lemonade-sdk 7.0.4__tar.gz → 8.0.0__tar.gz

{lemonade_sdk-7.0.4/src/lemonade_sdk.egg-info → lemonade_sdk-8.0.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: lemonade-sdk
-Version: 7.0.4
+Version: 8.0.0
 Summary: Lemonade SDK: Your LLM Aide for Validation and Deployment
 Author-email: lemonade@amd.com
 Requires-Python: >=3.10, <3.12
@@ -9,36 +9,42 @@ License-File: LICENSE
 License-File: NOTICE.md
 Requires-Dist: invoke>=2.0.0
 Requires-Dist: onnx<1.18.0,>=1.11.0
-Requires-Dist: torch>=1.12.1
 Requires-Dist: pyyaml>=5.4
 Requires-Dist: typeguard>=2.3.13
 Requires-Dist: packaging>=20.9
 Requires-Dist: numpy<2.0.0
-Requires-Dist: pandas>=1.5.3
 Requires-Dist: fasteners
 Requires-Dist: GitPython>=3.1.40
 Requires-Dist: psutil>=6.1.1
 Requires-Dist: wmi
+Requires-Dist: py-cpuinfo
 Requires-Dist: pytz
 Requires-Dist: zstandard
-Requires-Dist: matplotlib
+Requires-Dist: fastapi
+Requires-Dist: uvicorn[standard]
+Requires-Dist: openai>=1.81.0
+Requires-Dist: transformers<=4.51.3
+Requires-Dist: jinja2
 Requires-Dist: tabulate
 Requires-Dist: huggingface-hub==0.30.2
+Provides-Extra: oga-hybrid-minimal
+Requires-Dist: onnx==1.16.1; extra == "oga-hybrid-minimal"
+Requires-Dist: numpy==1.26.4; extra == "oga-hybrid-minimal"
+Requires-Dist: protobuf>=6.30.1; extra == "oga-hybrid-minimal"
+Provides-Extra: oga-cpu-minimal
+Requires-Dist: onnxruntime-genai==0.6.0; extra == "oga-cpu-minimal"
+Requires-Dist: onnxruntime<1.22.0,>=1.10.1; extra == "oga-cpu-minimal"
 Provides-Extra: llm
 Requires-Dist: torch>=2.6.0; extra == "llm"
-Requires-Dist: transformers<=4.51.3; extra == "llm"
 Requires-Dist: accelerate; extra == "llm"
-Requires-Dist: py-cpuinfo; extra == "llm"
 Requires-Dist: sentencepiece; extra == "llm"
 Requires-Dist: datasets; extra == "llm"
+Requires-Dist: pandas>=1.5.3; extra == "llm"
+Requires-Dist: matplotlib; extra == "llm"
 Requires-Dist: human-eval-windows==1.0.4; extra == "llm"
-Requires-Dist: fastapi; extra == "llm"
-Requires-Dist: uvicorn[standard]; extra == "llm"
-Requires-Dist: openai>=1.81.0; extra == "llm"
 Requires-Dist: lm-eval[api]; extra == "llm"
 Provides-Extra: llm-oga-cpu
-Requires-Dist: onnxruntime-genai==0.6.0; extra == "llm-oga-cpu"
-Requires-Dist: onnxruntime<1.22.0,>=1.10.1; extra == "llm-oga-cpu"
+Requires-Dist: lemonade-sdk[oga-cpu-minimal]; extra == "llm-oga-cpu"
 Requires-Dist: lemonade-sdk[llm]; extra == "llm-oga-cpu"
 Provides-Extra: llm-oga-igpu
 Requires-Dist: onnxruntime-genai-directml==0.6.0; extra == "llm-oga-igpu"
@@ -57,9 +63,7 @@ Requires-Dist: numpy==1.26.4; extra == "llm-oga-npu"
 Requires-Dist: protobuf>=6.30.1; extra == "llm-oga-npu"
 Requires-Dist: lemonade-sdk[llm]; extra == "llm-oga-npu"
 Provides-Extra: llm-oga-hybrid
-Requires-Dist: onnx==1.16.1; extra == "llm-oga-hybrid"
-Requires-Dist: numpy==1.26.4; extra == "llm-oga-hybrid"
-Requires-Dist: protobuf>=6.30.1; extra == "llm-oga-hybrid"
+Requires-Dist: lemonade-sdk[oga-hybrid-minimal]; extra == "llm-oga-hybrid"
 Requires-Dist: lemonade-sdk[llm]; extra == "llm-oga-hybrid"
 Provides-Extra: llm-oga-unified
 Requires-Dist: lemonade-sdk[llm-oga-hybrid]; extra == "llm-oga-unified"
@@ -78,24 +82,82 @@ Dynamic: summary
 
