together 1.1.2__tar.gz → 1.1.4__tar.gz

{together-1.1.2 → together-1.1.4}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: together
-Version: 1.1.2
+Version: 1.1.4
 Summary: Python client for Together's Cloud Platform!
 Home-page: https://github.com/togethercomputer/together-python
 License: Apache-2.0
@@ -17,7 +17,7 @@ Classifier: Programming Language :: Python :: 3.11
 Classifier: Programming Language :: Python :: 3.12
 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,<0.2.0)
+Requires-Dist: eval-type-backport (>=0.1.3,<0.3.0)
 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"
@@ -32,32 +32,46 @@ Project-URL: Bug Tracker, https://github.com/togethercomputer/together-python/is
 Project-URL: Repository, https://github.com/togethercomputer/together-python
 Description-Content-Type: text/markdown
 
-The [Together Python Library](https://pypi.org/project/together/) is the official Python client for Together's API platform, providing a convenient way for interacting with the REST APIs and enables easy integrations with Python 3.8+ applications with easy to use synchronous and asynchronous clients.
+<div align="center">
+  <a href="https://www.together.ai/">
+    <img alt="together.ai" height="100px" src="https://assets-global.website-files.com/64f6f2c0e3f4c5a91c1e823a/654693d569494912cfc0c0d4_favicon.svg">
+  </a>
+</div>
 
-# Installation
+# Together Python API library
+
+[![PyPI version](https://img.shields.io/pypi/v/together.svg)](https://pypi.org/project/together/)
+[![Discord](https://dcbadge.vercel.app/api/server/9Rk6sSeWEG?style=flat&compact=true)](https://discord.com/invite/9Rk6sSeWEG)
+[![Twitter](https://img.shields.io/twitter/url/https/twitter.com/togethercompute.svg?style=social&label=Follow%20%40togethercompute)](https://twitter.com/togethercompute)
+
+The [Together Python API Library](https://pypi.org/project/together/) is the official Python client for Together's API platform, providing a convenient way for interacting with the REST APIs and enables easy integrations with Python 3.8+ applications with easy to use synchronous and asynchronous clients.
+
+
+
+## Installation
 
 > 🚧
-> The library was rewritten in v1.0.0 released in April of 2024. There were significant changes made.
+> The Library was rewritten in v1.0.0 released in April of 2024. There were significant changes made.
 
-To install Together Python Library from PyPi, simply run:
+To install Together Python Library from PyPI, simply run:
 
 ```shell Shell
 pip install --upgrade together
 ```
 
-## Setting up API Key
+### Setting up API Key
 
 > 🚧 You will need to create an account with [Together.ai](https://api.together.xyz/) to obtain a Together API Key.
 
 Once logged in to the Together Playground, you can find available API keys in [this settings page](https://api.together.xyz/settings/api-keys).
 
-### Setting environment variable
+#### Setting environment variable
 
 ```shell
 export TOGETHER_API_KEY=xxxxx
 ```
 
-### Using the client
+#### Using the client
 
 ```python
 from together import Together
@@ -65,11 +79,11 @@ from together import Together
 client = Together(api_key="xxxxx")
 ```
 
-This library contains both a python library and a CLI. We'll demonstrate how to use both below.
+This repo contains both a Python Library and a CLI. We'll demonstrate how to use both below.
 
-# Usage – Python Client
+## Usage – Python Client
 
-## Chat Completions
+### Chat Completions
 
 ```python
 import os
@@ -84,7 +98,7 @@ response = client.chat.completions.create(
 print(response.choices[0].message.content)
 ```
 
-### Streaming
+#### Streaming
 
 ```python
 import os
@@ -101,7 +115,7 @@ for chunk in stream:
     print(chunk.choices[0].delta.content or "", end="", flush=True)
 ```
 
-### Async usage
+#### Async usage
 
 ```python
 import os, asyncio
@@ -130,7 +144,7 @@ async def async_chat_completion(messages):
 asyncio.run(async_chat_completion(messages))
 ```
 
-## Completions
+### Completions
 
 Completions are for code and language models shown [here](https://docs.together.ai/docs/inference-models). Below, a code model example is shown.
 
