together 1.5.8__py3-none-any.whl → 1.5.13__py3-none-any.whl

together/cli/api/endpoints.py

@@ -82,7 +82,7 @@ def endpoints(ctx: click.Context) -> None:
 @click.option(
     "--model",
     required=True,
-    help="The model to deploy (e.g. mistralai/Mixtral-8x7B-Instruct-v0.1)",
+    help="The model to deploy (e.g. meta-llama/Llama-4-Scout-17B-16E-Instruct)",
 )
 @click.option(
     "--min-replicas",

together/cli/api/finetune.py

@@ -1,10 +1,10 @@
 from __future__ import annotations
 
 import json
+import re
 from datetime import datetime, timezone
 from textwrap import wrap
 from typing import Any, Literal
-import re
 
 import click
 from click.core import ParameterSource  # type: ignore[attr-defined]
@@ -13,17 +13,17 @@ from tabulate import tabulate
 
 from together import Together
 from together.cli.api.utils import BOOL_WITH_AUTO, INT_WITH_MAX
+from together.types.finetune import (
+    DownloadCheckpointType,
+    FinetuneEventType,
+    FinetuneTrainingLimits,
+)
 from together.utils import (
     finetune_price_to_dollars,
+    format_timestamp,
     log_warn,
     log_warn_once,
     parse_timestamp,
-    format_timestamp,
-)
-from together.types.finetune import (
-    DownloadCheckpointType,
-    FinetuneTrainingLimits,
-    FinetuneEventType,
 )
 
 
@@ -82,7 +82,7 @@ def fine_tuning(ctx: click.Context) -> None:
 @click.option(
     "--lr-scheduler-type",
     type=click.Choice(["linear", "cosine"]),
-    default="linear",
+    default="cosine",
     help="Learning rate scheduler type",
 )
 @click.option(
@@ -348,9 +348,9 @@ def list(ctx: click.Context) -> None:
                 "Model Output Name": "\n".join(wrap(i.output_name or "", width=30)),
                 "Status": i.status,
                 "Created At": i.created_at,
-                "Price": f"""${finetune_price_to_dollars(
-                    float(str(i.total_price))
-                )}""",  # convert to string for mypy typing
+                "Price": f"""${
+                    finetune_price_to_dollars(float(str(i.total_price)))
+                }""",  # convert to string for mypy typing
             }
         )
     table = tabulate(display_list, headers="keys", tablefmt="grid", showindex=True)

together/client.py

@@ -23,6 +23,7 @@ class Together:
     fine_tuning: resources.FineTuning
     rerank: resources.Rerank
     audio: resources.Audio
+    batches: resources.Batches
     code_interpreter: CodeInterpreter
 
     # client options
@@ -90,6 +91,7 @@ class Together:
         self.audio = resources.Audio(self.client)
         self.endpoints = resources.Endpoints(self.client)
         self.code_interpreter = CodeInterpreter(self.client)
+        self.batches = resources.Batches(self.client)
 
 
 class AsyncTogether:
@@ -102,7 +104,7 @@ class AsyncTogether:
     fine_tuning: resources.AsyncFineTuning
     rerank: resources.AsyncRerank
     code_interpreter: CodeInterpreter
-
+    batches: resources.AsyncBatches
     # client options
     client: TogetherClient
 
@@ -166,6 +168,7 @@ class AsyncTogether:
         self.fine_tuning = resources.AsyncFineTuning(self.client)
         self.rerank = resources.AsyncRerank(self.client)
         self.code_interpreter = CodeInterpreter(self.client)
+        self.batches = resources.AsyncBatches(self.client)
 
 
 Client = Together

together/resources/__init__.py

@@ -8,6 +8,7 @@ from together.resources.finetune import AsyncFineTuning, FineTuning
 from together.resources.images import AsyncImages, Images
 from together.resources.models import AsyncModels, Models
 from together.resources.rerank import AsyncRerank, Rerank
+from together.resources.batch import Batches, AsyncBatches
 
 
 __all__ = [
@@ -31,4 +32,6 @@ __all__ = [
     "Audio",
     "AsyncEndpoints",
     "Endpoints",
+    "Batches",
+    "AsyncBatches",
 ]

