thinking-machines-lab/tinker
1
0
Fork 0
mirror of https://github.com/thinking-machines-lab/tinker.git synced 2026-04-19 12:58:01 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

Sync contents

Browse source
This commit is contained in:
Daniel Xu 2025-11-25 05:55:10 +00:00
parent 097e108902
commit 937c36e9b1
6 changed files with 109 additions and 106 deletions
Download patch file Download diff file Expand all files Collapse all files

45
docs/api/serviceclient.md
View file

@ -20,14 +20,17 @@ Args:
Example:
```python
# Near instant
client = ServiceClient()
# ^^^ near-instant
# Takes a moment as we initialize the model and assign resources
training_client = client.create_lora_training_client(base_model="Qwen/Qwen3-8B")
# ^^^ takes a moment as we initialize the model and assign resources
# Near-instant
sampling_client = client.create_sampling_client(base_model="Qwen/Qwen3-8B")
# ^^^ near-instant
# Near-instant
rest_client = client.create_rest_client()
# ^^^ near-instant
```
#### `get_server_capabilities`
@ -41,7 +44,7 @@ def get_server_capabilities() -> types.GetServerCapabilitiesResponse
Query the server's supported features and capabilities.
Returns:
GetServerCapabilitiesResponse with available models, features, and limits
- `GetServerCapabilitiesResponse` with available models, features, and limits
Example:
```python
@ -78,16 +81,16 @@ def create_lora_training_client(
Create a TrainingClient for LoRA fine-tuning.
Args:
base_model: Name of the base model to fine-tune (e.g., "Qwen/Qwen2.5-7B")
rank: LoRA rank controlling the size of adaptation matrices (default 32)
seed: Random seed for initialization. None means random seed.
train_mlp: Whether to train MLP layers (default True)
train_attn: Whether to train attention layers (default True)
train_unembed: Whether to train unembedding layers (default True)
user_metadata: Optional metadata to attach to the training run
- `base_model`: Name of the base model to fine-tune (e.g., "Qwen/Qwen2.5-7B")
- `rank`: LoRA rank controlling the size of adaptation matrices (default 32)
- `seed`: Random seed for initialization. None means random seed.
- `train_mlp`: Whether to train MLP layers (default True)
- `train_attn`: Whether to train attention layers (default True)
- `train_unembed`: Whether to train unembedding layers (default True)
- `user_metadata`: Optional metadata to attach to the training run
Returns:
TrainingClient configured for LoRA training
- `TrainingClient` configured for LoRA training
Example:
```python
@ -129,11 +132,11 @@ def create_training_client_from_state(
Create a TrainingClient from saved model weights.
Args:
path: Tinker path to saved weights (e.g., "tinker://run-id/weights/checkpoint-001")
user_metadata: Optional metadata to attach to the new training run
- `path`: Tinker path to saved weights (e.g., "tinker://run-id/weights/checkpoint-001")
- `user_metadata`: Optional metadata to attach to the new training run
Returns:
TrainingClient loaded with the specified weights
- `TrainingClient` loaded with the specified weights
Example:
```python
@ -168,12 +171,12 @@ def create_sampling_client(
Create a SamplingClient for text generation.
Args:
model_path: Path to saved model weights (e.g., "tinker://run-id/weights/checkpoint-001")
base_model: Name of base model to use (e.g., "Qwen/Qwen2.5-7B")
retry_config: Optional configuration for retrying failed requests
- `model_path`: Path to saved model weights (e.g., "tinker://run-id/weights/checkpoint-001")
- `base_model`: Name of base model to use (e.g., "Qwen/Qwen2.5-7B")
- `retry_config`: Optional configuration for retrying failed requests
Returns:
SamplingClient configured for text generation
- `SamplingClient` configured for text generation
Raises:
ValueError: If neither model_path nor base_model is provided
@ -216,7 +219,7 @@ The RestClient provides access to various REST endpoints for querying
model information, checkpoints, sessions, and managing checkpoint visibility.
Returns:
RestClient for accessing REST API endpoints
- `RestClient` for accessing REST API endpoints
Example:
```python