@@ -143,11 +157,12 @@ client = Together(api_key=os.environ.get("TOGETHER_API_KEY"))
 response = client.completions.create(
     model="codellama/CodeLlama-34b-Python-hf",
     prompt="Write a Next.js component with TailwindCSS for a header component.",
+    max_tokens=200,
 )
 print(response.choices[0].text)
 ```
 
-### Streaming
+#### Streaming
 
 ```python
 import os
@@ -164,7 +179,7 @@ for chunk in stream:
     print(chunk.choices[0].delta.content or "", end="", flush=True)
 ```
 
-### Async usage
+#### Async usage
 
 ```python
 import os, asyncio
@@ -193,7 +208,7 @@ async def async_chat_completion(prompts):
 asyncio.run(async_chat_completion(prompts))
 ```
 
-## Image generation
+### Image generation
 
 ```python
 import os
@@ -210,7 +225,7 @@ response = client.images.generate(
 print(response.data[0].b64_json)
 ```
 
-## Embeddings
+### Embeddings
 
 ```python
 from typing import List
@@ -229,7 +244,7 @@ embeddings = get_embeddings(input_texts, model='togethercomputer/m2-bert-80M-8k-
 print(embeddings)
 ```
 
-## Files
+### Files
 
 The files API is used for fine-tuning and allows developers to upload data to fine-tune on. It also has several methods to list all files, retrive files, and delete files. Please refer to our fine-tuning docs [here](https://docs.together.ai/docs/fine-tuning-python).
 
@@ -246,7 +261,7 @@ client.files.retrieve_content(id="file-d0d318cb-b7d9-493a-bd70-1cfe089d3815") #
 client.files.delete(id="file-d0d318cb-b7d9-493a-bd70-1cfe089d3815") # deletes a file
 ```
 
-## Fine-tunes
+### 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).
 
@@ -273,7 +288,7 @@ client.fine_tuning.list_events(id="ft-c66a5c18-1d6d-43c9-94bd-32d756425b4b") #
 client.fine_tuning.download(id="ft-c66a5c18-1d6d-43c9-94bd-32d756425b4b") # downloads compressed fine-tuned model or checkpoint to local disk
 ```
 
-## Models
+### Models
 
 This lists all the models that Together supports.
 
@@ -289,9 +304,9 @@ for model in models:
     print(model)
 ```
 
-# Usage – CLI
+## Usage – CLI
 
-## Chat Completions
+### Chat Completions
 
 ```bash
 together chat.completions \
@@ -302,7 +317,7 @@ together chat.completions \
 
 The Chat Completions CLI enables streaming tokens to stdout by default. To disable streaming, use `--no-stream`.
 
-## Completions
+### Completions
 
 ```bash
 together completions \
@@ -314,7 +329,7 @@ together completions \
 
 The Completions CLI enables streaming tokens to stdout by default. To disable streaming, use `--no-stream`.
 
-## Image Generations
+### Image Generations
 
 ```bash
 together images generate \
@@ -325,7 +340,7 @@ together images generate \
 
 The image is opened in the default image viewer by default. To disable this, use `--no-show`.
 
-## Files
+### Files
 
 ```bash
 # Help
@@ -350,7 +365,7 @@ together files retrieve-content file-6f50f9d1-5b95-416c-9040-0799b2b4b894
 together files delete file-6f50f9d1-5b95-416c-9040-0799b2b4b894
 ```
 
-## Fine-tuning
+### Fine-tuning
 
 ```bash
 # Help
@@ -377,7 +392,7 @@ together fine-tuning cancel ft-c66a5c18-1d6d-43c9-94bd-32d756425b4b
 together fine-tuning download ft-c66a5c18-1d6d-43c9-94bd-32d756425b4b
 ```
 
-## Models
+### Models
 
 ```bash
 # Help

{together-1.1.2 → together-1.1.4}/README.md

@@ -1,29 +1,43 @@
-The [Together Python Library](https://pypi.org/project/together/) is the official Python client for Together's API platform, providing a convenient way for interacting with the REST APIs and enables easy integrations with Python 3.8+ applications with easy to use synchronous and asynchronous clients.
+<div align="center">
+  <a href="https://www.together.ai/">
+    <img alt="together.ai" height="100px" src="https://assets-global.website-files.com/64f6f2c0e3f4c5a91c1e823a/654693d569494912cfc0c0d4_favicon.svg">
+  </a>
+</div>
 
-# Installation
+# Together Python API library
+
+[![PyPI version](https://img.shields.io/pypi/v/together.svg)](https://pypi.org/project/together/)
+[![Discord](https://dcbadge.vercel.app/api/server/9Rk6sSeWEG?style=flat&compact=true)](https://discord.com/invite/9Rk6sSeWEG)
+[![Twitter](https://img.shields.io/twitter/url/https/twitter.com/togethercompute.svg?style=social&label=Follow%20%40togethercompute)](https://twitter.com/togethercompute)
+
+The [Together Python API Library](https://pypi.org/project/together/) is the official Python client for Together's API platform, providing a convenient way for interacting with the REST APIs and enables easy integrations with Python 3.8+ applications with easy to use synchronous and asynchronous clients.
+
+
+
+## Installation
 
 > 🚧
-> The library was rewritten in v1.0.0 released in April of 2024. There were significant changes made.
+> The Library was rewritten in v1.0.0 released in April of 2024. There were significant changes made.
 
