depuzzle 0.2.1__tar.gz
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- depuzzle-0.2.1/.github/workflows/ci.yml +55 -0
- depuzzle-0.2.1/.gitignore +5 -0
- depuzzle-0.2.1/.python-version +1 -0
- depuzzle-0.2.1/CHANGELOG.md +77 -0
- depuzzle-0.2.1/LICENSE +21 -0
- depuzzle-0.2.1/PKG-INFO +366 -0
- depuzzle-0.2.1/README.md +348 -0
- depuzzle-0.2.1/depuzzle/__init__.py +3 -0
- depuzzle-0.2.1/depuzzle/backends/__init__.py +0 -0
- depuzzle-0.2.1/depuzzle/backends/base.py +14 -0
- depuzzle-0.2.1/depuzzle/backends/ollama.py +86 -0
- depuzzle-0.2.1/depuzzle/cli.py +232 -0
- depuzzle-0.2.1/depuzzle/exporters.py +23 -0
- depuzzle-0.2.1/depuzzle/loaders.py +88 -0
- depuzzle-0.2.1/depuzzle/mcp_server/__init__.py +0 -0
- depuzzle-0.2.1/depuzzle/mcp_server/server.py +12 -0
- depuzzle-0.2.1/depuzzle/mcp_server/tools.py +14 -0
- depuzzle-0.2.1/depuzzle/metrics.py +45 -0
- depuzzle-0.2.1/depuzzle/models.py +44 -0
- depuzzle-0.2.1/depuzzle/profiler.py +54 -0
- depuzzle-0.2.1/examples/ollama_demo.py +18 -0
- depuzzle-0.2.1/examples/try_loaders.py +7 -0
- depuzzle-0.2.1/pyproject.toml +40 -0
- depuzzle-0.2.1/pytest.ini +3 -0
- depuzzle-0.2.1/run3.json +1979 -0
- depuzzle-0.2.1/sample_traces/cpu_cache_trace.json +1994 -0
- depuzzle-0.2.1/sample_traces/run1.json +2542 -0
- depuzzle-0.2.1/sample_traces/run2.json +2230 -0
- depuzzle-0.2.1/scripts/test_mcp_client.py +34 -0
- depuzzle-0.2.1/tests/conftest.py +59 -0
- depuzzle-0.2.1/tests/integration/test_ollama_backend.py +40 -0
- depuzzle-0.2.1/tests/test_backend_info.py +13 -0
- depuzzle-0.2.1/tests/test_compare_metrics.py +62 -0
- depuzzle-0.2.1/tests/test_loaders.py +117 -0
- depuzzle-0.2.1/tests/test_mcp_protocol.py +17 -0
- depuzzle-0.2.1/tests/test_mcp_server.py +13 -0
- depuzzle-0.2.1/tests/test_mcp_tools.py +13 -0
- depuzzle-0.2.1/tests/test_metrics.py +106 -0
- depuzzle-0.2.1/tests/test_models.py +60 -0
- depuzzle-0.2.1/tests/test_ollama_parser.py +18 -0
- depuzzle-0.2.1/tests/test_profiler.py +56 -0
|
@@ -0,0 +1,55 @@
|
|
|
1
|
+
name: CI
|
|
2
|
+
|
|
3
|
+
on:
|
|
4
|
+
push:
|
|
5
|
+
branches:
|
|
6
|
+
- "**"
|
|
7
|
+
pull_request:
|
|
8
|
+
branches:
|
|
9
|
+
- main
|
|
10
|
+
|
|
11
|
+
jobs:
|
|
12
|
+
test:
|
|
13
|
+
runs-on: ubuntu-latest
|
|
14
|
+
|
|
15
|
+
steps:
|
|
16
|
+
- name: Checkout repository
|
|
17
|
+
uses: actions/checkout@v4
|
|
18
|
+
|
|
19
|
+
- name: Set up Python
|
|
20
|
+
uses: actions/setup-python@v5
|
|
21
|
+
with:
|
|
22
|
+
python-version: "3.12"
|
|
23
|
+
|
|
24
|
+
- name: Install project
|
|
25
|
+
run: |
|
|
26
|
+
python -m pip install -e .