 ## 🍋 Lemonade SDK: Quickly serve, benchmark and deploy LLMs
 
-The [Lemonade SDK](./docs/README.md) is designed to make it easy to serve, benchmark, and deploy large language models (LLMs) on a variety of hardware platforms, including CPU, GPU, and NPU. 
+The [Lemonade SDK](./docs/README.md) makes it easy to run Large Language Models (LLMs) on your PC. Our focus is using the best tools, such as neural processing units (NPUs) and Vulkan GPU acceleration, to maximize LLM speed and responsiveness.
 
 <div align="center">
   <img src="https://download.amd.com/images/lemonade_640x480_1.gif" alt="Lemonade Demo" title="Lemonade in Action">
 </div>
 
+### Features
+
 The [Lemonade SDK](./docs/README.md) is comprised of the following:
 
-- 🌐 **Lemonade Server**: A server interface that uses the standard Open AI API, allowing applications to integrate with local LLMs.
-- 🐍 **Lemonade Python API**: Offers High-Level API for easy integration of Lemonade LLMs into Python applications and Low-Level API for custom experiments.
-- 🖥️ **Lemonade CLI**: The `lemonade` CLI lets you mix-and-match LLMs, frameworks (PyTorch, ONNX, GGUF), and measurement tools to run experiments. The available tools are:
-  - Prompting an LLM.
-  - Measuring the accuracy of an LLM using a variety of tests.
-  - Benchmarking an LLM to get the time-to-first-token and tokens per second.
-  - Profiling the memory usage of an LLM.
+- 🌐 **[Lemonade Server](https://lemonade-server.ai/docs)**: A local LLM server for running ONNX and GGUF models using the OpenAI API standard. Install and enable your applications with NPU and GPU acceleration in minutes.
+- 🐍 **Lemonade API**: High-level Python API to directly integrate Lemonade LLMs into Python applications.
+- 🖥️ **Lemonade CLI**: The `lemonade` CLI lets you mix-and-match LLMs (ONNX, GGUF, SafeTensors) with measurement tools to characterize your models on your hardware. The available tools are:
+  - Prompting with templates.
+  - Measuring accuracy with a variety of tests.
+  - Benchmarking to get the time-to-first-token and tokens per second.
+  - Profiling the memory utilization.
 
 ### [Click here to get started with Lemonade.](./docs/README.md)
 
+### Supported Configurations
+
+Maximum LLM performance requires the right hardware accelerator with the right inference engine for your scenario. Lemonade supports the following configurations, while also making it easy to switch between them at runtime.
+
+<table border="1" cellpadding="6" cellspacing="0">
+  <thead>
+    <tr>
+      <th rowspan="2">Hardware</th>
+      <th colspan="3" align="center">🛠️ Engine Support</th>
+      <th colspan="2" align="center">🖥️ OS (x86/x64)</th>
+    </tr>
+    <tr>
+      <th align="center">OGA</th>
+      <th align="center">llamacpp</th>
+      <th align="center">HF</th>
+      <th align="center">Windows</th>
+      <th align="center">Linux</th>
+    </tr>
+  </thead>
+  <tbody>
+    <tr>
+      <td>🧠 CPU</td>
+      <td align="center">All platforms</td>
+      <td align="center">All platforms</td>
+      <td align="center">All platforms</td>
+      <td align="center">✅</td>
+      <td align="center">✅</td>
+    </tr>
+    <tr>
+      <td>🎮 GPU</td>
+      <td align="center">—</td>
+      <td align="center">Vulkan: All platforms<br><small>Focus: Radeon™ 7000/9000</small></td>
+      <td align="center">—</td>
+      <td align="center">✅</td>
+      <td align="center">✅</td>
+    </tr>
+    <tr>
+      <td>🤖 NPU</td>
+      <td align="center">AMD Ryzen™ AI 300 series</td>
+      <td align="center">—</td>
+      <td align="center">—</td>
+      <td align="center">✅</td>
+      <td align="center">—</td>
+    </tr>
+  </tbody>
+</table>
+
+
+
+#### Inference Engines Overview
+| Engine | Description |
+| :--- | :--- |
+| **OnnxRuntime GenAI (OGA)** | Microsoft engine that runs `.onnx` models and enables hardware vendors to provide their own execution providers (EPs) to support specialized hardware, such as neural processing units (NPUs). |
+| **llamacpp** | Community-driven engine with strong GPU acceleration, support for thousands of `.gguf` models, and advanced features such as vision-language models (VLMs) and mixture-of-experts (MoEs). |
+| **Hugging Face (HF)** | Hugging Face's `transformers` library can run the original `.safetensors` trained weights for models on Meta's PyTorch engine, which provides a source of truth for accuracy measurement. |
+
 ## Contributing
 
