Skip to content
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 1 addition & 1 deletion docs/core_concepts/model_tiers.md
Original file line number Diff line number Diff line change
Expand Up @@ -60,7 +60,7 @@ guide.
```{note}
**Version pinning.** For stable-only projects, pin a compatible range such as
`deeptab>=2.0,<3.0`. For projects that use experimental models, pin the exact
version (`deeptab==2.0.0`), since their APIs may change between releases.
version (`deeptab==2.1.0`), since their APIs may change between releases.
```

## Next Steps
Expand Down
2 changes: 1 addition & 1 deletion docs/getting_started/migration.md
Original file line number Diff line number Diff line change
Expand Up @@ -260,7 +260,7 @@ from deeptab.models.experimental import ModernNCAClassifier, TangosRegressor, Tr

```{warning}
Experimental models may change in minor releases. Pin an exact version such as
`deeptab==2.0.0` if you depend on one in production.
`deeptab==2.1.0` if you depend on one in production.
```

## What did not change
Expand Down
2 changes: 1 addition & 1 deletion docs/model_zoo/comparison_tables.md
Original file line number Diff line number Diff line change
Expand Up @@ -127,7 +127,7 @@ The table below gives practical guidance on whether each model trains comfortabl
The "CPU comfort zone" is where training at default config finishes in reasonable wall-clock time on a modern CPU. Beyond it, a CUDA, MPS, or similar accelerator provides meaningful speedup.

```{tip}
**Apple Silicon (MPS):** All models run on MPS via PyTorch's MPS backend. Set `accelerator="mps"` in `TrainerConfig`. MPS provides meaningful speedup for most models except those with Mamba CUDA kernels, which fall back to CPU on MPS unless a dedicated MPS implementation is available.
**Apple Silicon (MPS):** All models run on MPS via PyTorch's MPS backend. Pass `accelerator="mps"` to `fit()`. MPS provides meaningful speedup for most models except those with Mamba CUDA kernels, which fall back to CPU on MPS unless a dedicated MPS implementation is available.
```

```{note}
Expand Down
2 changes: 1 addition & 1 deletion docs/tutorials/experimental.md
Original file line number Diff line number Diff line change
Expand Up @@ -33,7 +33,7 @@ DeepTab sorts every model into one of two tiers. The tier is a contract about AP
| ------------------- | -------------------------------------------------- | ------------------------------------------------------- |
| **Import path** | `deeptab.models.experimental` | `deeptab.models` |
| **API stability** | May change without a deprecation cycle | Frozen under semantic versioning |
| **Recommended pin** | Exact version (`deeptab==2.0.0`) | Range (`deeptab>=2.0,<3.0`) |
| **Recommended pin** | Exact version (`deeptab==2.1.0`) | Range (`deeptab>=2.0,<3.0`) |
| **Best for** | Evaluating recent architectures, research feedback | Production, long-running baselines, reproducible suites |

Before an experimental model graduates to the stable zoo it has to clear a documented bar: a conventional public API, a model-zoo page with a limitations section, a runnable end-to-end example, working `save`/`load` with a prediction round-trip test, passing behavioural tests in CI, no open correctness bugs, and registration in the model registry. Until then, treat its defaults as provisional.
Expand Down
4 changes: 2 additions & 2 deletions docs/tutorials/notebooks/experimental.ipynb
Original file line number Diff line number Diff line change
Expand Up @@ -58,13 +58,13 @@
"| --- | --- | --- |\n",
"| **Import path** | `deeptab.models.experimental` | `deeptab.models` |\n",
"| **API stability** | May change without a deprecation cycle | Frozen under semantic versioning |\n",
"| **Recommended pin** | Exact version (`deeptab==1.8.0`) | Range (`deeptab>=1.8,<2.0`) |\n",
"| **Recommended pin** | Exact version (`deeptab==2.1.0`) | Range (`deeptab>=2.0,<3.0`) |\n",
"| **Best for** | Evaluating recent architectures, research feedback | Production, long-running baselines, reproducible suites |\n",
"\n",
"Before an experimental model graduates to the stable zoo it has to clear a documented bar: a conventional public API, a model-zoo page with a limitations section, a runnable end-to-end example, working `save`/`load` with a prediction round-trip test, passing behavioural tests in CI, no open correctness bugs, and registration in the model registry. Until then, treat its defaults as provisional.\n",
"\n",
"```{warning}\n",
"Pin the **exact** DeepTab version whenever experimental results go into a paper, a benchmark table, or anything you might need to reproduce later. A range such as `deeptab>=1.8` can silently pull a release that changes an experimental model's behaviour.\n",
"Pin the **exact** DeepTab version whenever experimental results go into a paper, a benchmark table, or anything you might need to reproduce later. A range such as `deeptab>=2.0` can silently pull a release that changes an experimental model's behaviour.\n",
"```"
]
},
Expand Down
Loading