Pretrained weights for ProphetNet. ProphetNet is a new pre-trained language model for sequence-to-sequence learning with a novel self-supervised objective called future n-gram prediction.
Runs On
What it takes to serve prophetnet-large-uncased (391M 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 | 0.8 GB | 0.9 GB | 1x MI300X (192 GB) Vultr |
$1.85 | 1x H100 $1.99 · 1x MI325X $2.00 |
| 8-bit | 0.4 GB | 0.5 GB | 1x MI300X (192 GB) Vultr |
$1.85 | 1x H100 $1.99 · 1x MI325X $2.00 |
| 4-bit | 0.2 GB | 0.2 GB | 1x MI300X (192 GB) Vultr |
$1.85 | 1x H100 $1.99 · 1x 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, read Sep 18, 2026.
Model Card
Pretrained weights for ProphetNet. ProphetNet is a new pre-trained language model for sequence-to-sequence learning with a novel self-supervised objective called future n-gram prediction. ProphetNet is able to predict more future tokens with a n-stream decoder. The original implementation is Fairseq version at github repo. This pre-trained model can be fine-tuned on sequence-to-sequence tasks. The model could e.g. be trained on headline generation as follows
Excerpt from the card by Microsoft.
Configuration
- Architecture
- ProphetNetForConditionalGeneration
- Context length (tokens)
- 512
- Hidden size
- 1,024
- Vocabulary size
- 30,522
- Model type
- prophetnet
Identity and Version
- Repository
- microsoft/prophetnet-large-uncased
- Publisher
- Microsoft
- Task
- Not stated by the source
- Modality
- Other
- Library
- transformers
- Parameters
- 391M parameters
- Languages
- en
- Revision
- c5b84da76e7f132c85b5b361e508145eaf2c24cd
- First published
- 2022-03-02
- Last updated
- 2023-04-27
Files and Weights
11 files, 5.1 GB in total. The weights are 3 files totalling 5.1 GB in bin, ot, safetensors.
Every file
| File | Type | Size | SHA-256 |
|---|---|---|---|
| model.safetensors | Weights | 1.6 GB | 060fb9eca2c5 |
| pytorch_model.bin | Weights | 1.6 GB | cedef47113e2 |
| rust_model.ot | Weights | 1.9 GB | fd416c555453 |
| config.json | Configuration | 1.4 KB | — |
| generation_config.json | Configuration | 293 B | — |
| generation_config_for_summarization.json | Configuration | 293 B | — |
| special_tokens_map.json | Configuration | 90 B | — |
| README.md | Documentation | 1.7 KB | — |
| .gitattributes | Repository | 399 B | — |
| prophetnet.tokenizer | Tokenizer | 231.5 KB | — |
| tokenizer_config.json | Tokenizer | 141 B | — |
License and Download
- License
- Not stated by the source
- Access
- Open weights, no gate
- Download size
- 5.1 GB
Released by Microsoft through its official repository on Hugging Face.
Built From
- Described by arXiv:2001.04063
Memory Requirements
| Precision | Weights in memory |
|---|---|
| As published | 5.1 GB |
| 16-bit | 0.8 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.
Questions About prophetnet-large-uncased
How much GPU memory does prophetnet-large-uncased need?
About 0.9 GB at 16-bit and 0.2 GB at 4-bit: the weights (391M parameters) plus a working margin. A long context needs more.
What is the cheapest GPU to run prophetnet-large-uncased 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.
What is prophetnet-large-uncased's context length?
512 tokens, from the maximum position embeddings in its published configuration.