mostlyai-engine 1.1.4__tar.gz → 1.1.6__tar.gz

{mostlyai_engine-1.1.4 → mostlyai_engine-1.1.6}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: mostlyai-engine
-Version: 1.1.4
+Version: 1.1.6
 Summary: Synthetic Data Engine
 Project-URL: homepage, https://github.com/mostly-ai/mostlyai-engine
 Project-URL: repository, https://github.com/mostly-ai/mostlyai-engine
@@ -45,15 +45,22 @@ Provides-Extra: cpu
 Requires-Dist: torch==2.5.1+cpu; (sys_platform == 'linux') and extra == 'cpu'
 Requires-Dist: torch>=2.5.1; (sys_platform != 'linux') and extra == 'cpu'
 Provides-Extra: gpu
-Requires-Dist: bitsandbytes<0.44,>=0.43.3; extra == 'gpu'
+Requires-Dist: bitsandbytes<0.43,>=0.42.0; (sys_platform == 'darwin') and extra == 'gpu'
+Requires-Dist: bitsandbytes<0.44,>=0.43.3; (sys_platform == 'linux') and extra == 'gpu'
 Requires-Dist: torch>=2.5.1; extra == 'gpu'
-Requires-Dist: vllm<0.8,>=0.7.2; extra == 'gpu'
+Requires-Dist: vllm<0.7.3,>=0.7.2; (sys_platform == 'linux') and extra == 'gpu'
+Requires-Dist: vllm<0.8,>=0.7.3; (sys_platform == 'darwin') and extra == 'gpu'
 Description-Content-Type: text/markdown
 
 # Synthetic Data Engine 💎
-[![Documentation](https://img.shields.io/badge/docs-latest-green)](https://mostly-ai.github.io/mostlyai-engine/) ![license](https://img.shields.io/github/license/mostly-ai/mostlyai-engine) ![GitHub Release](https://img.shields.io/github/v/release/mostly-ai/mostlyai-engine) ![PyPI - Python Version](https://img.shields.io/pypi/pyversions/mostlyai-engine) [![stats](https://pepy.tech/badge/mostlyai-engine)](https://pypi.org/project/mostlyai-engine/)
 
-[Package Documentation](https://mostly-ai.github.io/mostlyai-engine/) | [Platform Documentation](https://mostly.ai/docs)
+![GitHub Release](https://img.shields.io/github/v/release/mostly-ai/mostlyai-engine)
+[![Documentation](https://img.shields.io/badge/docs-latest-green)](https://mostly-ai.github.io/mostlyai-engine/)
+[![stats](https://pepy.tech/badge/mostlyai-engine)](https://pypi.org/project/mostlyai-engine/)
+![license](https://img.shields.io/github/license/mostly-ai/mostlyai-engine)
+![PyPI - Python Version](https://img.shields.io/pypi/pyversions/mostlyai-engine)
+
+[Documentation](https://mostly-ai.github.io/mostlyai-engine/) | [Technical Paper](https://arxiv.org/abs/2501.12012) | [Free Cloud Service](https://app.mostly.ai/)
 
 Create high-fidelity privacy-safe synthetic data:
 
@@ -69,7 +76,7 @@ Create high-fidelity privacy-safe synthetic data:
 
 ...all within your safe compute environment, all with a few lines of Python code 💥.
 
-Note: This library is the underlying model engine of the [Synthetic Data SDK ✨](https://github.com/mostly-ai/mostlyai). Please refer to the latter, for an easy-to-use, higher-level software toolkit.
+Note: This library is the underlying model engine of the [Synthetic Data SDK](https://github.com/mostly-ai/mostlyai). Please refer to the latter, for an easy-to-use, higher-level software toolkit.
 
 
 ## Installation

{mostlyai_engine-1.1.4 → mostlyai_engine-1.1.6}/README.md

@@ -1,7 +1,12 @@
 # Synthetic Data Engine 💎
-[![Documentation](https://img.shields.io/badge/docs-latest-green)](https://mostly-ai.github.io/mostlyai-engine/) ![license](https://img.shields.io/github/license/mostly-ai/mostlyai-engine) ![GitHub Release](https://img.shields.io/github/v/release/mostly-ai/mostlyai-engine) ![PyPI - Python Version](https://img.shields.io/pypi/pyversions/mostlyai-engine) [![stats](https://pepy.tech/badge/mostlyai-engine)](https://pypi.org/project/mostlyai-engine/)
 
-[Package Documentation](https://mostly-ai.github.io/mostlyai-engine/) | [Platform Documentation](https://mostly.ai/docs)
+![GitHub Release](https://img.shields.io/github/v/release/mostly-ai/mostlyai-engine)
+[![Documentation](https://img.shields.io/badge/docs-latest-green)](https://mostly-ai.github.io/mostlyai-engine/)
+[![stats](https://pepy.tech/badge/mostlyai-engine)](https://pypi.org/project/mostlyai-engine/)
+![license](https://img.shields.io/github/license/mostly-ai/mostlyai-engine)
+![PyPI - Python Version](https://img.shields.io/pypi/pyversions/mostlyai-engine)
+
+[Documentation](https://mostly-ai.github.io/mostlyai-engine/) | [Technical Paper](https://arxiv.org/abs/2501.12012) | [Free Cloud Service](https://app.mostly.ai/)
 
 Create high-fidelity privacy-safe synthetic data:
 
@@ -17,7 +22,7 @@ Create high-fidelity privacy-safe synthetic data:
 
 ...all within your safe compute environment, all with a few lines of Python code 💥.
 
