Sync contents

docs/api/samplingclient.md

@@ -110,3 +110,14 @@ async def compute_logprobs_async(
 ```
 
 Async version of compute_logprobs.
+
+#### `get_tokenizer`
+
+```python
+def get_tokenizer() -> PreTrainedTokenizer
+```
+
+Get the tokenizer for the current model.
+
+Returns:
+- `PreTrainedTokenizer` compatible with the model

docs/api/trainingclient.md

@@ -176,6 +176,11 @@ def optim_step(
 
 Update model parameters using Adam optimizer.
 
+The Adam optimizer used by tinker is identical
+to [torch.optim.AdamW](https://docs.pytorch.org/docs/stable/generated/torch.optim.AdamW.html).
+Note that unlike PyTorch, Tinker's default weight decay value is 0.0 (no weight decay).
+
+
 Args:
 - `adam_params`: Adam optimizer parameters (learning_rate, betas, eps, weight_decay)
 
@@ -212,13 +217,17 @@ Async version of optim_step.
 #### `save_state`
 
 ```python
-def save_state(name: str) -> APIFuture[types.SaveWeightsResponse]
+def save_state(
+        name: str,
+        ttl_seconds: int | None = None
+) -> APIFuture[types.SaveWeightsResponse]
 ```
 
 Save model weights to persistent storage.
 
 Args:
 - `name`: Name for the saved checkpoint
+- `ttl_seconds`: Optional TTL in seconds for the checkpoint (None = never expires)
 
 Returns:
 - `APIFuture` containing the save response with checkpoint path
@@ -234,7 +243,10 @@ print(f"Saved to: {result.path}")
 #### `save_state_async`
 
 ```python
-async def save_state_async(name: str) -> APIFuture[types.SaveWeightsResponse]
+async def save_state_async(
+        name: str,
+        ttl_seconds: int | None = None
+) -> APIFuture[types.SaveWeightsResponse]
 ```
 
 Async version of save_state.
@@ -310,13 +322,16 @@ Async version of load_state_with_optimizer.
 
 ```python
 def save_weights_for_sampler(
-        name: str) -> APIFuture[types.SaveWeightsForSamplerResponse]
+    name: str,
+    ttl_seconds: int | None = None
+) -> APIFuture[types.SaveWeightsForSamplerResponse]
 ```
 
 Save model weights for use with a SamplingClient.
 
 Args:
 - `name`: Name for the saved sampler weights
+- `ttl_seconds`: Optional TTL in seconds for the checkpoint (None = never expires)
 
 Returns:
 - `APIFuture` containing the save response with sampler path
@@ -338,7 +353,9 @@ sampling_client = service_client.create_sampling_client(
 
 ```python
 async def save_weights_for_sampler_async(
-        name: str) -> APIFuture[types.SaveWeightsForSamplerResponse]
+    name: str,
+    ttl_seconds: int | None = None
+) -> APIFuture[types.SaveWeightsForSamplerResponse]
 ```
 
 Async version of save_weights_for_sampler.

docs/api/types.md

@@ -26,7 +26,7 @@ Weight decay for the optimizer. Uses decoupled weight decay.
 
 #### `grad_clip_norm`
 
-Gradient clip norm for the optimizer. 0.0 means no clipping.
+Maximum global gradient norm. If the global gradient norm is greater than this value, it will be clipped to this value. 0.0 means no clipping.
 
 ## `SupportedModel` Objects
 
@@ -159,6 +159,10 @@ The size of the checkpoint in bytes
 
 Whether the checkpoint is publicly accessible
 
+#### `expires_at`
+
+When this checkpoint expires (None = never expires)
+
 ## `ParsedCheckpointTinkerPath` Objects
 
 ```python
@@ -725,6 +729,10 @@ class SaveWeightsRequest(StrictBase)
 
 A file/directory name for the weights
 
+#### `ttl_seconds`
+
+TTL in seconds for this checkpoint (None = never expires)
+
 ## `LoraConfig` Objects
 
 ```python
@@ -834,6 +842,10 @@ class SaveWeightsForSamplerRequest(StrictBase)
 
 A file/directory name for the weights
 
+#### `ttl_seconds`
+
+TTL in seconds for this checkpoint (None = never expires)
+
 ## `SamplingParams` Objects
 
 ```python