-To install Together Python Library from PyPi, simply run:
+To install Together Python Library from PyPI, simply run:
 
 ```shell Shell
 pip install --upgrade together
 ```
 
-## Setting up API Key
+### Setting up API Key
 
 > 🚧 You will need to create an account with [Together.ai](https://api.together.xyz/) to obtain a Together API Key.
 
 Once logged in to the Together Playground, you can find available API keys in [this settings page](https://api.together.xyz/settings/api-keys).
 
-### Setting environment variable
+#### Setting environment variable
 
 ```shell
 export TOGETHER_API_KEY=xxxxx
 ```
 
-### Using the client
+#### Using the client
 
 ```python
 from together import Together
@@ -31,11 +45,11 @@ from together import Together
 client = Together(api_key="xxxxx")
 ```
 
-This library contains both a python library and a CLI. We'll demonstrate how to use both below.
+This repo contains both a Python Library and a CLI. We'll demonstrate how to use both below.
 
-# Usage – Python Client
+## Usage – Python Client
 
-## Chat Completions
+### Chat Completions
 
 ```python
 import os
@@ -50,7 +64,7 @@ response = client.chat.completions.create(
 print(response.choices[0].message.content)
 ```
 
-### Streaming
+#### Streaming
 
 ```python
 import os
@@ -67,7 +81,7 @@ for chunk in stream:
     print(chunk.choices[0].delta.content or "", end="", flush=True)
 ```
 
-### Async usage
+#### Async usage
 
 ```python
 import os, asyncio
@@ -96,7 +110,7 @@ async def async_chat_completion(messages):
 asyncio.run(async_chat_completion(messages))
 ```
 
-## Completions
+### Completions
 
 Completions are for code and language models shown [here](https://docs.together.ai/docs/inference-models). Below, a code model example is shown.
 
@@ -109,11 +123,12 @@ client = Together(api_key=os.environ.get("TOGETHER_API_KEY"))
 response = client.completions.create(
     model="codellama/CodeLlama-34b-Python-hf",
     prompt="Write a Next.js component with TailwindCSS for a header component.",
+    max_tokens=200,
 )
 print(response.choices[0].text)
 ```
 
-### Streaming
+#### Streaming
 
 ```python
 import os
@@ -130,7 +145,7 @@ for chunk in stream:
     print(chunk.choices[0].delta.content or "", end="", flush=True)
 ```
 
-### Async usage
+#### Async usage
 
 ```python
 import os, asyncio
@@ -159,7 +174,7 @@ async def async_chat_completion(prompts):
 asyncio.run(async_chat_completion(prompts))
 ```
 
-## Image generation
+### Image generation
 
 ```python
 import os
@@ -176,7 +191,7 @@ response = client.images.generate(
 print(response.data[0].b64_json)
 ```
 
-## Embeddings
+### Embeddings
 
 ```python
 from typing import List
@@ -195,7 +210,7 @@ embeddings = get_embeddings(input_texts, model='togethercomputer/m2-bert-80M-8k-
 print(embeddings)
 ```
 
-## Files
+### Files
 
 The files API is used for fine-tuning and allows developers to upload data to fine-tune on. It also has several methods to list all files, retrive files, and delete files. Please refer to our fine-tuning docs [here](https://docs.together.ai/docs/fine-tuning-python).
 
@@ -212,7 +227,7 @@ client.files.retrieve_content(id="file-d0d318cb-b7d9-493a-bd70-1cfe089d3815") #
 client.files.delete(id="file-d0d318cb-b7d9-493a-bd70-1cfe089d3815") # deletes a file
 ```
 
-## Fine-tunes
+### 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).
 