-Note: This library is the underlying model engine of the [Synthetic Data SDK ✨](https://github.com/mostly-ai/mostlyai). Please refer to the latter, for an easy-to-use, higher-level software toolkit.
+Note: This library is the underlying model engine of the [Synthetic Data SDK](https://github.com/mostly-ai/mostlyai). Please refer to the latter, for an easy-to-use, higher-level software toolkit.
 
 
 ## Installation

{mostlyai_engine-1.1.4 → mostlyai_engine-1.1.6}/mostlyai/engine/__init__.py

@@ -22,7 +22,7 @@ from mostlyai.engine.generation import generate
 
 
 __all__ = ["split", "analyze", "encode", "train", "generate", "init_logging"]
-__version__ = "1.1.4"
+__version__ = "1.1.6"
 
 # suppress specific warning related to os.fork() in multi-threaded processes
 warnings.filterwarnings("ignore", category=DeprecationWarning, message=".*multi-threaded.*fork.*")

{mostlyai_engine-1.1.4 → mostlyai_engine-1.1.6}/mostlyai/engine/_language/generation.py

@@ -13,8 +13,10 @@
 # limitations under the License.
 
 import contextlib
+import importlib
 import json
 import os
+import platform
 
 import json_repair
 import logging
@@ -23,7 +25,6 @@ from pathlib import Path
 
 import pandas as pd
 import torch
-from huggingface_hub import constants as hf_constants
 from transformers import (
     PreTrainedTokenizerBase,
 )
@@ -101,15 +102,10 @@ def decode_buffered_samples(
     tgt_data = tgt_data.map(
         lambda x: x.encode("utf-8", errors="surrogatepass").decode("utf-8", errors="replace") if not pd.isna(x) else x
     )
-    # overwrite generated columns with the seeded values
-    tgt_data.update(tgt_seed)
 
     # prepend the context keys to the data (if not dummy context)
     if ctx_keys.name != DUMMY_CONTEXT_KEY:
         tgt_data = pd.concat([ctx_keys, tgt_data], axis=1)
-    invalid_percentage = ((tgt_data[tgt_stats["columns"].keys()] == INVALID_VALUE).sum() / len(tgt_data) * 100.0).map(
-        "{:.2f}%".format
-    )
 
     for col in tgt_stats["columns"].keys():
         col_stats = tgt_stats["columns"][col]
@@ -122,6 +118,12 @@ def decode_buffered_samples(
         else:
             tgt_data[col] = decode_text(tgt_data[col], col_stats)
 
+    # overwrite generated columns with the seeded values
+    tgt_data.update(tgt_seed)
+
+    invalid_percentage = ((tgt_data[tgt_stats["columns"].keys()] == INVALID_VALUE).sum() / len(tgt_data) * 100.0).map(
+        "{:.2f}%".format
+    )
     _LOG.info(f"percentage of invalid values: {invalid_percentage.to_dict()}")
     _LOG.info(f"decoded {tgt_data.shape} from {len(buffer.buffer)} batches in {time.time() - t0:.2f}s")
     return tgt_data
@@ -236,14 +238,14 @@ def generate(
         _LOG.info(f"{max_new_tokens=}")
 
         t0 = time.time()
-        hf_constants.HF_HUB_OFFLINE = (
-            False  # needed for gated hf models that are not sharded, otherwise GatedRepoError in vLLM
-        )
-        # set the default env var so that we don't pass it explicitly to vLLM
-        os.environ["HF_TOKEN"] = os.getenv("MOSTLY_HUGGING_FACE_TOKEN") or os.getenv("HF_TOKEN", "")
+        # use MOSTLY_HUGGING_FACE_TOKEN if available, otherwise HF_TOKEN should be unset or with a pre-set value as is
+        if os.getenv("MOSTLY_HUGGING_FACE_TOKEN"):
+            os.environ["HF_TOKEN"] = os.environ["MOSTLY_HUGGING_FACE_TOKEN"]
 
         is_peft_adapter = (workspace.model_path / "adapter_config.json").exists()
-        if is_peft_adapter and device.type == "cuda":
+        if is_peft_adapter and (
+            device.type == "cuda" or (platform.system() == "Darwin" and importlib.util.find_spec("vllm") is not None)
+        ):
             from mostlyai.engine._language.engine.vllm_engine import VLLMEngine
 
             engine = VLLMEngine(workspace.model_path, device, max_new_tokens, MAX_LENGTH)

{mostlyai_engine-1.1.4 → mostlyai_engine-1.1.6}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "mostlyai-engine"
-version = "1.1.4"
+version = "1.1.6"
 description = "Synthetic Data Engine"
 authors = [{ name = "MOSTLY AI", email = "dev@mostly.ai" }]
 requires-python = ">=3.10"
@@ -47,8 +47,10 @@ dependencies = [
 [project.optional-dependencies]
 gpu = [
     "torch>=2.5.1",
-    "bitsandbytes>=0.43.3,<0.44",
-    "vllm>=0.7.2,<0.8",
+    "bitsandbytes>=0.42.0,<0.43; sys_platform == 'darwin'",
+    "vllm>=0.7.3,<0.8; sys_platform == 'darwin'",
+    "bitsandbytes>=0.43.3,<0.44; sys_platform == 'linux'",
+    "vllm>=0.7.2,<0.7.3; sys_platform == 'linux'",
 ]
 cpu = [
     "torch==2.5.1+cpu; sys_platform == 'linux'",