|
|
27
|
+
|
|
28
|
+
- name: Install dependencies
|
|
29
|
+
run: |
|
|
30
|
+
python -m pip install --upgrade pip
|
|
31
|
+
pip install -e ".[dev]"
|
|
32
|
+
|
|
33
|
+
- name: Debug package
|
|
34
|
+
run: |
|
|
35
|
+
pwd
|
|
36
|
+
ls -la
|
|
37
|
+
ls -la depuzzle
|
|
38
|
+
python -c "import sys; print(sys.executable)"
|
|
39
|
+
python -c "import depuzzle; print(depuzzle.__file__)"
|
|
40
|
+
|
|
41
|
+
- name: Run tests
|
|
42
|
+
run: |
|
|
43
|
+
python -m pytest -m "not integration"
|
|
44
|
+
|
|
45
|
+
- name: Check formatting
|
|
46
|
+
run: |
|
|
47
|
+
python -m black --check .
|
|
48
|
+
|
|
49
|
+
- name: Run linter
|
|
50
|
+
run: |
|
|
51
|
+
python -m ruff check .
|
|
52
|
+
|
|
53
|
+
- name: Type check
|
|
54
|
+
run: |
|
|
55
|
+
python -m mypy depuzzle
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
3.12.12
|
|
@@ -0,0 +1,77 @@
|
|
|
1
|
+
# Changelog
|
|
2
|
+
|
|
3
|
+
All notable changes to this project will be documented in this file.
|
|
4
|
+
|
|
5
|
+
## [0.2.1] - 2026-08-12
|
|
6
|
+
|
|
7
|
+
### Fixed
|
|
8
|
+
|
|
9
|
+
- TestPyPI package upload and successful clean install
|
|
10
|
+
|
|
11
|
+
## [0.2.0] - 2026-07-28
|
|
12
|
+
|
|
13
|
+
### Added
|
|
14
|
+
|
|
15
|
+
- Added runtime metadata collection for inference runs.
|
|
16
|
+
- Added backend runtime information to trace summaries:
|
|
17
|
+
- Backend name
|
|
18
|
+
- Processor utilization (CPU/GPU split)
|
|
19
|
+
- Context length
|
|
20
|
+
- Added runtime metadata to JSON trace outputs.
|
|
21
|
+
- Added runtime comparison support for `depuzzle compare`.
|
|
22
|
+
- Added shared test fixtures using `conftest.py`.
|
|
23
|
+
- Added integration test coverage for Ollama backend functionality.
|
|
24
|
+
|
|
25
|
+
### Changed
|
|
26
|
+
|
|
27
|
+
- Updated trace summaries to display runtime information separately from inference metrics.
|
|
28
|
+
- Improved comparison output by separating runtime configuration from performance metrics.
|
|
29
|
+
- Refactored tests to use reusable fake backends and fake traces.
|
|
30
|
+
- Improved handling of missing runtime information.
|
|
31
|
+
|
|
32
|
+
### Fixed
|
|
33
|
+
|
|
34
|
+
- Fixed trace summary output formatting.
|
|
35
|
+
- Fixed comparison handling when runtime metadata is unavailable.
|
|
36
|
+
- Fixed CI failures caused by integration tests requiring Ollama.
|
|
37
|
+
|
|
38
|
+
### Testing
|
|
39
|
+
|
|
40
|
+
- Added unit tests for runtime metadata.
|
|
41
|
+
- Added tests for runtime comparison.
|
|
42
|
+
- Added Ollama integration test markers.
|
|
43
|
+
- Improved GitHub Actions CI workflow.
|
|
44
|
+
|
|
45
|
+
## [0.1.0] - 2026-07-24
|
|
46
|
+
|
|
47
|
+
### Added
|
|
48
|
+
|
|
49
|
+
- Initial release of depuzzle.
|
|
50
|
+
- CLI for profiling local LLM inference.
|
|
51
|
+
- Support for profiling Ollama models.
|
|
52
|
+
- Token-level inference trace collection.
|
|
53
|
+
- JSON trace export.
|
|
54
|
+
- Trace loading utilities.
|
|
55
|
+
- Inference metrics including:
|
|
56
|
+
- token count
|
|
57
|
+
- total latency
|
|
58
|
+
- time to first token (TTFT)
|
|
59
|
+
- tokens per second
|
|
60
|
+
- Trace comparison command.
|
|
61
|
+
- Rich terminal output for summaries and comparisons.
|
|
62
|
+
- Unit test suite.
|
|
63
|
+
- GitHub Actions continuous integration.
|
|
64
|
+
- Project documentation and installation instructions.