 We are actively seeking collaborators from across the industry. If you would like to contribute to this project, please check out our [contribution guide](./docs/contribute.md).

lemonade_sdk-8.0.0/README.md

@@ -0,0 +1,97 @@
+[![Lemonade tests](https://github.com/lemonade-sdk/lemonade/actions/workflows/test_lemonade.yml/badge.svg)](https://github.com/lemonade-sdk/lemonade/tree/main/test "Check out our tests")
+[![OS - Windows | Linux](https://img.shields.io/badge/OS-windows%20%7C%20linux-blue)](docs/README.md#installation "Check out our instructions")
+[![Made with Python](https://img.shields.io/badge/Python-3.8,3.10-blue?logo=python&logoColor=white)](docs/README.md#installation "Check out our instructions")
+
+## 🍋 Lemonade SDK: Quickly serve, benchmark and deploy LLMs
+
+The [Lemonade SDK](./docs/README.md) makes it easy to run Large Language Models (LLMs) on your PC. Our focus is using the best tools, such as neural processing units (NPUs) and Vulkan GPU acceleration, to maximize LLM speed and responsiveness.
+
+<div align="center">
+  <img src="https://download.amd.com/images/lemonade_640x480_1.gif" alt="Lemonade Demo" title="Lemonade in Action">
+</div>
+
+### Features
+
+The [Lemonade SDK](./docs/README.md) is comprised of the following:
+
+- 🌐 **[Lemonade Server](https://lemonade-server.ai/docs)**: A local LLM server for running ONNX and GGUF models using the OpenAI API standard. Install and enable your applications with NPU and GPU acceleration in minutes.
+- 🐍 **Lemonade API**: High-level Python API to directly integrate Lemonade LLMs into Python applications.
+- 🖥️ **Lemonade CLI**: The `lemonade` CLI lets you mix-and-match LLMs (ONNX, GGUF, SafeTensors) with measurement tools to characterize your models on your hardware. The available tools are:
+  - Prompting with templates.
+  - Measuring accuracy with a variety of tests.
+  - Benchmarking to get the time-to-first-token and tokens per second.
+  - Profiling the memory utilization.
+
+### [Click here to get started with Lemonade.](./docs/README.md)
+
+### Supported Configurations
+
+Maximum LLM performance requires the right hardware accelerator with the right inference engine for your scenario. Lemonade supports the following configurations, while also making it easy to switch between them at runtime.
+
+<table border="1" cellpadding="6" cellspacing="0">
+  <thead>
+    <tr>
+      <th rowspan="2">Hardware</th>
+      <th colspan="3" align="center">🛠️ Engine Support</th>
+      <th colspan="2" align="center">🖥️ OS (x86/x64)</th>
+    </tr>
+    <tr>
+      <th align="center">OGA</th>
+      <th align="center">llamacpp</th>
+      <th align="center">HF</th>
+      <th align="center">Windows</th>
+      <th align="center">Linux</th>
+    </tr>
+  </thead>
+  <tbody>
+    <tr>
+      <td>🧠 CPU</td>
+      <td align="center">All platforms</td>
+      <td align="center">All platforms</td>
+      <td align="center">All platforms</td>
+      <td align="center">✅</td>
+      <td align="center">✅</td>
+    </tr>
+    <tr>
+      <td>🎮 GPU</td>
+      <td align="center">—</td>
+      <td align="center">Vulkan: All platforms<br><small>Focus: Radeon™ 7000/9000</small></td>
+      <td align="center">—</td>
+      <td align="center">✅</td>
+      <td align="center">✅</td>
+    </tr>
+    <tr>
+      <td>🤖 NPU</td>
+      <td align="center">AMD Ryzen™ AI 300 series</td>
+      <td align="center">—</td>
+      <td align="center">—</td>
+      <td align="center">✅</td>
+      <td align="center">—</td>
+    </tr>
+  </tbody>
+</table>
+
+
+
+#### Inference Engines Overview
+| Engine | Description |
+| :--- | :--- |
+| **OnnxRuntime GenAI (OGA)** | Microsoft engine that runs `.onnx` models and enables hardware vendors to provide their own execution providers (EPs) to support specialized hardware, such as neural processing units (NPUs). |
+| **llamacpp** | Community-driven engine with strong GPU acceleration, support for thousands of `.gguf` models, and advanced features such as vision-language models (VLMs) and mixture-of-experts (MoEs). |
+| **Hugging Face (HF)** | Hugging Face's `transformers` library can run the original `.safetensors` trained weights for models on Meta's PyTorch engine, which provides a source of truth for accuracy measurement. |
+
+## Contributing
+
+We are actively seeking collaborators from across the industry. If you would like to contribute to this project, please check out our [contribution guide](./docs/contribute.md).
+
+## Maintainers
+
+This project is sponsored by AMD. It is maintained by @danielholanda @jeremyfowers @ramkrishna @vgodsoe in equal measure. You can reach us by filing an [issue](https://github.com/lemonade-sdk/lemonade/issues) or email [lemonade@amd.com](mailto:lemonade@amd.com).
+
+## License
+
+This project is licensed under the [Apache 2.0 License](https://github.com/lemonade-sdk/lemonade/blob/main/LICENSE). Portions of the project are licensed as described in [NOTICE.md](./NOTICE.md).
+
+<!--This file was originally licensed under Apache 2.0. It has been modified.
+Modifications Copyright (c) 2025 AMD-->
+