together/resources/batch.py

@@ -0,0 +1,136 @@
+from __future__ import annotations
+
+from typing import List
+
+from together.abstract import api_requestor
+from together.together_response import TogetherResponse
+from together.types import (
+    TogetherClient,
+    TogetherRequest,
+    BatchJob,
+)
+
+
+class Batches:
+    def __init__(self, client: TogetherClient) -> None:
+        self._client = client
+
+    def create_batch(self, file_id: str, endpoint: str) -> BatchJob:
+
+        requestor = api_requestor.APIRequestor(
+            client=self._client,
+        )
+
+        parameter_payload = {
+            "input_file_id": file_id,
+            "endpoint": endpoint,
+            "completion_window": "24h",
+        }
+
+        response, _, _ = requestor.request(
+            options=TogetherRequest(
+                method="POST",
+                url=f"batches",
+                params=parameter_payload,
+            ),
+            stream=False,
+        )
+
+        assert isinstance(response, TogetherResponse)
+        response_body = response.data.get("job", {})
+        return BatchJob(**response_body)
+
+    def get_batch(self, batch_job_id: str) -> BatchJob:
+        requestor = api_requestor.APIRequestor(
+            client=self._client,
+        )
+
+        response, _, _ = requestor.request(
+            options=TogetherRequest(
+                method="GET",
+                url=f"batches/{batch_job_id}",
+            ),
+            stream=False,
+        )
+
+        assert isinstance(response, TogetherResponse)
+        return BatchJob(**response.data)
+
+    def list_batches(self) -> List[BatchJob]:
+        requestor = api_requestor.APIRequestor(
+            client=self._client,
+        )
+
+        response, _, _ = requestor.request(
+            options=TogetherRequest(
+                method="GET",
+                url="batches",
+            ),
+            stream=False,
+        )
+
+        assert isinstance(response, TogetherResponse)
+        jobs = response.data or []
+        return [BatchJob(**job) for job in jobs]
+
+
+class AsyncBatches:
+    def __init__(self, client: TogetherClient) -> None:
+        self._client = client
+
+    async def create_batch(self, file_id: str, endpoint: str) -> BatchJob:
+        requestor = api_requestor.APIRequestor(
+            client=self._client,
+        )
+
+        parameter_payload = {
+            "input_file_id": file_id,
+            "endpoint": endpoint,
+            "completion_window": "24h",
+        }
+
+        response, _, _ = await requestor.arequest(
+            options=TogetherRequest(
+                method="POST",
+                url=f"batches",
+                params=parameter_payload,
+            ),
+            stream=False,
+        )
+
+        assert isinstance(response, TogetherResponse)
+        response_body = response.data.get("job", {})
+        return BatchJob(**response_body)
+
+    async def get_batch(self, batch_job_id: str) -> BatchJob:
+        requestor = api_requestor.APIRequestor(
+            client=self._client,
+        )
+
+        response, _, _ = await requestor.arequest(
+            options=TogetherRequest(
+                method="GET",
+                url=f"batches/{batch_job_id}",
+            ),
+            stream=False,
+        )
+
+        assert isinstance(response, TogetherResponse)
+        return BatchJob(**response.data)
+
+    async def list_batches(self) -> List[BatchJob]:
+        requestor = api_requestor.APIRequestor(
+            client=self._client,
+        )
+
+        response, _, _ = await requestor.arequest(
+            options=TogetherRequest(
+                method="GET",
+                url="batches",
+            ),
+            stream=False,
+        )
+
+        assert isinstance(response, TogetherResponse)
+        jobs = response.data or []
+        return [BatchJob(**job) for job in jobs]

together/resources/files.py

@@ -32,7 +32,7 @@ class Files:
     ) -> FileResponse:
         upload_manager = UploadManager(self._client)
 
-        if check:
+        if check and purpose == FilePurpose.FineTune:
             report_dict = check_file(file)
             if not report_dict["is_check_passed"]:
                 raise FileTypeError(

together/resources/finetune.py

@@ -2,7 +2,7 @@ from __future__ import annotations
 
 import re
 from pathlib import Path
-from typing import Dict, List, Literal
+from typing import List, Dict, Literal
 
 from rich import print as rprint
 
@@ -53,7 +53,7 @@ def create_finetune_request(
     n_checkpoints: int | None = 1,
     batch_size: int | Literal["max"] = "max",
     learning_rate: float | None = 0.00001,
-    lr_scheduler_type: Literal["linear", "cosine"] = "linear",
+    lr_scheduler_type: Literal["linear", "cosine"] = "cosine",
     min_lr_ratio: float = 0.0,
     scheduler_num_cycles: float = 0.5,
     warmup_ratio: float | None = None,
@@ -69,7 +69,7 @@ def create_finetune_request(
     wandb_base_url: str | None = None,
     wandb_project_name: str | None = None,
     wandb_name: str | None = None,
-    train_on_inputs: bool | Literal["auto"] = "auto",
+    train_on_inputs: bool | Literal["auto"] | None = None,
     training_method: str = "sft",
     dpo_beta: float | None = None,
     from_checkpoint: str | None = None,
@@ -101,6 +101,11 @@ def create_finetune_request(
             raise ValueError(
                 f"LoRA adapters are not supported for the selected model ({model_or_checkpoint})."
             )
+
+        if lora_dropout is not None:
+            if not 0 <= lora_dropout < 1.0:
+                raise ValueError("LoRA dropout must be in [0, 1) range.")
+
         lora_r = lora_r if lora_r is not None else model_limits.lora_training.max_rank
         lora_alpha = lora_alpha if lora_alpha is not None else lora_r * 2
         training_type = LoRATrainingType(
@@ -166,6 +171,18 @@ def create_finetune_request(
             f"training_method must be one of {', '.join(AVAILABLE_TRAINING_METHODS)}"
         )
 