|
|
65
|
+
|
|
66
|
+
### Changed
|
|
67
|
+
|
|
68
|
+
- Established the initial project architecture around:
|
|
69
|
+
- Backends
|
|
70
|
+
- Profiler
|
|
71
|
+
- Trace models
|
|
72
|
+
- Metrics
|
|
73
|
+
- CLI
|
|
74
|
+
|
|
75
|
+
### Fixed
|
|
76
|
+
|
|
77
|
+
- Initial public release.
|
depuzzle-0.2.1/LICENSE
ADDED
|
@@ -0,0 +1,21 @@
|
|
|
1
|
+
MIT License
|
|
2
|
+
|
|
3
|
+
Copyright (c) 2026 Manan Raheja
|
|
4
|
+
|
|
5
|
+
Permission is hereby granted, free of charge, to any person obtaining a copy
|
|
6
|
+
of this software and associated documentation files (the "Software"), to deal
|
|
7
|
+
in the Software without restriction, including without limitation the rights
|
|
8
|
+
to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
|
|
9
|
+
copies of the Software, and to permit persons to whom the Software is
|
|
10
|
+
furnished to do so, subject to the following conditions:
|
|
11
|
+
|
|
12
|
+
The above copyright notice and this permission notice shall be included in all
|
|
13
|
+
copies or substantial portions of the Software.
|
|
14
|
+
|
|
15
|
+
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
|
|
16
|
+
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
|
|
17
|
+
FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
|
|
18
|
+
AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
|
|
19
|
+
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
|
|
20
|
+
OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
|
|
21
|
+
SOFTWARE.
|
depuzzle-0.2.1/PKG-INFO
ADDED
|
@@ -0,0 +1,366 @@
|
|
|
1
|
+
Metadata-Version: 2.5
|
|
2
|
+
Name: depuzzle
|
|
3
|
+
Version: 0.2.1
|
|
4
|
+
Summary: Inference profiler for local LLMs
|
|
5
|
+
License-File: LICENSE
|
|
6
|
+
Requires-Python: <3.14,>=3.12
|
|
7
|
+
Requires-Dist: fastmcp
|
|
8
|
+
Requires-Dist: httpx
|
|
9
|
+
Requires-Dist: mcp
|
|
10
|
+
Requires-Dist: typer
|
|
11
|
+
Provides-Extra: dev
|
|
12
|
+
Requires-Dist: black; extra == 'dev'
|
|
13
|
+
Requires-Dist: mypy; extra == 'dev'
|
|
14
|
+
Requires-Dist: pytest; extra == 'dev'
|
|
15
|
+
Requires-Dist: pytest-asyncio; extra == 'dev'
|
|
16
|
+
Requires-Dist: ruff; extra == 'dev'
|
|
17
|
+
Description-Content-Type: text/markdown
|
|
18
|
+
|
|
19
|
+
# depuzzle
|
|
20
|
+
|
|
21
|
+
[](https://github.com/mananraheja/depuzzle/actions)
|
|
22
|
+
|
|
23
|
+
A lightweight Python tool for profiling, analyzing, and comparing LLM inference performance.
|
|
24
|
+
|
|
25
|
+
`depuzzle` helps developers profile local LLM inference runs, capture token-level traces, and compare performance across models and configurations.
|
|
26
|
+
|
|
27
|
+
---
|
|
28
|
+
|
|
29
|
+
## Usage
|
|
30
|
+
|
|
31
|
+
### Profile an LLM inference run
|
|
32
|
+
|
|
33
|
+
depuzzle can profile local LLM inference requests and capture token-level timing information.
|
|
34
|
+
|
|
35
|
+
Example:
|
|
36
|
+
|
|
37
|
+
```bash
|
|
38
|
+
depuzzle profile run \
|
|
39
|
+
--model qwen2.5:3b \
|
|
40
|
+
--prompt "Explain virtual memory in one paragraph." \
|
|
41
|
+
--output run1.json
|
|
42
|
+
```
|
|
43
|
+
|
|
44
|
+
Example output:
|
|
45
|
+
|
|
46
|
+
```bash
|
|
47
|
+
--- Trace Summary ---
|
|
48
|
+
|
|
49
|
+
model: qwen2.5:3b
|
|
50
|
+
tokens: 633
|
|
51
|
+
latency_seconds: 31.13s
|
|
52
|
+
ttft_seconds: 4.71s
|
|
53
|
+
tokens_per_second: 20.33
|
|
54
|
+
|
|
55
|
+
runtime:
|
|
56
|
+
backend: ollama
|
|
57
|
+
processor: 72%/28% CPU/GPU
|
|
58
|
+
context_length: 131072
|
|
59
|
+
|
|
60
|
+
Trace saved to run1.json
|
|
61
|
+
```
|
|
62
|
+
|
|
63
|
+
### Compare inference runs
|
|
64
|
+
|
|
65
|
+
depuzzle can compare two inference traces to evaluate model performance differences.