{lemonade_sdk-7.0.4 → lemonade_sdk-8.0.0}/setup.py

@@ -14,54 +14,73 @@ setup(
         "lemonade.profilers",
         "lemonade.common",
         "lemonade.tools",
-        "lemonade.tools.ort_genai",
+        "lemonade.tools.huggingface",
+        "lemonade.tools.oga",
+        "lemonade.tools.llamacpp",
         "lemonade.tools.quark",
         "lemonade.tools.report",
+        "lemonade.tools.server.utils",
         "lemonade.tools.server",
         "lemonade_install",
         "lemonade_server",
     ],
     install_requires=[
+        # Minimal dependencies required for end-users who are running
+        # apps deployed on Lemonade SDK
         "invoke>=2.0.0",
         "onnx>=1.11.0,<1.18.0",
-        "torch>=1.12.1",
         "pyyaml>=5.4",
         "typeguard>=2.3.13",
         "packaging>=20.9",
         # Necessary until upstream packages account for the breaking
         # change to numpy
         "numpy<2.0.0",
-        "pandas>=1.5.3",
         "fasteners",
         "GitPython>=3.1.40",
         "psutil>=6.1.1",
         "wmi",
+        "py-cpuinfo",
         "pytz",
         "zstandard",
-        "matplotlib",
+        "fastapi",
+        "uvicorn[standard]",
+        "openai>=1.81.0",
+        "transformers<=4.51.3",
+        "jinja2",
         "tabulate",
         # huggingface-hub==0.31.0 introduces a new transfer protocol that was causing us issues
         "huggingface-hub==0.30.2",
     ],
     extras_require={
+        # The -minimal extras are meant to deploy specific backends into end-user
+        # applications, without including developer-focused tools
+        "oga-hybrid-minimal": [
+            # Note: `lemonade-install --ryzenai hybrid` is necessary
+            # to complete installation
+            "onnx==1.16.1",
+            "numpy==1.26.4",
+            "protobuf>=6.30.1",
+        ],
+        "oga-cpu-minimal": [
+            "onnxruntime-genai==0.6.0",
+            "onnxruntime >=1.10.1,<1.22.0",
+        ],
         "llm": [
+            # Minimal dependencies for developers to use all features of
+            # Lemonade SDK, including building and optimizing models
             "torch>=2.6.0",
-            "transformers<=4.51.3",
             "accelerate",
-            "py-cpuinfo",
             "sentencepiece",
             "datasets",
+            "pandas>=1.5.3",
+            "matplotlib",
             # Install human-eval from a forked repo with Windows support until the
             # PR (https://github.com/openai/human-eval/pull/53) is merged
             "human-eval-windows==1.0.4",
-            "fastapi",
-            "uvicorn[standard]",
-            "openai>=1.81.0",
             "lm-eval[api]",
         ],
         "llm-oga-cpu": [
-            "onnxruntime-genai==0.6.0",
-            "onnxruntime >=1.10.1,<1.22.0",
+            "lemonade-sdk[oga-cpu-minimal]",
             "lemonade-sdk[llm]",
         ],
         "llm-oga-igpu": [
@@ -84,9 +103,7 @@ setup(
             "lemonade-sdk[llm]",
         ],
         "llm-oga-hybrid": [
-            "onnx==1.16.1",
-            "numpy==1.26.4",
-            "protobuf>=6.30.1",
+            "lemonade-sdk[oga-hybrid-minimal]",
             "lemonade-sdk[llm]",
         ],
         "llm-oga-unified": [

