Megatron-LM
E760435
Megatron-LM is a large-scale language model training framework developed by NVIDIA, designed to efficiently train massive transformer models through model, tensor, and pipeline parallelism.
All labels observed (1)
| Label | Occurrences |
|---|---|
| Megatron-LM canonical | 2 |
How this entity was disambiguated
This entity first appeared as the object of triple T8823605 — resolving that mention is where its identity was fixed. The disambiguator weighed these candidate entities and picked the highlighted one (or “None”, minting a new entity). This is how homonymy is resolved: the same surface form can point to different entities.
NED1
Entity disambiguation (via context triple)
gpt-5-mini-2025-08-07
Target entity: Megatron-LM Context triple: [NCCL, usedBy, Megatron-LM]
-
A.
LLaMA
LLaMA is a family of large language models developed by Meta AI, designed for efficient training and inference across a range of natural language processing tasks.
-
B.
GPT-Neo
GPT-Neo is an open-source family of autoregressive language models developed by EleutherAI as a free alternative to OpenAI’s GPT-3.
-
C.
PaLM 2
PaLM 2 is a large-scale language model developed by Google, known for powering various AI features across Google products before being succeeded by the Gemini family of models.
-
D.
GPT-2
GPT-2 is a large transformer-based language model known for generating coherent, human-like text and sparking widespread discussion about the implications of advanced AI text generation.
-
E.
GPT-3
GPT-3 is a large-scale autoregressive language model known for generating human-like text and performing a wide range of natural language tasks with minimal fine-tuning.
- F. None of above. chosen
- G. Unsure - the case is ambiguous/there is not enough information to decide.
NED2
Entity disambiguation (via description)
gpt-5-mini-2025-08-07
Target entity: Megatron-LM Target entity description: Megatron-LM is a large-scale language model training framework developed by NVIDIA, designed to efficiently train massive transformer models through model, tensor, and pipeline parallelism.
-
A.
LLaMA
LLaMA is a family of large language models developed by Meta AI, designed for efficient training and inference across a range of natural language processing tasks.
-
B.
GPT-Neo
GPT-Neo is an open-source family of autoregressive language models developed by EleutherAI as a free alternative to OpenAI’s GPT-3.
-
C.
PaLM 2
PaLM 2 is a large-scale language model developed by Google, known for powering various AI features across Google products before being succeeded by the Gemini family of models.
-
D.
GPT-2
GPT-2 is a large transformer-based language model known for generating coherent, human-like text and sparking widespread discussion about the implications of advanced AI text generation.
-
E.
GPT-3
GPT-3 is a large-scale autoregressive language model known for generating human-like text and performing a wide range of natural language tasks with minimal fine-tuning.
- F. None of above. chosen
Statements (49)
| Predicate | Object |
|---|---|
| instanceOf |
large-scale language model training framework
ⓘ
open-source software project ⓘ |
| basedOn | PyTorch ⓘ |
| developer |
NVIDIA
ⓘ
linked to:
NVIDIA Corporation
|
| domain |
large-scale deep learning
ⓘ
natural language processing ⓘ |
| hasFeature |
configurable parallelism degrees
ⓘ
fused attention kernels ⓘ fused bias and activation operations ⓘ fused layer normalization ⓘ highly optimized fused CUDA kernels ⓘ scalable tokenizer and dataset tools ⓘ scripted training pipelines ⓘ |
| implements |
3D parallelism
ⓘ
activation checkpointing ⓘ data parallelism ⓘ distributed training ⓘ gradient checkpointing ⓘ mixed precision training ⓘ model parallelism ⓘ optimizer state sharding ⓘ pipeline parallelism ⓘ tensor parallelism ⓘ |
| influenced |
Megatron-DeepSpeed
ⓘ
linked to:
DeepSpeed
|
| license | Apache License 2.0 ⓘ |
| optimizedFor |
NVIDIA GPUs
ⓘ
linked to:
GPU
multi-GPU systems ⓘ multi-node clusters ⓘ |
| programmingLanguage | Python ⓘ |
| repository | https://github.com/NVIDIA/Megatron-LM ⓘ |
| supports |
BERT-style language models
ⓘ
BF16 precision ⓘ FP16 precision ⓘ GPT-style language models ⓘ T5-style models ⓘ ZeRO-style optimizer partitioning ⓘ causal language modeling ⓘ checkpoints for large models ⓘ curriculum learning setups ⓘ decoder-only transformer architectures ⓘ distributed data loading ⓘ encoder-decoder transformer architectures ⓘ masked language modeling ⓘ sequence-to-sequence training ⓘ transformer models ⓘ |
| usedFor |
scaling transformer models to hundreds of billions of parameters
ⓘ
training very large language models ⓘ |
| usedIn | NVIDIA large language model research ⓘ |
| uses |
NVIDIA NCCL
ⓘ
linked to:
NCCL
|
How these facts were elicited
The pipeline generated the facts above by prompting gpt-5.1 with this entity's name + description and the instruction below.
Instruction
You are a knowledge base construction expert. Given a subject entity and a description of it, return factual statements that you know for the subject as a JSON list of dictionaries(triples), where keys must be "subject", "predicate" and "object". The number of facts may be very high, between 25 to 50 or more, for very popular subjects. For less popular subjects, the number of facts can be very low, like 5 or 10. # Requirements - If you don't know the subject at all, return an empty list. - If the subject is not a named entity, return an empty list. - Include at least one triple where predicate is "instanceOf". - Do not get too wordy. - Separate several objects into multiple triples with one object.
Input
Subject: Megatron-LM Description of subject: Megatron-LM is a large-scale language model training framework developed by NVIDIA, designed to efficiently train massive transformer models through model, tensor, and pipeline parallelism.
Referenced by (2)
Full triples — surface form annotated when it differs from this entity's canonical label.