+    if train_on_inputs is not None and training_method != "sft":
+        raise ValueError("train_on_inputs is only supported for SFT training")
+
+    if train_on_inputs is None and training_method == "sft":
+        log_warn_once(
+            "train_on_inputs is not set for SFT training, it will be set to 'auto'"
+        )
+        train_on_inputs = "auto"
+
+    if dpo_beta is not None and training_method != "dpo":
+        raise ValueError("dpo_beta is only supported for DPO training")
+
     lr_scheduler: FinetuneLRScheduler
     if lr_scheduler_type == "cosine":
         if scheduler_num_cycles <= 0.0:
@@ -183,8 +200,10 @@ def create_finetune_request(
             lr_scheduler_args=LinearLRSchedulerArgs(min_lr_ratio=min_lr_ratio),
         )
 
-    training_method_cls: TrainingMethodSFT | TrainingMethodDPO = TrainingMethodSFT()
-    if training_method == "dpo":
+    training_method_cls: TrainingMethodSFT | TrainingMethodDPO
+    if training_method == "sft":
+        training_method_cls = TrainingMethodSFT(train_on_inputs=train_on_inputs)
+    elif training_method == "dpo":
         training_method_cls = TrainingMethodDPO(dpo_beta=dpo_beta)
 
     finetune_request = FinetuneRequest(
@@ -206,7 +225,6 @@ def create_finetune_request(
         wandb_base_url=wandb_base_url,
         wandb_project_name=wandb_project_name,
         wandb_name=wandb_name,
-        train_on_inputs=train_on_inputs,
         training_method=training_method_cls,
         from_checkpoint=from_checkpoint,
     )
@@ -263,7 +281,7 @@ class FineTuning:
         n_checkpoints: int | None = 1,
         batch_size: int | Literal["max"] = "max",
         learning_rate: float | None = 0.00001,
-        lr_scheduler_type: Literal["linear", "cosine"] = "linear",
+        lr_scheduler_type: Literal["linear", "cosine"] = "cosine",
         min_lr_ratio: float = 0.0,
         scheduler_num_cycles: float = 0.5,
         warmup_ratio: float = 0.0,
@@ -281,7 +299,7 @@ class FineTuning:
         wandb_name: str | None = None,
         verbose: bool = False,
         model_limits: FinetuneTrainingLimits | None = None,
-        train_on_inputs: bool | Literal["auto"] = "auto",
+        train_on_inputs: bool | Literal["auto"] | None = None,
         training_method: str = "sft",
         dpo_beta: float | None = None,
         from_checkpoint: str | None = None,
@@ -300,7 +318,7 @@ class FineTuning:
             batch_size (int or "max"): Batch size for fine-tuning. Defaults to max.
             learning_rate (float, optional): Learning rate multiplier to use for training
                 Defaults to 0.00001.
-            lr_scheduler_type (Literal["linear", "cosine"]): Learning rate scheduler type. Defaults to "linear".
+            lr_scheduler_type (Literal["linear", "cosine"]): Learning rate scheduler type. Defaults to "cosine".
             min_lr_ratio (float, optional): Min learning rate ratio of the initial learning rate for
                 the learning rate scheduler. Defaults to 0.0.
             scheduler_num_cycles (float, optional): Number or fraction of cycles for the cosine learning rate scheduler. Defaults to 0.5.
@@ -326,12 +344,12 @@ class FineTuning:
                 Defaults to False.
             model_limits (FinetuneTrainingLimits, optional): Limits for the hyperparameters the model in Fine-tuning.
                 Defaults to None.
-            train_on_inputs (bool or "auto"): Whether to mask the user messages in conversational data or prompts in instruction data.
+            train_on_inputs (bool or "auto", optional): Whether to mask the user messages in conversational data or prompts in instruction data.
                 "auto" will automatically determine whether to mask the inputs based on the data format.
                 For datasets with the "text" field (general format), inputs will not be masked.
                 For datasets with the "messages" field (conversational format) or "prompt" and "completion" fields
                 (Instruction format), inputs will be masked.
-                Defaults to "auto".
+                Defaults to None, or "auto" if training_method is "sft" (set in create_finetune_request).
             training_method (str, optional): Training method. Defaults to "sft".
                 Supported methods: "sft", "dpo".
             dpo_beta (float, optional): DPO beta parameter. Defaults to None.
@@ -545,7 +563,7 @@ class FineTuning:
         *,
         output: Path | str | None = None,
         checkpoint_step: int | None = None,
-        checkpoint_type: DownloadCheckpointType = DownloadCheckpointType.DEFAULT,
+        checkpoint_type: DownloadCheckpointType | str = DownloadCheckpointType.DEFAULT,
     ) -> FinetuneDownloadResult:
         """
         Downloads compressed fine-tuned model or checkpoint to local disk.
@@ -558,7 +576,7 @@ class FineTuning:
                 Defaults to None.
             checkpoint_step (int, optional): Specifies step number for checkpoint to download.
                 Defaults to -1 (download the final model)
-            checkpoint_type (CheckpointType, optional): Specifies which checkpoint to download.
+            checkpoint_type (CheckpointType | str, optional): Specifies which checkpoint to download.
                 Defaults to CheckpointType.DEFAULT.
 