@@ -239,7 +254,7 @@ client.fine_tuning.list_events(id="ft-c66a5c18-1d6d-43c9-94bd-32d756425b4b") #
 client.fine_tuning.download(id="ft-c66a5c18-1d6d-43c9-94bd-32d756425b4b") # downloads compressed fine-tuned model or checkpoint to local disk
 ```
 
-## Models
+### Models
 
 This lists all the models that Together supports.
 
@@ -255,9 +270,9 @@ for model in models:
     print(model)
 ```
 
-# Usage – CLI
+## Usage – CLI
 
-## Chat Completions
+### Chat Completions
 
 ```bash
 together chat.completions \
@@ -268,7 +283,7 @@ together chat.completions \
 
 The Chat Completions CLI enables streaming tokens to stdout by default. To disable streaming, use `--no-stream`.
 
-## Completions
+### Completions
 
 ```bash
 together completions \
@@ -280,7 +295,7 @@ together completions \
 
 The Completions CLI enables streaming tokens to stdout by default. To disable streaming, use `--no-stream`.
 
-## Image Generations
+### Image Generations
 
 ```bash
 together images generate \
@@ -291,7 +306,7 @@ together images generate \
 
 The image is opened in the default image viewer by default. To disable this, use `--no-show`.
 
-## Files
+### Files
 
 ```bash
 # Help
@@ -316,7 +331,7 @@ together files retrieve-content file-6f50f9d1-5b95-416c-9040-0799b2b4b894
 together files delete file-6f50f9d1-5b95-416c-9040-0799b2b4b894
 ```
 
-## Fine-tuning
+### Fine-tuning
 
 ```bash
 # Help
@@ -343,7 +358,7 @@ together fine-tuning cancel ft-c66a5c18-1d6d-43c9-94bd-32d756425b4b
 together fine-tuning download ft-c66a5c18-1d6d-43c9-94bd-32d756425b4b
 ```
 
-## Models
+### Models
 
 ```bash
 # Help

{together-1.1.2 → together-1.1.4}/pyproject.toml

@@ -12,7 +12,7 @@ build-backend = "poetry.masonry.api"
 
 [tool.poetry]
 name = "together"
-version = "1.1.2"
+version = "1.1.4"
 authors = [
     "Together AI <support@together.ai>"
 ]
@@ -36,7 +36,7 @@ tabulate = "^0.9.0"
 pydantic = "^2.6.3"
 aiohttp = "^3.9.3"
 filelock = "^3.13.1"
-eval-type-backport = "^0.1.3"
+eval-type-backport = ">=0.1.3,<0.3.0"
 click = "^8.1.7"
 pillow = "^10.3.0"
 pyarrow = ">=10.0.1"
@@ -50,7 +50,7 @@ optional = true
 
 [tool.poetry.group.quality.dependencies]
 black = ">=23.1,<25.0"
-ruff = "^0.3.2"
+ruff = ">=0.3.2,<0.5.0"
 types-tqdm = "^4.65.0.0"
 types-tabulate = "^0.9.0.3"
 pre-commit = "3.5.0"
@@ -66,6 +66,13 @@ pytest = ">=7.4.2,<9.0.0"
 pytest-watch = "^4.2.0"
 tox = "^4.14.1"
 
+[tool.poetry.group.examples]
+optional = true
+
+[tool.poetry.group.examples.dependencies]
+datasets = "^2.18.0"
+transformers = "^4.39.3"
+
 
 [tool.poetry.urls]
 "Homepage" = "https://github.com/togethercomputer/together-python"

{together-1.1.2 → together-1.1.4}/src/together/cli/api/chat.py

@@ -28,6 +28,9 @@ class ChatShell(cmd.Cmd):
         top_p: float | None = None,
         top_k: int | None = None,
         repetition_penalty: float | None = None,
+        presence_penalty: float | None = None,
+        frequency_penalty: float | None = None,
+        min_p: float | None = None,
         safety_model: str | None = None,
         system_message: str | None = None,
     ) -> None:
@@ -40,6 +43,9 @@ class ChatShell(cmd.Cmd):
         self.top_p = top_p
         self.top_k = top_k
         self.repetition_penalty = repetition_penalty
+        self.presence_penalty = presence_penalty
+        self.frequency_penalty = frequency_penalty
+        self.min_p = min_p
         self.safety_model = safety_model
         self.system_message = system_message
 
@@ -69,6 +75,9 @@ class ChatShell(cmd.Cmd):
             top_p=self.top_p,
             top_k=self.top_k,
             repetition_penalty=self.repetition_penalty,
+            presence_penalty=self.presence_penalty,
+            frequency_penalty=self.frequency_penalty,
+            min_p=self.min_p,
             safety_model=self.safety_model,
             stream=True,
         ):
@@ -76,13 +85,12 @@ class ChatShell(cmd.Cmd):
             assert isinstance(chunk, ChatCompletionChunk)
             assert chunk.choices
             assert chunk.choices[0].delta
-            assert chunk.choices[0].delta.content
 
             token = chunk.choices[0].delta.content
 
             click.echo(token, nl=False)
 