|
|
66
|
+
|
|
67
|
+
Example:
|
|
68
|
+
|
|
69
|
+
```bash
|
|
70
|
+
depuzzle compare sample_traces/run1.json sample_traces/run2.json
|
|
71
|
+
```
|
|
72
|
+
|
|
73
|
+
Example output:
|
|
74
|
+
|
|
75
|
+
```bash
|
|
76
|
+
--- Comparison ---
|
|
77
|
+
|
|
78
|
+
Metric Run 1 Run 2 Change
|
|
79
|
+
------------------------------------------------------------------------------------------
|
|
80
|
+
model qwen2.5:3b llama3.2:3b -
|
|
81
|
+
tokens 633 555 -12.32%
|
|
82
|
+
latency_seconds 31.13 28.41 -8.74% (faster)
|
|
83
|
+
ttft_seconds 4.71 6.98 +48.07% (slower)
|
|
84
|
+
tokens_per_second 20.33 19.53 -3.93% (worse)
|
|
85
|
+
|
|
86
|
+
|
|
87
|
+
runtime
|
|
88
|
+
backend ollama ollama same
|
|
89
|
+
processor 72%/28% CPU/GPU 100% CPU changed
|
|
90
|
+
context_length 131072 131072 same
|
|
91
|
+
```
|
|
92
|
+
|
|
93
|
+
---
|
|
94
|
+
|
|
95
|
+
## Features
|
|
96
|
+
|
|
97
|
+
- Profile local LLM inference runs
|
|
98
|
+
- Capture token-level inference traces
|
|
99
|
+
- Measure:
|
|
100
|
+
- Total latency
|
|
101
|
+
- Time to First Token (TTFT)
|
|
102
|
+
- Tokens per second
|
|
103
|
+
- Record runtime information
|
|
104
|
+
- Backend
|
|
105
|
+
- CPU/GPU processor split
|
|
106
|
+
- Context length
|
|
107
|
+
- Export traces as JSON
|
|
108
|
+
- Compare inference runs
|
|
109
|
+
- Automated testing with GitHub Actions
|
|
110
|
+
|
|
111
|
+
---
|
|
112
|
+
|
|
113
|
+
## Supported Backends
|
|
114
|
+
|
|
115
|
+
Current:
|
|
116
|
+
|
|
117
|
+
- Ollama
|
|
118
|
+
|
|
119
|
+
Planned:
|
|
120
|
+
|
|
121
|
+
- MLX
|
|
122
|
+
- llama.cpp
|
|
123
|
+
- vLLM
|
|
124
|
+
- TensorRT-LLM
|
|
125
|
+
|
|
126
|
+
---
|
|
127
|
+
|
|
128
|
+
## Installation
|
|
129
|
+
|
|
130
|
+
Clone the repository:
|
|
131
|
+
|
|
132
|
+
```bash
|
|
133
|
+
git clone https://github.com/mananraheja/depuzzle.git
|
|
134
|
+
cd depuzzle
|
|
135
|
+
```
|
|
136
|
+
|
|
137
|
+
Create a virtual environment:
|
|
138
|
+
|
|
139
|
+
```bash
|
|
140
|
+
python -m venv .venv
|
|
141
|
+
source .venv/bin/activate
|
|
142
|
+
```
|
|
143
|
+
|
|
144
|
+
Install:
|
|
145
|
+
|
|
146
|
+
```bash
|
|
147
|
+
pip install -e .
