# TensorWave Training API

**Canonical:** https://apis.io/apis/tensorwave/tensorwave-training-api/  
**Provider:** TensorWave — https://apis.io/providers/tensorwave/  
**Base URL:** https://gpt-oss.cray-lm.com  
**Documentation:** https://www.scalarlm.com/docs/

TensorWave Training API is one of 5 APIs that [TensorWave](https://apis.io/providers/tensorwave/) publishes on the [APIs.io](https://apis.io/) network, described by a machine-readable OpenAPI specification. Tagged areas include Training. The published artifact set on APIs.io includes an OpenAPI specification, API documentation, an API reference, a getting-started guide, and authentication docs.

Megatron-LM training jobs, chunked dataset upload, checkpoints and Hugging Face publishing.

## Operations (21)

| Method | Path | Summary |
|---|---|---|
| POST | `/v1/megatron/train` | Submit a training job |
| POST | `/v1/megatron/upload/init` | Begin a chunked training-data upload |
| POST | `/v1/megatron/upload/chunk` | Upload one chunk of training data |
| POST | `/v1/megatron/upload/finalize` | Finalize a chunked upload and launch training |
| GET | `/v1/megatron/train/{job_hash}` | Get training job info |
| GET | `/v1/megatron/train/{job_hash}/dataset` | Page through a job's training dataset |
| GET | `/v1/megatron/train/{job_hash}/dataset/download` | Download a job's dataset.jsonlines |
| GET | `/v1/megatron/train/{job_hash}/checkpoints` | List checkpoints written by a training job |
| POST | `/v1/megatron/train/{job_hash}/publish` | Publish a checkpoint to Hugging Face Hub |
| GET | `/v1/megatron/train/{job_hash}/publish/status` | Latest publish status for a training job |
| POST | `/v1/megatron/train/{job_hash}/publish/cancel` | Cancel an in-flight publish job |
| GET | `/v1/megatron/train/{job_hash}/publish/logs` | Tail the publish job logs |
| GET | `/v1/megatron/train/logs/{model_name}` | Stream training logs for a model |
| POST | `/v1/megatron/cancel/{job_hash}` | Cancel a training job |
| POST | `/v1/megatron/restart/{job_hash}` | Restart a training job |
| POST | `/v1/megatron/delete/{job_hash}` | Delete a training job |
| GET | `/v1/megatron/list_models` | List models deployed in this ScalarLM instance |
| GET | `/v1/megatron/squeue` | Slurm queue as seen by the training API |
| GET | `/v1/megatron/gpu_count` | Number of GPUs visible to the deployment |
| GET | `/v1/megatron/node_count` | Number of nodes visible to the deployment |
| GET | `/v1/megatron/endpoints` | List routes registered on the training router |

## Machine-readable artifacts (9)

- **OpenAPI** — https://raw.githubusercontent.com/api-evangelist/tensorwave/refs/heads/main/openapi/tensorwave-training-api-openapi.yml
- **Documentation** — https://www.scalarlm.com/docs/
- **APIReference** — https://www.scalarlm.com/inference/
- **GettingStarted** — https://www.scalarlm.com/quick-start/
- **SourceCode** — https://github.com/tensorwavecloud/ScalarLM
- **Authentication** — https://raw.githubusercontent.com/api-evangelist/tensorwave/refs/heads/main/authentication/tensorwave-authentication.yml
- **Conventions** — https://raw.githubusercontent.com/api-evangelist/tensorwave/refs/heads/main/conventions/tensorwave-conventions.yml
- **ErrorCatalog** — https://raw.githubusercontent.com/api-evangelist/tensorwave/refs/heads/main/errors/tensorwave-problem-types.yml
- **DataModel** — https://raw.githubusercontent.com/api-evangelist/tensorwave/refs/heads/main/data-model/tensorwave-data-model.yml

## Other TensorWave APIs (4)

- [TensorWave Generate API](https://apis.io/apis/tensorwave/tensorwave-generate-api/)
- [TensorWave Health API](https://apis.io/apis/tensorwave/tensorwave-health-api/)
- [TensorWave Open AI API](https://apis.io/apis/tensorwave/tensorwave-openai-api/)
- [TensorWave Slurm API](https://apis.io/apis/tensorwave/tensorwave-slurm-api/)

## Tags

Training

---

Profiled by [API Evangelist](https://apievangelist.com) and published on [APIs.io](https://apis.io/apis/tensorwave/tensorwave-training-api/). The API's provider profile, Kin Score and agent-readiness rating are at https://apis.io/providers/tensorwave/.