-            output += token
+            output += token or ""
 
         click.echo("\n")
 
@@ -109,6 +117,10 @@ class ChatShell(cmd.Cmd):
 @click.option("--temperature", type=float, help="Sampling temperature")
 @click.option("--top-p", type=int, help="Top p sampling")
 @click.option("--top-k", type=float, help="Top k sampling")
+@click.option("--repetition-penalty", type=float, help="Repetition penalty")
+@click.option("--presence-penalty", type=float, help="Presence penalty")
+@click.option("--frequency-penalty", type=float, help="Frequency penalty")
+@click.option("--min-p", type=float, help="Minimum p")
 @click.option("--safety-model", type=str, help="Moderation model")
 @click.option("--system-message", type=str, help="System message to use for the chat")
 def interactive(
@@ -120,6 +132,9 @@ def interactive(
     top_p: float | None = None,
     top_k: int | None = None,
     repetition_penalty: float | None = None,
+    presence_penalty: float | None = None,
+    frequency_penalty: float | None = None,
+    min_p: float | None = None,
     safety_model: str | None = None,
     system_message: str | None = None,
 ) -> None:
@@ -135,6 +150,9 @@ def interactive(
         top_p=top_p,
         top_k=top_k,
         repetition_penalty=repetition_penalty,
+        presence_penalty=presence_penalty,
+        frequency_penalty=frequency_penalty,
+        min_p=min_p,
         safety_model=safety_model,
         system_message=system_message,
     ).cmdloop()
@@ -158,6 +176,11 @@ def interactive(
 @click.option("--top-p", type=int, help="Top p sampling")
 @click.option("--top-k", type=float, help="Top k sampling")
 @click.option("--repetition-penalty", type=float, help="Repetition penalty")
+@click.option("--presence-penalty", type=float, help="Presence penalty sampling method")
+@click.option(
+    "--frequency-penalty", type=float, help="Frequency penalty sampling method"
+)
+@click.option("--min-p", type=float, help="Min p sampling")
 @click.option("--no-stream", is_flag=True, help="Disable streaming")
 @click.option("--logprobs", type=int, help="Return logprobs. Only works with --raw.")
 @click.option("--echo", is_flag=True, help="Echo prompt. Only works with --raw.")
@@ -174,6 +197,9 @@ def chat(
     top_p: float | None = None,
     top_k: int | None = None,
     repetition_penalty: float | None = None,
+    presence_penalty: float | None = None,
+    frequency_penalty: float | None = None,
+    min_p: float | None = None,
     no_stream: bool = False,
     logprobs: int | None = None,
     echo: bool | None = None,
@@ -195,6 +221,9 @@ def chat(
         max_tokens=max_tokens,
         stop=stop,
         repetition_penalty=repetition_penalty,
+        presence_penalty=presence_penalty,
+        frequency_penalty=frequency_penalty,
+        min_p=min_p,
         stream=not no_stream,
         logprobs=logprobs,
         echo=echo,

{together-1.1.2 → together-1.1.4}/src/together/cli/api/completions.py

@@ -14,7 +14,6 @@ from together.types.completions import CompletionChoicesChunk, CompletionRespons
 @click.pass_context
 @click.argument("prompt", type=str, required=True)
 @click.option("--model", type=str, required=True, help="Model name")
-@click.option("--no-stream", is_flag=True, help="Disable streaming")
 @click.option("--max-tokens", type=int, help="Max tokens to generate")
 @click.option(
     "--stop", type=str, multiple=True, help="List of strings to stop generation"
@@ -22,6 +21,11 @@ from together.types.completions import CompletionChoicesChunk, CompletionRespons
 @click.option("--temperature", type=float, help="Sampling temperature")
 @click.option("--top-p", type=int, help="Top p sampling")
 @click.option("--top-k", type=float, help="Top k sampling")
+@click.option("--repetition-penalty", type=float, help="Repetition penalty")
+@click.option("--presence-penalty", type=float, help="Presence penalty")
+@click.option("--frequency-penalty", type=float, help="Frequency penalty")
+@click.option("--min-p", type=float, help="Minimum p")
+@click.option("--no-stream", is_flag=True, help="Disable streaming")
 @click.option("--logprobs", type=int, help="Return logprobs. Only works with --raw.")
 @click.option("--echo", is_flag=True, help="Echo prompt. Only works with --raw.")
 @click.option("--n", type=int, help="Number of output generations")
@@ -37,6 +41,9 @@ def completions(
     top_p: float | None = None,
     top_k: int | None = None,
     repetition_penalty: float | None = None,
+    presence_penalty: float | None = None,
+    frequency_penalty: float | None = None,
+    min_p: float | None = None,
     no_stream: bool = False,
     logprobs: int | None = None,
     echo: bool | None = None,
@@ -56,6 +63,9 @@ def completions(
         max_tokens=max_tokens,
         stop=stop,
         repetition_penalty=repetition_penalty,
+        presence_penalty=presence_penalty,
+        frequency_penalty=frequency_penalty,
+        min_p=min_p,
         stream=not no_stream,
         logprobs=logprobs,
         echo=echo,