{lemonade_sdk-7.0.4 → lemonade_sdk-8.0.0}/src/lemonade/api.py

@@ -57,7 +57,7 @@ def from_pretrained(
         # Huggingface supports all checkpoints, so there is nothing to check for
 
         import torch
-        from lemonade.tools.huggingface_load import HuggingfaceLoad
+        from lemonade.tools.huggingface.load import HuggingfaceLoad
 
         state = _make_state(recipe, checkpoint)
 
@@ -73,7 +73,7 @@ def from_pretrained(
         # Huggingface Transformers recipe for discrete GPU (Nvidia, Instinct, Radeon)
 
         import torch
-        from lemonade.tools.huggingface_load import HuggingfaceLoad
+        from lemonade.tools.huggingface.load import HuggingfaceLoad
 
         state = _make_state(recipe, checkpoint)
 
@@ -87,7 +87,7 @@ def from_pretrained(
         return state.model, state.tokenizer
 
     elif recipe.startswith("oga-"):
-        import lemonade.tools.ort_genai.oga as oga
+        import lemonade.tools.oga.load as oga
 
         # Make sure the user chose a supported runtime, e.g., oga-cpu
         user_backend = recipe.split("oga-")[1]

{lemonade_sdk-7.0.4 → lemonade_sdk-8.0.0}/src/lemonade/cli.py

@@ -1,4 +1,8 @@
 import os
+
+# pylint: disable=C0413
+# Prevent HF warnings from showing on every import
+os.environ["TRANSFORMERS_NO_ADVISORY_WARNINGS"] = "1"
 from lemonade.version import __version__ as version_number
 from lemonade.tools import FirstTool, NiceHelpFormatter
 from lemonade.profilers.memory_tracker import MemoryTracker
@@ -8,12 +12,12 @@ from lemonade.sequence import Sequence
 from lemonade.tools.management_tools import Cache, Version, SystemInfo
 from lemonade.state import State
 
-from lemonade.tools.huggingface_load import HuggingfaceLoad
-
-from lemonade.tools.huggingface_bench import HuggingfaceBench
-from lemonade.tools.ort_genai.oga_bench import OgaBench
-from lemonade.tools.llamacpp_bench import LlamaCppBench
-from lemonade.tools.llamacpp import LoadLlamaCpp
+from lemonade.tools.huggingface.load import HuggingfaceLoad
+from lemonade.tools.huggingface.bench import HuggingfaceBench
+from lemonade.tools.oga.load import OgaLoad
+from lemonade.tools.oga.bench import OgaBench
+from lemonade.tools.llamacpp.bench import LlamaCppBench
+from lemonade.tools.llamacpp.load import LoadLlamaCpp
 
 import lemonade.cache as cache
 from lemonade.tools.mmlu import AccuracyMMLU
@@ -24,7 +28,6 @@ from lemonade.tools.prompt import LLMPrompt
 from lemonade.tools.quark.quark_load import QuarkLoad
 from lemonade.tools.quark.quark_quantize import QuarkQuantize
 from lemonade.tools.report.llm_report import LemonadeReport
-from lemonade.tools.server.serve import Server
 
 
 def main():
@@ -40,26 +43,17 @@ def main():
         LMEvalHarness,
         LLMPrompt,
         HuggingfaceBench,
+        OgaLoad,
         OgaBench,
         QuarkQuantize,
         QuarkLoad,
         LemonadeReport,
-        Server,
         # Inherited from lemonade
         Cache,
         Version,
         SystemInfo,
     ]
 