         Returns:
@@ -582,6 +600,16 @@ class FineTuning:
 
         ft_job = self.retrieve(id)
 
+        # convert str to DownloadCheckpointType
+        if isinstance(checkpoint_type, str):
+            try:
+                checkpoint_type = DownloadCheckpointType(checkpoint_type.lower())
+            except ValueError:
+                enum_strs = ", ".join(e.value for e in DownloadCheckpointType)
+                raise ValueError(
+                    f"Invalid checkpoint type: {checkpoint_type}. Choose one of {{{enum_strs}}}."
+                )
+
         if isinstance(ft_job.training_type, FullTrainingType):
             if checkpoint_type != DownloadCheckpointType.DEFAULT:
                 raise ValueError(
@@ -592,10 +620,11 @@ class FineTuning:
             if checkpoint_type == DownloadCheckpointType.DEFAULT:
                 checkpoint_type = DownloadCheckpointType.MERGED
 
-            if checkpoint_type == DownloadCheckpointType.MERGED:
-                url += f"&checkpoint={DownloadCheckpointType.MERGED.value}"
-            elif checkpoint_type == DownloadCheckpointType.ADAPTER:
-                url += f"&checkpoint={DownloadCheckpointType.ADAPTER.value}"
+            if checkpoint_type in {
+                DownloadCheckpointType.MERGED,
+                DownloadCheckpointType.ADAPTER,
+            }:
+                url += f"&checkpoint={checkpoint_type.value}"
             else:
                 raise ValueError(
                     f"Invalid checkpoint type for LoRATrainingType: {checkpoint_type}"
@@ -664,7 +693,7 @@ class AsyncFineTuning:
         n_checkpoints: int | None = 1,
         batch_size: int | Literal["max"] = "max",
         learning_rate: float | None = 0.00001,
-        lr_scheduler_type: Literal["linear", "cosine"] = "linear",
+        lr_scheduler_type: Literal["linear", "cosine"] = "cosine",
         min_lr_ratio: float = 0.0,
         scheduler_num_cycles: float = 0.5,
         warmup_ratio: float = 0.0,
@@ -682,7 +711,7 @@ class AsyncFineTuning:
         wandb_name: str | None = None,
         verbose: bool = False,
         model_limits: FinetuneTrainingLimits | None = None,
-        train_on_inputs: bool | Literal["auto"] = "auto",
+        train_on_inputs: bool | Literal["auto"] | None = None,
         training_method: str = "sft",
         dpo_beta: float | None = None,
         from_checkpoint: str | None = None,
@@ -701,7 +730,7 @@ class AsyncFineTuning:
             batch_size (int, optional): Batch size for fine-tuning. Defaults to max.
             learning_rate (float, optional): Learning rate multiplier to use for training
                 Defaults to 0.00001.
-            lr_scheduler_type (Literal["linear", "cosine"]): Learning rate scheduler type. Defaults to "linear".
+            lr_scheduler_type (Literal["linear", "cosine"]): Learning rate scheduler type. Defaults to "cosine".
             min_lr_ratio (float, optional): Min learning rate ratio of the initial learning rate for
                 the learning rate scheduler. Defaults to 0.0.
             scheduler_num_cycles (float, optional): Number or fraction of cycles for the cosine learning rate scheduler. Defaults to 0.5.
@@ -732,7 +761,7 @@ class AsyncFineTuning:
                 For datasets with the "text" field (general format), inputs will not be masked.
                 For datasets with the "messages" field (conversational format) or "prompt" and "completion" fields
                 (Instruction format), inputs will be masked.
-                Defaults to "auto".
+                Defaults to None, or "auto" if training_method is "sft" (set in create_finetune_request).
             training_method (str, optional): Training method. Defaults to "sft".
                 Supported methods: "sft", "dpo".
             dpo_beta (float, optional): DPO beta parameter. Defaults to None.

together/types/__init__.py

@@ -52,6 +52,7 @@ from together.types.finetune import (
 from together.types.images import ImageRequest, ImageResponse
 from together.types.models import ModelObject
 from together.types.rerank import RerankRequest, RerankResponse
+from together.types.batch import BatchJob, BatchJobStatus, BatchEndpoint
 