{together-1.1.2 → together-1.1.4}/src/together/cli/api/finetune.py

@@ -107,10 +107,20 @@ def retrieve(ctx: click.Context, fine_tune_id: str) -> None:
 @fine_tuning.command()
 @click.pass_context
 @click.argument("fine_tune_id", type=str, required=True)
-def cancel(ctx: click.Context, fine_tune_id: str) -> None:
+@click.option(
+    "--quiet", is_flag=True, help="Do not prompt for confirmation before cancelling job"
+)
+def cancel(ctx: click.Context, fine_tune_id: str, quiet: bool = False) -> None:
     """Cancel fine-tuning job"""
     client: Together = ctx.obj
-
+    if not quiet:
+        confirm_response = input(
+            "You will be billed for any completed training steps upon cancellation. "
+            f"Do you want to cancel job {fine_tune_id}? [y/N]"
+        )
+        if "y" not in confirm_response.lower():
+            click.echo({"status": "Cancel not submitted"})
+            return
     response = client.fine_tuning.cancel(fine_tune_id)
 
     click.echo(json.dumps(response.model_dump(), indent=4))

{together-1.1.2 → together-1.1.4}/src/together/resources/chat/completions.py

@@ -28,6 +28,10 @@ class ChatCompletions:
         top_p: float | None = None,
         top_k: int | None = None,
         repetition_penalty: float | None = None,
+        presence_penalty: float | None = None,
+        frequency_penalty: float | None = None,
+        min_p: float | None = None,
+        logit_bias: Dict[str, float] | None = None,
         stream: bool = False,
         logprobs: int | None = None,
         echo: bool | None = None,
@@ -59,6 +63,21 @@ class ChatCompletions:
             repetition_penalty (float, optional): A number that controls the diversity of generated text
                     by reducing the likelihood of repeated sequences. Higher values decrease repetition.
                 Defaults to None.
+            presence_penalty (float, optional): A number that controls the likelihood of tokens based on if they have
+                    appeared in the text. Positive values decrease the likelihood of repeated tokens or phrases.
+                    Must be in the range [-2, 2].
+                Defaults to None.
+            frequency_penalty (float, optional): A number that controls the likelihood of tokens based on the frequency
+                    of their appearance in the text. Positive decrease the likelihood of repeated tokens or phrases.
+                    Must be in the range [-2, 2].
+                Defaults to None.
+            min_p (float, optional): A number that controls the minimum percentage value that a token must reach to
+                be considered during sampling.
+                Must be in the range [0, 1].
+                Defaults to None.
+            logit_bias (Dict[str, float], optional): A dictionary of tokens and their bias values that modify the
+                likelihood of specific tokens being sampled. Bias values must be in the range [-100, 100].
+                Defaults to None.
             stream (bool, optional): Flag indicating whether to stream the generated completions.
                 Defaults to False.
             logprobs (int, optional): Number of top-k logprobs to return
@@ -100,6 +119,10 @@ class ChatCompletions:
             max_tokens=max_tokens,
             stop=stop,
             repetition_penalty=repetition_penalty,
+            presence_penalty=presence_penalty,
+            frequency_penalty=frequency_penalty,
+            min_p=min_p,
+            logit_bias=logit_bias,
             stream=stream,
             logprobs=logprobs,
             echo=echo,
@@ -142,6 +165,10 @@ class AsyncChatCompletions:
         top_p: float | None = None,
         top_k: int | None = None,
         repetition_penalty: float | None = None,
+        presence_penalty: float | None = None,
+        frequency_penalty: float | None = None,
+        min_p: float | None = None,
+        logit_bias: Dict[str, float] | None = None,
         stream: bool = False,
         logprobs: int | None = None,
         echo: bool | None = None,
@@ -173,6 +200,21 @@ class AsyncChatCompletions:
             repetition_penalty (float, optional): A number that controls the diversity of generated text
                     by reducing the likelihood of repeated sequences. Higher values decrease repetition.
                 Defaults to None.
+            presence_penalty (float, optional): A number that controls the likelihood of tokens based on if they have
+                    appeared in the text. Positive values decrease the likelihood of repeated tokens or phrases.
+                    Must be in the range [-2, 2].
+                Defaults to None.
+            frequency_penalty (float, optional): A number that controls the likelihood of tokens based on the frequency
+                    of their appearance in the text. Positive decrease the likelihood of repeated tokens or phrases.
+                    Must be in the range [-2, 2].
+                Defaults to None.
+            min_p (float, optional): A number that controls the minimum percentage value that a token must reach to
+                be considered during sampling.
+                Must be in the range [0, 1].
+                Defaults to None.
+            logit_bias (Dict[str, float], optional): A dictionary of tokens and their bias values that modify the
+                likelihood of specific tokens being sampled. Bias values must be in the range [-100, 100].
+                Defaults to None.
             stream (bool, optional): Flag indicating whether to stream the generated completions.
                 Defaults to False.
             logprobs (int, optional): Number of top-k logprobs to return
@@ -214,6 +256,10 @@ class AsyncChatCompletions:
             max_tokens=max_tokens,
             stop=stop,
             repetition_penalty=repetition_penalty,
+            presence_penalty=presence_penalty,
+            frequency_penalty=frequency_penalty,
+            min_p=min_p,
+            logit_bias=logit_bias,
             stream=stream,
             logprobs=logprobs,
             echo=echo,

