://huggingface.co/datasets/multi_nli) corpus. The model is a pretrained model on English language text using a masked language modeling (MLM) objective. - Developed by: See [GitHub Repo](https://github.com/facebookresearch/fairseq/tree/main/examples/roberta) for model developers - Model Type: Transformer-based language model - Language(s): English - License: MIT - Parent Model: This model is a fine-tuned version of the RoBERTa large model. Users should see the [RoBERTa large model card](https://huggingface.co/roberta-large) for relevant information. - Resources for more information: - [Research Paper](https://savrn.com/papers/roberta-a-robustly-optimized-bert-pretraining-approach) - [GitHub Repo](https://github.com/facebookresearch/fairseq/tree/main/examples/roberta) ### How to Get Started with the Model Use the code below to get started with the model. The model can be loaded with the zero-shot-classification pipeline like so: ``` from transformers import pipeline classifier = pipeline('zero-shot-classification', model='roberta-large-mnli') ``` [Read the full model card (1,214 words)](https://savrn.com/models/roberta-large-mnli/card) ## Configuration Architecture RobertaForSequenceClassification Context length (tokens) 514 Layers 24 Hidden size 1,024 Feed-forward size 4,096 Attention heads 16 Vocabulary size 50,265 Model type roberta ## Identity and Version Repository FacebookAI/roberta-large-mnli Publisher Facebook AI community Task Text classification Modality Text Library transformers Parameters 356M parameters Languages en Revision 2a8f12d27941090092df78e4ba6f0928eb5eac98 First published 2022-03-02 Last updated 2024-02-19 ## Files and Weights 11 files, 5.7 GB in total. The weights are 4 files totalling 5.7 GB in bin, h5, msgpack, safetensors. Weights4 files · 5.7 GB Configuration1 file · 688 B Tokenizer4 files · 2.7 MB Documentation1 file · 10.7 KB Repository1 file · 445 B Every file | File | Type | Size | SHA-256 | | --- | --- | --- | --- | | flax_model.msgpack | Weights | 1.4 GB | 3268be444292 | | model.safetensors | Weights | 1.4 GB | f4dbab1bceb1 | | pytorch_model.bin | Weights | 1.4 GB | 5a7be848f7dd | | tf_model.h5 | Weights | 1.4 GB | cf9b30ac6668 | | config.json | Configuration | 688 B | — | | README.md | Documentation | 10.7 KB | — | | .gitattributes | Repository | 445 B | — | | merges.txt | Tokenizer | 456.3 KB | — | | tokenizer.json | Tokenizer | 1.4 MB | — | | tokenizer_config.json | Tokenizer | 25 B | — | | vocab.json | Tokenizer | 898.8 KB | — | ## License and Download License mit Access Open weights, no gate Download size 5.7 GB [Download from Facebook AI community](https://huggingface.co/FacebookAI/roberta-large-mnli) Released by Facebook AI community through its official repository on Hugging Face. [Read the license](https://opensource.org/license/mit). ## Built From - Described by arXiv:1508.05326 - Described by [arXiv:1704.05426](https://savrn.com/papers/a-broad-coverage-challenge-corpus-for-sentence-understanding-through-inference) - Described by [arXiv:1804.07461](https://savrn.com/papers/glue-a-multi-task-benchmark-and-analysis-platform-for-natural-language-understanding) - Described by [arXiv:1806.02847](https://savrn.com/papers/a-simple-method-for-commonsense-reasoning) - Described by [arXiv:1809.05053](https://savrn.com/papers/xnli-evaluating-cross-lingual-sentence-representations) - Described by [arXiv:1907.11692](https://savrn.com/papers/roberta-a-robustly-optimized-bert-pretraining-approach) - Described by [arXiv:1910.09700](https://savrn.com/papers/quantifying-the-carbon-emissions-of-machine-learning) - Trained on (disclosed) bookcorpus - Trained on (disclosed) multi_nli - Trained on (disclosed) wikipedia ## Memory Requirements | Precision | Weights in memory | | --- | --- | | As published | 5.7 GB | | 16-bit | 0.7 GB | | 8-bit | 0.4 GB | | 4-bit | 0.2 GB | Weights only, from the