 
 __all__ = [
@@ -104,4 +105,7 @@ __all__ = [
     "DedicatedEndpoint",
     "ListEndpoint",
     "Autoscaling",
+    "BatchJob",
+    "BatchJobStatus",
+    "BatchEndpoint",
 ]

together/types/batch.py

@@ -0,0 +1,53 @@
+from __future__ import annotations
+
+from enum import Enum
+from typing import Optional
+from datetime import datetime
+
+from pydantic import Field
+
+from together.types.abstract import BaseModel
+
+
+class BatchJobStatus(str, Enum):
+    """
+    The status of a batch job
+    """
+
+    VALIDATING = "VALIDATING"
+    IN_PROGRESS = "IN_PROGRESS"
+    COMPLETED = "COMPLETED"
+    FAILED = "FAILED"
+    EXPIRED = "EXPIRED"
+    CANCELLED = "CANCELLED"
+
+
+class BatchEndpoint(str, Enum):
+    """
+    The endpoint of a batch job
+    """
+
+    COMPLETIONS = "/v1/completions"
+    CHAT_COMPLETIONS = "/v1/chat/completions"
+    # More endpoints can be added here as needed
+
+
+class BatchJob(BaseModel):
+    """
+    A batch job object
+    """
+
+    id: str
+    user_id: str
+    input_file_id: str
+    file_size_bytes: int
+    status: BatchJobStatus
+    job_deadline: datetime
+    created_at: datetime
+    endpoint: str
+    progress: float = 0.0
+    model_id: Optional[str] = None
+    output_file_id: Optional[str] = None
+    error_file_id: Optional[str] = None
+    error: Optional[str] = None
+    completed_at: Optional[datetime] = None

together/types/files.py

@@ -13,6 +13,7 @@ from together.types.common import (
 
 class FilePurpose(str, Enum):
     FineTune = "fine-tune"
+    BatchAPI = "batch-api"
 
 
 class FileType(str, Enum):

together/types/finetune.py

@@ -3,7 +3,7 @@ from __future__ import annotations
 from enum import Enum
 from typing import List, Literal, Any
 
-from pydantic import StrictBool, Field, field_validator
+from pydantic import Field, StrictBool, field_validator
 
 from together.types.abstract import BaseModel
 from together.types.common import (
@@ -149,6 +149,7 @@ class TrainingMethodSFT(TrainingMethod):
     """
 
     method: Literal["sft"] = "sft"
+    train_on_inputs: StrictBool | Literal["auto"] = "auto"
 
 
 class TrainingMethodDPO(TrainingMethod):
@@ -201,8 +202,6 @@ class FinetuneRequest(BaseModel):
     wandb_name: str | None = None
     # training type
     training_type: FullTrainingType | LoRATrainingType | None = None
-    # train on inputs
-    train_on_inputs: StrictBool | Literal["auto"] = "auto"
     # training method
     training_method: TrainingMethodSFT | TrainingMethodDPO = Field(
         default_factory=TrainingMethodSFT

together/utils/files.py

@@ -6,7 +6,6 @@ from pathlib import Path
 from traceback import format_exc
 from typing import Any, Dict, List
 
-from pyarrow import ArrowInvalid, parquet
 
 from together.constants import (
     MAX_FILE_SIZE_GB,
@@ -372,6 +371,14 @@ def _check_jsonl(file: Path) -> Dict[str, Any]:
 
 
 def _check_parquet(file: Path) -> Dict[str, Any]:
+    try:
+        # Pyarrow is optional as it's large (~80MB) and isn't compatible with older systems.
+        from pyarrow import ArrowInvalid, parquet
+    except ImportError:
+        raise ImportError(
+            "pyarrow is not installed and is required to use parquet files. Please install it via `pip install together[pyarrow]`"
+        )
+
     report_dict: Dict[str, Any] = {}
 
     try:

{together-1.5.8.dist-info → together-1.5.13.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.3
 Name: together
-Version: 1.5.8
+Version: 1.5.13
 Summary: Python client for Together's Cloud Platform!
 License: Apache-2.0
 Author: Together AI
@@ -13,6 +13,7 @@ Classifier: Programming Language :: Python :: 3.10
 Classifier: Programming Language :: Python :: 3.11
 Classifier: Programming Language :: Python :: 3.12
 Classifier: Programming Language :: Python :: 3.13
+Provides-Extra: pyarrow
 Requires-Dist: aiohttp (>=3.9.3,<4.0.0)
 Requires-Dist: click (>=8.1.7,<9.0.0)
 Requires-Dist: eval-type-backport (>=0.1.3,<0.3.0)
@@ -20,7 +21,7 @@ Requires-Dist: filelock (>=3.13.1,<4.0.0)
 Requires-Dist: numpy (>=1.23.5) ; python_version < "3.12"
 Requires-Dist: numpy (>=1.26.0) ; python_version >= "3.12"
 Requires-Dist: pillow (>=11.1.0,<12.0.0)
-Requires-Dist: pyarrow (>=10.0.1)
+Requires-Dist: pyarrow (>=10.0.1) ; extra == "pyarrow"
 Requires-Dist: pydantic (>=2.6.3,<3.0.0)
 Requires-Dist: requests (>=2.31.0,<3.0.0)
 Requires-Dist: rich (>=13.8.1,<15.0.0)
@@ -92,7 +93,7 @@ client = Together()
 