|
|
148
|
+
```
|
|
149
|
+
|
|
150
|
+
For development:
|
|
151
|
+
|
|
152
|
+
```bash
|
|
153
|
+
pip install -e ".[dev]"
|
|
154
|
+
```
|
|
155
|
+
|
|
156
|
+
---
|
|
157
|
+
|
|
158
|
+
## Quick Start
|
|
159
|
+
|
|
160
|
+
Run profiler:
|
|
161
|
+
|
|
162
|
+
```bash
|
|
163
|
+
depuzzle profile run \
|
|
164
|
+
--model <model_name> \
|
|
165
|
+
--prompt "<sample_prompt>" \
|
|
166
|
+
--output <path_to_output_file.json>
|
|
167
|
+
```
|
|
168
|
+
|
|
169
|
+
Compare two inference runs:
|
|
170
|
+
|
|
171
|
+
```bash
|
|
172
|
+
depuzzle compare run1.json run2.json
|
|
173
|
+
```
|
|
174
|
+
|
|
175
|
+
Example:
|
|
176
|
+
|
|
177
|
+
```bash
|
|
178
|
+
Metric Run 1 Run 2 Change
|
|
179
|
+
|
|
180
|
+
Total Latency 12.5s 10.2s -18%
|
|
181
|
+
Time To First Token 0.8s 0.6s -25%
|
|
182
|
+
Tokens Generated 256 256 -
|
|
183
|
+
Tokens/sec 20.4 25.1 +23%
|
|
184
|
+
```
|
|
185
|
+
|
|
186
|
+
---
|
|
187
|
+
|
|
188
|
+
## Trace Format
|
|
189
|
+
|
|
190
|
+
`depuzzle` expects traces in JSON format:
|
|
191
|
+
```json
|
|
192
|
+
{
|
|
193
|
+
"model": "llama3.2:3b",
|
|
194
|
+
"prompt": "Hello world",
|
|
195
|
+
"start_time": "2026-07-22T10:00:00",
|
|
196
|
+
"end_time": "2026-07-22T10:00:12",
|
|
197
|
+
"tokens": [
|
|
198
|
+
{
|
|
199
|
+
"token": "Hello",
|
|
200
|
+
"timestamp": "2026-07-22T10:00:01"
|
|
201
|
+
}
|
|
202
|
+
],
|
|
203
|
+
"total_latency": 12.0,
|
|
204
|
+
"time_to_first_token": 1.0,
|
|
205
|
+
"backend_info": {
|
|
206
|
+
"backend": "ollama",
|
|
207
|
+
"processor": "72%/28% CPU/GPU",
|
|
208
|
+
"context_length": 131072
|
|
209
|
+
}
|
|
210
|
+
}
|
|
211
|
+
```
|
|
212
|
+
|
|
213
|
+
---
|
|
214
|
+
|
|
215
|
+
## Architecture
|
|
216
|
+
|
|
217
|
+
`depuzzle` consists of two main workflows:
|
|
218
|
+
|
|
219
|
+
### 1. Profile Run
|
|
220
|
+
|
|
221
|
+
The profiling workflow executes an inference run and captures performance data.
|
|
222
|
+
|
|
223
|
+
```bash
|
|
224
|
+
User
|
|
225
|
+
|
|
|
226
|
+
v
|
|
227
|
+
depuzzle profile run
|
|
228
|
+
|
|
|
229
|
+
v
|
|
230
|
+
Profiler
|
|
231
|
+
|
|
|
232
|
+
v
|
|
233
|
+
Backend Adapter
|
|
234
|
+
|
|
|
235
|
+
v
|
|
236
|
+
Trace Recorder
|
|
237
|
+
|
|
|
238
|
+
v
|
|
239
|
+
trace.json
|
|
240
|
+
```
|
|
241
|
+
|
|
242
|
+
|
|
243
|
+
The generated trace contains:
|
|
244
|
+
- model information
|
|
245
|
+
- prompt metadata
|
|
246
|
+
- token generation timestamps
|
|
247
|
+
- latency measurements
|
|
248
|
+
- time-to-first-token
|
|
249
|
+
- backend information
|
|
250
|
+
- processor utilization
|
|
251
|
+
- context length
|
|
252
|
+
|
|
253
|
+
### 2. Analyze and Compare
|
|
254
|
+
|
|
255
|
+
The analysis workflow processes captured traces.