-    # Import onnxruntime-genai recipes
-    try:
-        from lemonade.tools.ort_genai.oga import OgaLoad
-
-        tools = tools + [OgaLoad]
-
-    except ModuleNotFoundError:
-        pass
-
     # List the available profilers
     profilers = [MemoryTracker]
 

{lemonade_sdk-7.0.4 → lemonade_sdk-8.0.0}/src/lemonade/common/build.py

@@ -6,8 +6,6 @@ from typing import Dict
 import hashlib
 import psutil
 import yaml
-import torch
-import numpy as np
 import lemonade.common.exceptions as exp
 
 state_file_name = "state.yaml"
@@ -101,51 +99,6 @@ def unique_id():
     return hashlib.sha256(f"{pid}{start_time}".encode()).hexdigest()
 
 
-def get_shapes_and_dtypes(inputs: dict):
-    """
-    Return the shape and data type of each value in the inputs dict
-    """
-    shapes = {}
-    dtypes = {}
-    for key in sorted(inputs):
-        value = inputs[key]
-        if isinstance(
-            value,
-            (list, tuple),
-        ):
-            for v, i in zip(value, range(len(value))):
-                if isinstance(v, (list, tuple)):
-                    # Handle nested lists/tuples, for example past_key_values
-                    # in an LLM that has KV-caching enabled
-                    for v2, i2 in zip(v, range(len(v))):
-                        subsubkey = f"{key}[{i}][{i2}]"
-                        shapes[subsubkey] = np.array(v2).shape
-                        dtypes[subsubkey] = np.array(v2).dtype.name
-                else:
-                    # Handle single list/tuple
-                    subkey = f"{key}[{i}]"
-                    shapes[subkey] = np.array(v).shape
-                    dtypes[subkey] = np.array(v).dtype.name
-        elif torch.is_tensor(value):
-            shapes[key] = np.array(value.detach()).shape
-            dtypes[key] = np.array(value.detach()).dtype.name
-        elif isinstance(value, np.ndarray):
-            shapes[key] = value.shape
-            dtypes[key] = value.dtype.name
-        elif isinstance(value, (bool, int, float)):
-            shapes[key] = (1,)
-            dtypes[key] = type(value).__name__
-        elif value is None:
-            pass
-        else:
-            raise exp.Error(
-                "One of the provided inputs contains the unsupported "
-                f' type {type(value)} at key "{key}".'
-            )
-
-    return shapes, dtypes
-
-
 class Logger:
     """
     Redirects stdout to file (and console if needed)

lemonade_sdk-8.0.0/src/lemonade/common/network.py

@@ -0,0 +1,50 @@
+import os
+from typing import Optional
+import socket
+from huggingface_hub import model_info
+
+
+def is_offline():
+    """
+    Check if the system is offline by attempting to connect to huggingface.co.
+
+    Returns:
+        bool: True if the system is offline (cannot connect to huggingface.co),
+              False otherwise.
+    """
+    if os.environ.get("LEMONADE_OFFLINE"):
+        return True
+    try:
+        socket.gethostbyname("huggingface.co")
+        return False
+    except socket.gaierror:
+        return True
+
+
+def get_base_model(checkpoint: str) -> Optional[str]:
+    """
+    Get the base model information for a given checkpoint from the Hugging Face Hub.
+    Will auto-detect if we're offline and skip the network call in that case.
+
+    Args:
+        checkpoint: The model checkpoint to query
+
+    Returns:
+        The base model name if found, or None if not found or error occurs
+    """
+    # Skip network call in offline mode
+    if is_offline():
+        return None
+
+    try:
+        info = model_info(checkpoint)
+        if info.cardData and "base_model" in info.cardData:
+            if info.cardData["base_model"] is not None:
+                # This is a derived model
+                return info.cardData["base_model"]
+            else:
+                # This is itself a base model
+                return [checkpoint]
+    except Exception:  # pylint: disable=broad-except
+        pass
+    return None