 # Simple text message
 response = client.chat.completions.create(
-    model="mistralai/Mixtral-8x7B-Instruct-v0.1",
+    model="meta-llama/Llama-4-Scout-17B-16E-Instruct",
     messages=[{"role": "user", "content": "tell me about new york"}],
 )
 print(response.choices[0].message.content)
@@ -182,7 +183,7 @@ from together import Together
 
 client = Together()
 stream = client.chat.completions.create(
-    model="mistralai/Mixtral-8x7B-Instruct-v0.1",
+    model="meta-llama/Llama-4-Scout-17B-16E-Instruct",
     messages=[{"role": "user", "content": "tell me about new york"}],
     stream=True,
 )
@@ -207,7 +208,7 @@ async def async_chat_completion(messages):
     async_client = AsyncTogether()
     tasks = [
         async_client.chat.completions.create(
-            model="mistralai/Mixtral-8x7B-Instruct-v0.1",
+            model="meta-llama/Llama-4-Scout-17B-16E-Instruct",
             messages=[{"role": "user", "content": message}],
         )
         for message in messages
@@ -230,7 +231,7 @@ from together import Together
 client = Together()
 
 response = client.chat.completions.create(
-    model="mistralai/Mixtral-8x7B-Instruct-v0.1",
+    model="meta-llama/Llama-3.2-3B-Instruct-Turbo",
     messages=[{"role": "user", "content": "tell me about new york"}],
     logprobs=1
 )
@@ -381,7 +382,7 @@ client.files.delete(id="file-d0d318cb-b7d9-493a-bd70-1cfe089d3815") # deletes a
 
 ### Fine-tunes
 
-The finetune API is used for fine-tuning and allows developers to create finetuning jobs. It also has several methods to list all jobs, retrive statuses and get checkpoints. Please refer to our fine-tuning docs [here](https://docs.together.ai/docs/fine-tuning-python).
+The finetune API is used for fine-tuning and allows developers to create finetuning jobs. It also has several methods to list all jobs, retrive statuses and get checkpoints. Please refer to our fine-tuning docs [here](https://docs.together.ai/docs/fine-tuning-quickstart).
 
 ```python
 from together import Together
@@ -390,7 +391,7 @@ client = Together()
 
 client.fine_tuning.create(
   training_file = 'file-d0d318cb-b7d9-493a-bd70-1cfe089d3815',
-  model = 'mistralai/Mixtral-8x7B-Instruct-v0.1',
+  model = 'meta-llama/Llama-3.2-3B-Instruct',
   n_epochs = 3,
   n_checkpoints = 1,
   batch_size = "max",
@@ -428,7 +429,7 @@ for model in models:
 together chat.completions \
   --message "system" "You are a helpful assistant named Together" \
   --message "user" "What is your name?" \
-  --model mistralai/Mixtral-8x7B-Instruct-v0.1
+  --model meta-llama/Llama-4-Scout-17B-16E-Instruct
 ```
 
 The Chat Completions CLI enables streaming tokens to stdout by default. To disable streaming, use `--no-stream`.
@@ -438,7 +439,7 @@ The Chat Completions CLI enables streaming tokens to stdout by default. To disab
 ```bash
 together completions \
   "Large language models are " \
-  --model mistralai/Mixtral-8x7B-v0.1 \
+  --model meta-llama/Llama-4-Scout-17B-16E-Instruct \
   --max-tokens 512 \
   --stop "."
 ```