{together-1.1.2 → together-1.1.4}/src/together/resources/completions.py

@@ -1,6 +1,6 @@
 from __future__ import annotations
 
-from typing import AsyncGenerator, Iterator, List
+from typing import AsyncGenerator, Dict, Iterator, List
 
 from together.abstract import api_requestor
 from together.together_response import TogetherResponse
@@ -28,6 +28,10 @@ class Completions:
         top_p: float | None = None,
         top_k: int | None = None,
         repetition_penalty: float | None = None,
+        presence_penalty: float | None = None,
+        frequency_penalty: float | None = None,
+        min_p: float | None = None,
+        logit_bias: Dict[str, float] | None = None,
         stream: bool = False,
         logprobs: int | None = None,
         echo: bool | None = None,
@@ -55,6 +59,21 @@ class Completions:
             repetition_penalty (float, optional): A number that controls the diversity of generated text
                     by reducing the likelihood of repeated sequences. Higher values decrease repetition.
                 Defaults to None.
+            presence_penalty (float, optional): A number that controls the likelihood of tokens based on if they have
+                    appeared in the text. Positive values decrease the likelihood of repeated tokens or phrases.
+                    Must be in the range [-2, 2].
+                Defaults to None.
+            frequency_penalty (float, optional): A number that controls the likelihood of tokens based on the frequency
+                    of their appearance in the text. Positive decrease the likelihood of repeated tokens or phrases.
+                    Must be in the range [-2, 2].
+                Defaults to None.
+            min_p (float, optional): A number that controls the minimum percentage value that a token must reach to
+                be considered during sampling.
+                Must be in the range [0, 1].
+                Defaults to None.
+            logit_bias (Dict[str, float], optional): A dictionary of tokens and their bias values that modify the
+                likelihood of specific tokens being sampled. Bias values must be in the range [-100, 100].
+                Defaults to None.
             stream (bool, optional): Flag indicating whether to stream the generated completions.
                 Defaults to False.
             logprobs (int, optional): Number of top-k logprobs to return
@@ -85,6 +104,10 @@ class Completions:
             max_tokens=max_tokens,
             stop=stop,
             repetition_penalty=repetition_penalty,
+            presence_penalty=presence_penalty,
+            frequency_penalty=frequency_penalty,
+            min_p=min_p,
+            logit_bias=logit_bias,
             stream=stream,
             logprobs=logprobs,
             echo=echo,
@@ -124,6 +147,10 @@ class AsyncCompletions:
         top_p: float | None = None,
         top_k: int | None = None,
         repetition_penalty: float | None = None,
+        presence_penalty: float | None = None,
+        frequency_penalty: float | None = None,
+        min_p: float | None = None,
+        logit_bias: Dict[str, float] | None = None,
         stream: bool = False,
         logprobs: int | None = None,
         echo: bool | None = None,
@@ -151,6 +178,21 @@ class AsyncCompletions:
             repetition_penalty (float, optional): A number that controls the diversity of generated text
                     by reducing the likelihood of repeated sequences. Higher values decrease repetition.
                 Defaults to None.
+            presence_penalty (float, optional): A number that controls the likelihood of tokens based on if they have
+                    appeared in the text. Positive values decrease the likelihood of repeated tokens or phrases.
+                    Must be in the range [-2, 2].
+                Defaults to None.
+            frequency_penalty (float, optional): A number that controls the likelihood of tokens based on the frequency
+                    of their appearance in the text. Positive decrease the likelihood of repeated tokens or phrases.
+                    Must be in the range [-2, 2].
+                Defaults to None.
+            min_p (float, optional): A number that controls the minimum percentage value that a token must reach to
+                be considered during sampling.
+                Must be in the range [0, 1].
+                Defaults to None.
+            logit_bias (Dict[str, float], optional): A dictionary of tokens and their bias values that modify the
+                likelihood of specific tokens being sampled. Bias values must be in the range [-100, 100].
+                Defaults to None.
             stream (bool, optional): Flag indicating whether to stream the generated completions.
                 Defaults to False.
             logprobs (int, optional): Number of top-k logprobs to return
@@ -181,6 +223,10 @@ class AsyncCompletions:
             max_tokens=max_tokens,
             stop=stop,
             repetition_penalty=repetition_penalty,
+            presence_penalty=presence_penalty,
+            frequency_penalty=frequency_penalty,
+            min_p=min_p,
+            logit_bias=logit_bias,
             stream=stream,
             logprobs=logprobs,
             echo=echo,