{lemonade_sdk-7.0.4 → lemonade_sdk-8.0.0}/src/lemonade/common/status.py

@@ -7,12 +7,10 @@ import dataclasses
 from typing import Callable, List, Union, Dict, Optional
 import textwrap
 import psutil
-import torch
 from lemonade.common import printing
 from lemonade.state import State
 import lemonade.common.build as build
 import lemonade.common.filesystem as fs
-import lemonade.common.analyze_model as analyze_model
 
 
 def _pretty_print_key(key: str) -> str:
@@ -64,7 +62,6 @@ class SkipFields:
 
     file_name: bool = False
     model_name: bool = False
-    parameters: bool = False
     location: bool = False
     input_shape: bool = False
     build_dir: bool = False
@@ -147,18 +144,6 @@ class UniqueInvocationInfo(BasicInfo):
                 print(f", line {self.line}")
             self.skip.location = True
 
-    def _print_parameters(self):
-        if self.skip.parameters or self.params is None:
-            return
-
-        # Display number of parameters and size
-        parameters_size = parameters_to_size(self.params)
-        print(
-            f"{self.indent}\tParameters:\t{'{:,}'.format(self.params)} ({parameters_size})"
-        )
-
-        self.skip.parameters = True
-
     def _print_unique_input_shape(
         self,
         exec_time_formatted: str,
@@ -348,7 +333,6 @@ class UniqueInvocationInfo(BasicInfo):
         if (self.depth == 0 and not model_visited) or (self.depth != 0):
             # Print this information only once per model
             self._print_location()
-            self._print_parameters()
         self._print_unique_input_shape(
             exec_time_formatted, invocation_idx, multiple_unique_invocations
         )
@@ -362,16 +346,13 @@ class UniqueInvocationInfo(BasicInfo):
 
 @dataclasses.dataclass
 class ModelInfo(BasicInfo):
-    model: torch.nn.Module = None
+    model: str = None
     old_forward: Union[Callable, None] = None
     unique_invocations: Union[Dict[str, UniqueInvocationInfo], None] = (
         dataclasses.field(default_factory=dict)
     )
     last_unique_invocation_executed: Union[str, None] = None
 
-    def __post_init__(self):
-        self.params = analyze_model.count_parameters(self.model)
-
 
 def recursive_print(
     models_found: Dict[str, ModelInfo],
@@ -447,7 +428,7 @@ def stop_logger_forward() -> None:
 def add_to_state(
     state: State,
     name: str,
-    model: Union[str, torch.nn.Module],
+    model: str,
     extension: str = "",
     input_shapes: Optional[Dict] = None,
 ):

{lemonade_sdk-7.0.4 → lemonade_sdk-8.0.0}/src/lemonade/common/system_info.py

@@ -3,6 +3,7 @@ import importlib.metadata
 import platform
 import re
 import subprocess
+import ctypes
 
 
 class SystemInfo(ABC):
@@ -184,11 +185,25 @@ class WindowsSystemInfo(SystemInfo):
             str: Windows power setting.
         """
         try:
-            out = subprocess.check_output(["powercfg", "/getactivescheme"]).decode()
-            return re.search(r"\((.*?)\)", out).group(1)
+            # Capture output as bytes
+            out_bytes = subprocess.check_output(["powercfg", "/getactivescheme"])
+
+            # Get system's OEM code page (e.g., cp437, cp850)
+            oem_cp = "cp" + str(ctypes.windll.kernel32.GetOEMCP())
+
+            # Decode using detected OEM code page
+            out = out_bytes.decode(oem_cp)
+
+            # Extract power scheme name from parentheses
+            match = re.search(r"\((.*?)\)", out)
+            if match:
+                return match.group(1)
+            return "Power scheme name not found in output"
+
         except subprocess.CalledProcessError:
-            pass
-        return "Windows power setting not found"
+            return "Windows power setting not found (command failed)"
+        except Exception as e:  # pylint: disable=broad-except
+            return f"Error retrieving power setting: {str(e)}"
 
     def get_dict(self) -> dict:
         """