{together-1.5.8.dist-info → together-1.5.13.dist-info}/RECORD

@@ -5,14 +5,14 @@ together/cli/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 together/cli/api/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 together/cli/api/chat.py,sha256=2PHRb-9T-lUEKhUJFtc7SxJv3shCVx40gq_8pzfsewM,9234
 together/cli/api/completions.py,sha256=l-Zw5t7hojL3w8xd_mitS2NRB72i5Z0xwkzH0rT5XMc,4263
-together/cli/api/endpoints.py,sha256=tM42PthzkpTjOkmL6twyWYXqEQw0DOqsEKTZdhp6vKU,13284
+together/cli/api/endpoints.py,sha256=f6KafWZvRF6n_ThWdr3y9uhE6wPF37PcD45w_EtgXmY,13289
 together/cli/api/files.py,sha256=QLYEXRkY8J2Gg1SbTCtzGfoTMvosoeACNK83L_oLubs,3397
-together/cli/api/finetune.py,sha256=AnVa5sDB4ZCR_S-CMJStj1vzi0b2AwYQT2c3rkUnAbQ,16238
+together/cli/api/finetune.py,sha256=mM8GF6xP-NM-eOQfn8eXBCLop77OesLVaATAfKm_HMo,16238
 together/cli/api/images.py,sha256=GADSeaNUHUVMtWovmccGuKc28IJ9E_v4vAEwYHJhu5o,2645
 together/cli/api/models.py,sha256=CXw8B1hqNkadogi58GIXhLg_dTJnvTBaE7Kq1_xQ-10,1423
 together/cli/api/utils.py,sha256=IuqYWPnLI38_Bqd7lj8V_SnGdYc59pRmMbQmciS4FsM,1326
 together/cli/cli.py,sha256=YCDzbXpC5is0rs2PEkUPrIhYuzdyrihQ8GVR_TlDv5s,2054
-together/client.py,sha256=lN_KfJs2gCdLfQ5GBrVKin-4ZL7L7UQf9wjofWQ7sXg,5682
+together/client.py,sha256=us5aE8hVzKmMCHZz52NcSPXByOsigd2sKbbyqe4x1m0,5861
 together/constants.py,sha256=UDJhEylJFmdm4bedBDpvqYXBj5Or3k7z9GWtkRY_dZQ,1526
 together/error.py,sha256=HU6247CyzCFjaxL9A0XYbXZ6fY_ebRg0FEYjI4Skogs,5515
 together/filemanager.py,sha256=lwNIYm-BAcnUPtyE0Q_8NpRNsxMlQrpIWFVUVJBBz88,11356
@@ -24,24 +24,26 @@ together/legacy/files.py,sha256=qmAqMiNTPWb6WvLV5Tsv6kxGRfQ31q7OkHZNFwkw8v0,4082
 together/legacy/finetune.py,sha256=XjZ4Dn2hSjMUVm64s6u1bbh9F7r9GbDKp-WLmzyEKRw,5123
 together/legacy/images.py,sha256=bJJRs-6C7-NexPyaeyHiYlHOU51yls5-QAiqtO4xrZU,626
 together/legacy/models.py,sha256=85ZN9Ids_FjdYNDRv5k7sgrtVWPKPHqkDplORtVUGHg,1087
-together/resources/__init__.py,sha256=OQ8tW9mUIX0Ezk0wvYEnnEym6wGsjBKgXFLU9Ffgb-o,984
+together/resources/__init__.py,sha256=jZ9O14K7wKb1U0bmIMosa9P3W3xPCJhoEJiaHw8anCc,1078
 together/resources/audio/__init__.py,sha256=e7xp0Lkp_nMAHXcuFHS7dLXP_YqTPMMZIilW1TW_sAI,551
 together/resources/audio/speech.py,sha256=81ib_gIo-Rxoaipx2Pi9ZsKnOTjeFPwSlBrcUkyX5xk,5211
+together/resources/batch.py,sha256=wSJR30CAFjzZ436vjYdXLCT0ahA5E0ud_qHMS5YvZ1M,3750
 together/resources/chat/__init__.py,sha256=RsTptdP8MeGjcdIjze896-J27cRvCbUoMft0X2BVlQ8,617
 together/resources/chat/completions.py,sha256=jYiNZsWa8RyEacL0VgxWj1egJ857oU4nxIY8uqGHcaU,14459
 together/resources/code_interpreter.py,sha256=vbN8Mh5MG6HQvqra7p61leIyfebgbgJTM_q2A_Fylhw,2948
 together/resources/completions.py,sha256=5Wa-ZjPCxRcam6CDe7KgGYlTA7yJZMmd5TrRgGCL_ug,11726
 together/resources/embeddings.py,sha256=PTvLb82yjG_-iQOyuhsilp77Fr7gZ0o6WD2KeRnKoxs,2675
 together/resources/endpoints.py,sha256=NNjp-wyzOotzlscGGrANhOHxQBjHTN8f5kTQTH_CLvE,17177
-together/resources/files.py,sha256=bnPbaF25e4InBRPvHwXHXT-oSX1Z1sZRsnQW5wq82U4,4990
-together/resources/finetune.py,sha256=6Al4QfqNid_1blWShl1RR9uis0xwrGruITc5Uxr_jkA,36367
+together/resources/files.py,sha256=y3Ri6UtyAa7fjCJ8_fp26Y2hzzi6Aoo21JKkVgljFl8,5026
+together/resources/finetune.py,sha256=M-nvOZMnL7ZyTayiKN9Vos8D1uYNj-ENmrOA2bkPF8A,37617
 together/resources/images.py,sha256=LQUjKPaFxWTqOAPnyF1Pp7Rz4NLOYhmoKwshpYiprEM,4923