{together-1.1.2 → together-1.1.4}/src/together/types/chat_completions.py

@@ -1,9 +1,11 @@
 from __future__ import annotations
 
+import warnings
 from enum import Enum
 from typing import Any, Dict, List
 
-from pydantic import Field
+from pydantic import Field, model_validator
+from typing_extensions import Self
 
 from together.types.abstract import BaseModel
 from together.types.common import (
@@ -20,6 +22,7 @@ class MessageRole(str, Enum):
     ASSISTANT = "assistant"
     SYSTEM = "system"
     USER = "user"
+    TOOL = "tool"
 
 
 class ResponseFormatType(str, Enum):
@@ -86,6 +89,10 @@ class ChatCompletionRequest(BaseModel):
     top_p: float | None = None
     top_k: int | None = None
     repetition_penalty: float | None = None
+    presence_penalty: float | None = None
+    frequency_penalty: float | None = None
+    min_p: float | None = None
+    logit_bias: Dict[str, float] | None = None
     # stream SSE token chunks
     stream: bool = False
     # return logprobs
@@ -102,6 +109,16 @@ class ChatCompletionRequest(BaseModel):
     tools: List[Tools] | None = None
     tool_choice: ToolChoice | ToolChoiceEnum | None = None
 
+    # Raise warning if repetition_penalty is used with presence_penalty or frequency_penalty
+    @model_validator(mode="after")
+    def verify_parameters(self) -> Self:
+        if self.repetition_penalty:
+            if self.presence_penalty or self.frequency_penalty:
+                warnings.warn(
+                    "repetition_penalty is not advisable to be used alongside presence_penalty or frequency_penalty"
+                )
+        return self
+
 
 class ChatCompletionChoicesData(BaseModel):
     index: int | None = None

{together-1.1.2 → together-1.1.4}/src/together/types/completions.py

@@ -1,6 +1,10 @@
 from __future__ import annotations
 
-from typing import List
+import warnings
+from typing import Dict, List
+
+from pydantic import model_validator
+from typing_extensions import Self
 
 from together.types.abstract import BaseModel
 from together.types.common import (
@@ -27,6 +31,10 @@ class CompletionRequest(BaseModel):
     top_p: float | None = None
     top_k: int | None = None
     repetition_penalty: float | None = None
+    presence_penalty: float | None = None
+    frequency_penalty: float | None = None
+    min_p: float | None = None
+    logit_bias: Dict[str, float] | None = None
     # stream SSE token chunks
     stream: bool = False
     # return logprobs
@@ -39,6 +47,16 @@ class CompletionRequest(BaseModel):
     # moderation model
     safety_model: str | None = None
 
+    # Raise warning if repetition_penalty is used with presence_penalty or frequency_penalty
+    @model_validator(mode="after")
+    def verify_parameters(self) -> Self:
+        if self.repetition_penalty:
+            if self.presence_penalty or self.frequency_penalty:
+                warnings.warn(
+                    "repetition_penalty is not advisable to be used alongside presence_penalty or frequency_penalty"
+                )
+        return self
+
 
 class CompletionChoicesData(BaseModel):
     index: int