published parameter count; the key-value cache and runtime add to this. ## Compare roberta-large-mnli - [bge-reranker-base vs roberta-large-mnli](https://savrn.com/models/compare/bge-reranker-base-vs-roberta-large-mnli) - [bge-reranker-v2-m3 vs roberta-large-mnli](https://savrn.com/models/compare/bge-reranker-v2-m3-vs-roberta-large-mnli) ## Questions About roberta-large-mnli ### How much GPU memory does roberta-large-mnli need? About 0.9 GB at 16-bit and 0.2 GB at 4-bit: the weights (356M parameters) plus a working margin. A long context needs more. ### What is the cheapest GPU to run roberta-large-mnli on? At 16-bit, 1x MI300X from $1.85 an hour; at 4-bit, 1x MI300X from $1.85 an hour, at the lowest on-demand prices the SAVRN Index lists. ### Can I use roberta-large-mnli commercially? Yes. roberta-large-mnli is released under MIT License. The MIT License is a short permissive license. It permits commercial use, modification and redistribution, provided the copyright notice and permission notice are included. ### What is roberta-large-mnli's context length? 514 tokens, from the maximum position embeddings in its published configuration. ## Similar Models Model · Text classification ### [klue-roberta-framing-3class](https://savrn.com/models/klue-roberta-framing-3class) [Joaram](https://savrn.com/model-publishers/likeramm) 같은 사실을 긍정적인 틀 또는 부정적인 틀로 제시하는 가치 프레이밍(valence framing)을 다룹니다. 예를 들어 "GDP 2.3% 성장"을 "견조한 성장세"로 쓰면 긍정, "성장 둔화 우려"로 쓰면 부정입니다. 코드: https://github.com/likeramm/econ-framing-bias - 방법: LLM 라벨을 소형 모델로 옮기는 지식 증류. klue/roberta-large를 파인튜닝했습니다. - 라벨: gpt-5.5가 라벨링 가이드라인 프롬프트(v2)로 판정한 결과 (OpenAI Batch API) - 학습 8,866 / 검증 1,900 / 테스트 1,901 (층화 분할, seed 42). 검증셋으로 모델을 고르고, 테스트셋으로 최종 보고 - 입력: 제목 [SEP] 본문 앞 500자, 최대 256 토큰 - 설정: lr 1e-5, batch 4 × gradient accumulation 2, 최대 6 epoch(조기 종료 patience 2), 클래스 가중 교차엔트로피, FP16, RTX 4060 테스트셋 (n=1,901, 정답 = gpt-5.5 라벨) 사람 라벨과의 일치도 (파일럿 50건, 학습 제외): 정확도 0.74, Cohen's κ 0.58 - 중립이 가장 어렵습니다. 사람과 LLM의 판단도 중립에서 가장 많이 엇갈립니다. - 사건의 좋고 나쁨에 끌리는 경향이 있습니다. "GDP 2.3% 성장… Open weights cc-by-sa-4.0 337M parameters 514 tokens [View model](https://savrn.com/models/klue-roberta-framing-3class) Model · Text classification ### [laya-multilingual-calculator-router](https://savrn.com/models/laya-multilingual-calculator-router) [Leo Borcherding](https://savrn.com/model-publishers/borcherding) A Laya decision model that answers one question before each agent step: does this step need a calculation? It also scores how much working the calculation takes (none, trivial, real working). Fine-tuned from convaiinnovations/laya (the multilingual checkpoint) on Borcherding/calculator-routing with Unsloth's FastDecisionModel. Ask the questions with the exact wording above; that is what the model was trained on. Held-out test split (1,200 rows, 600 per class). Test uses templates, names and entities that never appear in train, so it measures new phrasings of familiar kinds of step. Base Laya says "yes" to almost anything with a number in it. The tuned model fixes that: order IDs, versions… Open weights apache-2.0 322M parameters laya [View model](https://savrn.com/models/laya-multilingual-calculator-router) Model · Text classification ### [laya-multilingual](https://savrn.com/models/laya-multilingual) [Convai Innovations](https://savrn.com/model-publishers/convaiinnovations) Non-autoregressive System 1 decision model covering 100+ languages. Give it a state (text, email, ticket, or JSON) and typed questions; it returns typed answers with probabilities in a single forward pass. No text generation, so nothing to parse and nothing to hallucinate. Part of the Laya family — use this checkpoint for anything that is not English. Since laya 0.3.13 the default Router() keeps both english and this checkpoint resident, so a mixed workload no longer swaps checkpoints on every language change. For a server, load them up