 together/resources/models.py,sha256=qgmAXv61Cq4oLxytenEZBywA8shldDHYxJ_EAu_4JWQ,3864
 together/resources/rerank.py,sha256=3Ju_aRSyZ1s_3zCSNZnSnEJErUVmt2xa3M8z1nvejMA,3931
 together/together_response.py,sha256=a3dgKMPDrlfKQwxYENfNt2T4l2vSZxRWMixhHSy-q3E,1308
-together/types/__init__.py,sha256=VgIbE2AOK9c2TQUzkbRbyRkdia2COXJXl_wxPaoxR-M,2688
+together/types/__init__.py,sha256=_93XstLg1OOWratj_N1bsNN-2aS628uHH3SZj0wszyc,2820
 together/types/abstract.py,sha256=1lFQI_3WjsR_t1128AeKW0aTk6EiM6Gh1J3ZuyLLPao,642
 together/types/audio_speech.py,sha256=jlj8BZf3dkIDARF1P11fuenVLj4try8Yx4RN-EAkhOU,2609
+together/types/batch.py,sha256=FP0RuQ3EDy-FV1bh-biPICvyRS7WqLm38GHz5lzKyXM,1112
 together/types/chat_completions.py,sha256=qpBCMXEWtRwW_fmiu6cecm3d4h6mcK8gvr-8JkbAopQ,5104
 together/types/code_interpreter.py,sha256=cjF8TKgRkJllHS4i24dWQZBGTRsG557eHSewOiip0Kk,1770
 together/types/common.py,sha256=kxZ-N9xtBsGYZBmbIWnZ0rfT3Pn8PFB7sAbp3iv96pw,1525
@@ -49,19 +51,19 @@ together/types/completions.py,sha256=o3FR5ixsTUj-a3pmOUzbSQg-hESVhpqrC9UD__VCqr4
 together/types/embeddings.py,sha256=J7grkYYn7xhqeKaBO2T-8XQRtHhkzYzymovtGdIUK5A,751
 together/types/endpoints.py,sha256=EzNhHOoQ_D9fUdNQtxQPeSWiFzdFLqpNodN0YLmv_h0,4393
 together/types/error.py,sha256=OVlCs3cx_2WhZK4JzHT8SQyRIIqKOP1AZQ4y1PydjAE,370
-together/types/files.py,sha256=-rEUfsV6f2vZB9NrFxT4_933ubsDIUNkPB-3OlOFk4A,1954
-together/types/finetune.py,sha256=KVaos9wBjFZFvVoL4DBx3PlXjGAx7_0CTQxKzEZzf40,10940
+together/types/files.py,sha256=i-Ke57p8Svb1MbMZxu-Fo2zxIc6j-mDO2TLGNwPpGu0,1981
+together/types/finetune.py,sha256=Utdcm3kL_cDfBS3zjXwyHsuP2qFFjCQiQZOsPD-WlpE,10918
 together/types/images.py,sha256=xnC-FZGdZU30WSFTybfGneWxb-kj0ZGufJsgHtB8j0k,980
 together/types/models.py,sha256=nwQIZGHKZpX9I6mK8z56VW70YC6Ry6JGsVa0s99QVxc,1055
 together/types/rerank.py,sha256=qZfuXOn7MZ6ly8hpJ_MZ7OU_Bi1-cgYNSB20Wja8Qkk,1061
 together/utils/__init__.py,sha256=5fqvj4KT2rHxKSQot2TSyV_HcvkvkGiqAiaYuJwqtm0,786
 together/utils/_log.py,sha256=5IYNI-jYzxyIS-pUvhb0vE_Muo3MA7GgBhsu66TKP2w,1951
 together/utils/api_helpers.py,sha256=2K0O6qeEQ2zVFvi5NBN5m2kjZJaS3-JfKFecQ7SmGaw,3746
-together/utils/files.py,sha256=rfp10qU0urtWOXXFeasFtO9xp-1KIhM3S43JxcnHmL0,16438
+together/utils/files.py,sha256=btWQawwXbNKfPmCtRyObZViG1Xx-IPz45PrAtMXvcy8,16741
 together/utils/tools.py,sha256=H2MTJhEqtBllaDvOyZehIO_IVNK3P17rSDeILtJIVag,2964
 together/version.py,sha256=p03ivHyE0SyWU4jAnRTBi_sOwywVWoZPU4g2gzRgG-Y,126
-together-1.5.8.dist-info/LICENSE,sha256=xx0jnfkXJvxRnG63LTGOxlggYnIysveWIZ6H3PNdCrQ,11357
-together-1.5.8.dist-info/METADATA,sha256=lsO8JDj10zXplEJu-tVmwkMXNFLOE6RDQ6GQLYhBCXs,15415
-together-1.5.8.dist-info/WHEEL,sha256=b4K_helf-jlQoXBBETfwnf4B04YC67LOev0jo4fX5m8,88
-together-1.5.8.dist-info/entry_points.txt,sha256=G-b5NKW6lUUf1V1fH8IPTBb7jXnK7lhbX9H1zTEJXPs,50
-together-1.5.8.dist-info/RECORD,,
+together-1.5.13.dist-info/LICENSE,sha256=xx0jnfkXJvxRnG63LTGOxlggYnIysveWIZ6H3PNdCrQ,11357
+together-1.5.13.dist-info/METADATA,sha256=pH3dlk_0LIhPs8Z6HW5xnGFPLnxTT7t_-WA3WP4tp1s,15497
+together-1.5.13.dist-info/WHEEL,sha256=b4K_helf-jlQoXBBETfwnf4B04YC67LOev0jo4fX5m8,88
+together-1.5.13.dist-info/entry_points.txt,sha256=G-b5NKW6lUUf1V1fH8IPTBb7jXnK7lhbX9H1zTEJXPs,50
+together-1.5.13.dist-info/RECORD,,