|
|
256
|
+
|
|
257
|
+
```bash
|
|
258
|
+
trace.json
|
|
259
|
+
|
|
|
260
|
+
v
|
|
261
|
+
Trace Loader
|
|
262
|
+
|
|
|
263
|
+
v
|
|
264
|
+
InferenceTrace Model
|
|
265
|
+
|
|
|
266
|
+
v
|
|
267
|
+
Metrics Engine
|
|
268
|
+
|
|
|
269
|
+
v
|
|
270
|
+
Comparison Report
|
|
271
|
+
```
|
|
272
|
+
|
|
273
|
+
Core components:
|
|
274
|
+
|
|
275
|
+
- **Profiler**: Captures inference execution events and generates trace files.
|
|
276
|
+
- **Loader**: Parses and validates trace JSON files.
|
|
277
|
+
- **Models**: Provides structured representations of inference runs.
|
|
278
|
+
- **Metrics**: Calculates performance statistics.
|
|
279
|
+
- **Compare**: Compares multiple inference runs.
|
|
280
|
+
|
|
281
|
+
---
|
|
282
|
+
|
|
283
|
+
## Development
|
|
284
|
+
|
|
285
|
+
Run tests:
|
|
286
|
+
|
|
287
|
+
```bash
|
|
288
|
+
pytest -m "not integration"
|
|
289
|
+
```
|
|
290
|
+
|
|
291
|
+
Run integration tests (requires Ollama and downloaded local models):
|
|
292
|
+
```bash
|
|
293
|
+
pytest -m integration
|
|
294
|
+
```
|
|
295
|
+
|
|
296
|
+
Format code:
|
|
297
|
+
|
|
298
|
+
```bash
|
|
299
|
+
black .
|
|
300
|
+
```
|
|
301
|
+
|
|
302
|
+
Lint:
|
|
303
|
+
|
|
304
|
+
```bash
|
|
305
|
+
ruff check .
|
|
306
|
+
```
|
|
307
|
+
|
|
308
|
+
Type checking:
|
|
309
|
+
|
|
310
|
+
```bash
|
|
311
|
+
mypy depuzzle
|
|
312
|
+
```
|
|
313
|
+
|
|
314
|
+
---
|
|
315
|
+
|
|
316
|
+
## Roadmap
|
|
317
|
+
|
|
318
|
+
### v0.1.0
|
|
319
|
+
- [x] Profile local LLM inference runs
|
|
320
|
+
- [x] Export traces to JSON
|
|
321
|
+
- [x] Calculate latency and throughput metrics
|
|
322
|
+
- [x] Compare inference runs
|
|
323
|
+
- [x] GitHub Actions CI
|
|
324
|
+
|
|
325
|
+
### v0.2.1
|
|
326
|
+
- [x] Backend runtime metadata
|
|
327
|
+
- [x] Runtime information in trace summaries
|
|
328
|
+
- [x] Runtime comparison between inference runs
|
|
329
|
+
|
|
330
|
+
### Future
|
|
331
|
+
|
|
332
|
+
#### Additional Backends
|
|
333
|
+
- [ ] MLX
|
|
334
|
+
- [ ] llama.cpp
|
|
335
|
+
- [ ] vLLM
|
|
336
|
+
- [ ] TensorRT-LLM
|
|
337
|
+
- [ ] Hugging Face Transformers
|
|
338
|
+
|
|
339
|
+
#### Profiling
|
|
340
|
+
- [ ] Memory usage
|
|
341
|
+
- [ ] GPU utilization
|
|
342
|
+
- [ ] Live inference profiling
|
|
343
|
+
- [ ] Batch benchmarking
|
|
344
|
+
- [ ] Multi-run statistical summaries
|
|
345
|
+
|
|
346
|
+
#### Visualization
|
|
347
|
+
- [ ] Interactive timeline visualization
|
|
348
|
+
- [ ] Trace diff visualization
|
|
349
|
+
- [ ] HTML report generation
|
|
350
|
+
|
|
351
|
+
#### AI Agent Integration
|
|
352
|
+
- [ ] MCP server
|
|
353
|
+
- [ ] Additional MCP tools
|
|
354
|
+
- [ ] Benchmark orchestration through MCP
|
|
355
|
+
|
|
356
|
+
#### Long-term
|
|
357
|
+
- [ ] Layer-by-layer profiling
|
|
358
|
+
- [ ] Distributed inference profiling
|
|
359
|
+
- [ ] Multi-node benchmarking
|
|
360
|
+
- [ ] Plugin system for custom backends
|
|
361
|
+
|
|
362
|
+
---
|
|
363
|
+
|
|
364
|
+
## LICENSE
|
|
365
|
+
|
|
366
|
+
MIT
|