front so even the first request of each language is just a forward pass: router.attach("multilingual", agent) registers an Agent you already built, so a… Open weights apache-2.0 322M parameters transformers [View model](https://savrn.com/models/laya-multilingual) Model · Text classification ### [laya-coreai](https://savrn.com/models/laya-coreai) [Andrei](https://savrn.com/model-publishers/andyinqtr) Laya typed decisions on Apple Silicon, running on the Core AI runtime — the successor to Core ML. This is a.aimodel asset exported from via Apple's coreai-torch bridge. It outputs choice / score / noul probabilities (and RL action logits) with zero generated tokens and no PyTorch, Core ML, Transformers, or cloud API at inference time. macOS 27+ (Core AI runtime), Python 3.10+. Validated on M3 Max / macOS 27.2. Validate the download end-to-end (all three specializations, timing, contract checks): Snake demo with the model (terminal game, reuses the laya-coreml UI + safety shield; automatically uses the B3 asset when present for ~2x game throughput): ~3× faster per pass than the fastest Core… Open weights apache-2.0 322M parameters coreai [View model](https://savrn.com/models/laya-coreai) Model · Text classification ### [laya-pt-es-typed](https://savrn.com/models/laya-pt-es-typed) [Telepatia](https://savrn.com/model-publishers/telepatia-ai) This checkpoint fine-tunes convaiinnovations/laya-multilingual for native choice, score, and noul decisions in Portuguese and Spanish. It keeps the original 322M-parameter mmBERT architecture. It adds no inference component and does not generate text. It returns typed answers and probabilities in one forward pass. This is a text model. Inference takes a textual state plus typed questions. The second training stage used text decisions derived from public speech corpora, but this checkpoint does not accept audio by itself. The separate audio projector is not included. The official Laya SDK defines these primitives as follows: - choice: selects one key from a runtime-defined criteria object.… Open weights apache-2.0 322M parameters laya [View model](https://savrn.com/models/laya-pt-es-typed) Model · Text classification ### [laya-multilingual](https://savrn.com/models/laya-multilingual-2) [Scott Lamkin](https://savrn.com/model-publishers/scottrlamkin) Non-autoregressive System 1 decision model covering 100+ languages. Give it a state (text, email, ticket, or JSON) and typed questions; it returns typed answers with probabilities in a single forward pass. No text generation, so nothing to parse and nothing to hallucinate. Part of the Laya family — use this checkpoint for anything that is not English. The default Router() keeps both english and this checkpoint resident, so a mixed workload no longer swaps checkpoints on every language change. For a server, load them up front so even the first request of each language is just a forward pass: router.attach("multilingual", agent) registers an Agent you already built, so a process that loaded… Open weights apache-2.0 322M parameters transformers [View model](https://savrn.com/models/laya-multilingual-2) ## Facebook AI community [All models and datasets](https://savrn.com/model-publishers/facebookai) ## Versions - [2a8f12d27941](https://savrn.com/models/roberta-large-mnli/versions/2a8f12d27941) · current 2026-09-18 ## Explore More - [All text classification models](https://savrn.com/models/tasks/text-classification) - [All models under mit](https://savrn.com/models/licenses/mit) - [Model comparisons](https://savrn.com/models/comparisons) - [The model directory](https://savrn.com/models) - [Open model prices by host](https://savrn.com/ai-index/pricing/open-models) ## Source - Repository metadata, read 2026-10-20. - [Hugging Face record](https://huggingface.co/FacebookAI/roberta-large-mnli) - [How the hub is built](https://savrn.com/model-hub/methodology)