trio-core 0.3.0__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.
Files changed (193) hide show
  1. trio_core-0.3.0/.github/ISSUE_TEMPLATE/bug_report.md +46 -0
  2. trio_core-0.3.0/.github/ISSUE_TEMPLATE/feature_request.md +25 -0
  3. trio_core-0.3.0/.github/ISSUE_TEMPLATE/model_support.md +30 -0
  4. trio_core-0.3.0/.github/PULL_REQUEST_TEMPLATE.md +28 -0
  5. trio_core-0.3.0/.github/workflows/publish.yml +28 -0
  6. trio_core-0.3.0/.github/workflows/test.yml +28 -0
  7. trio_core-0.3.0/.gitignore +16 -0
  8. trio_core-0.3.0/CHANGELOG.md +61 -0
  9. trio_core-0.3.0/CONTRIBUTING.md +102 -0
  10. trio_core-0.3.0/LICENSE +190 -0
  11. trio_core-0.3.0/PKG-INFO +257 -0
  12. trio_core-0.3.0/README.md +205 -0
  13. trio_core-0.3.0/SECURITY.md +26 -0
  14. trio_core-0.3.0/examples/bench_generate_loop.py +231 -0
  15. trio_core-0.3.0/examples/bench_non_qwen.py +465 -0
  16. trio_core-0.3.0/examples/bench_prefix_cache.py +114 -0
  17. trio_core-0.3.0/examples/bench_sink.py +235 -0
  18. trio_core-0.3.0/examples/bench_sink_real.py +202 -0
  19. trio_core-0.3.0/examples/bench_streaming_qa.py +376 -0
  20. trio_core-0.3.0/examples/bench_visual_similarity.py +177 -0
  21. trio_core-0.3.0/examples/run_baseline_mlxvlm.py +175 -0
  22. trio_core-0.3.0/examples/run_benchmark.py +191 -0
  23. trio_core-0.3.0/examples/run_eval.py +110 -0
  24. trio_core-0.3.0/examples/run_regression.py +299 -0
  25. trio_core-0.3.0/examples/run_tier1_benchmark.py +313 -0
  26. trio_core-0.3.0/examples/screen_narrator.py +320 -0
  27. trio_core-0.3.0/examples/stream_monitor.py +120 -0
  28. trio_core-0.3.0/examples/video_analyze.py +84 -0
  29. trio_core-0.3.0/examples/webcam_gui.py +207 -0
  30. trio_core-0.3.0/examples/webcam_to_text.py +75 -0
  31. trio_core-0.3.0/pyproject.toml +54 -0
  32. trio_core-0.3.0/research/README.md +365 -0
  33. trio_core-0.3.0/research/eval-baseline-plan.md +206 -0
  34. trio_core-0.3.0/research/eval-results/bench_mvbench_8tasks_20_baseline.json +1619 -0
  35. trio_core-0.3.0/research/eval-results/bench_mvbench_8tasks_20_tome_r4.json +1619 -0
  36. trio_core-0.3.0/research/eval-results/bench_mvbench_clevrer_5_baseline.json +269 -0
  37. trio_core-0.3.0/research/eval-results/bench_pope_100_baseline.json +919 -0
  38. trio_core-0.3.0/research/eval-results/bench_pope_100_baseline_v2.json +919 -0
  39. trio_core-0.3.0/research/eval-results/bench_pope_100_qwen3vl_baseline.json +919 -0
  40. trio_core-0.3.0/research/eval-results/bench_pope_100_qwen3vl_tome_r4.json +919 -0
  41. trio_core-0.3.0/research/eval-results/bench_pope_100_tome_r4.json +919 -0
  42. trio_core-0.3.0/research/eval-results/bench_pope_100_tome_r4_hidden_mk02.json +919 -0
  43. trio_core-0.3.0/research/eval-results/bench_pope_100_tome_r4_hidden_mk025.json +919 -0
  44. trio_core-0.3.0/research/eval-results/bench_pope_100_tome_r4_hidden_v2.json +919 -0
  45. trio_core-0.3.0/research/eval-results/bench_pope_100_tome_r4_keys.json +919 -0
  46. trio_core-0.3.0/research/eval-results/bench_pope_100_tome_r4_keys_v2.json +919 -0
  47. trio_core-0.3.0/research/eval-results/bench_pope_100_tome_r8.json +919 -0
  48. trio_core-0.3.0/research/eval-results/bench_pope_random_30_baseline.json +289 -0
  49. trio_core-0.3.0/research/eval-results/bench_streaming_qa_w10_b100.json +110 -0
  50. trio_core-0.3.0/research/eval-results/bench_streaming_qa_w30_b100.json +35 -0
  51. trio_core-0.3.0/research/eval-results/compound-tome-fastv-benchmark.md +113 -0
  52. trio_core-0.3.0/research/eval-results/eval_1080p_baseline.json +322 -0
  53. trio_core-0.3.0/research/eval-results/eval_1080p_tome_r4.json +322 -0
  54. trio_core-0.3.0/research/eval-results/eval_1080p_tome_r4_hidden.json +322 -0
  55. trio_core-0.3.0/research/eval-results/eval_baseline.json +322 -0
  56. trio_core-0.3.0/research/eval-results/eval_compressed_50.json +322 -0
  57. trio_core-0.3.0/research/eval-results/eval_compressed_75.json +322 -0
  58. trio_core-0.3.0/research/eval-results/eval_qwen35_0.8b_baseline.json +322 -0
  59. trio_core-0.3.0/research/eval-results/eval_qwen35_0.8b_tome_r4.json +322 -0
  60. trio_core-0.3.0/research/eval-results/eval_qwen3vl_baseline.json +322 -0
  61. trio_core-0.3.0/research/eval-results/eval_qwen3vl_tome_r4.json +322 -0
  62. trio_core-0.3.0/research/eval-results/eval_tome_r8.json +322 -0
  63. trio_core-0.3.0/research/eval-results/mlxvlm-native-baselines.md +87 -0
  64. trio_core-0.3.0/research/eval-results/regression/SUMMARY.md +149 -0
  65. trio_core-0.3.0/research/eval-results/regression/default_baseline.json +26 -0
  66. trio_core-0.3.0/research/eval-results/regression/default_baseline_latest.json +36 -0
  67. trio_core-0.3.0/research/eval-results/regression/default_vs_0.95_latest.json +29 -0
  68. trio_core-0.3.0/research/eval-results/regression/gemma-3-4b-it_baseline.json +36 -0
  69. trio_core-0.3.0/research/eval-results/regression/gemma-3-4b-it_mlxvlm_native.json +36 -0
  70. trio_core-0.3.0/research/eval-results/regression/qwen2.5-vl-3b_baseline.json +36 -0
  71. trio_core-0.3.0/research/eval-results/regression/qwen2.5-vl-3b_baseline_latest.json +36 -0
  72. trio_core-0.3.0/research/eval-results/regression/qwen2.5-vl-3b_fastv_0.5.json +38 -0
  73. trio_core-0.3.0/research/eval-results/regression/qwen2.5-vl-3b_fastv_0.5_L16.json +38 -0
  74. trio_core-0.3.0/research/eval-results/regression/qwen2.5-vl-3b_fastv_0.5_L2.json +38 -0
  75. trio_core-0.3.0/research/eval-results/regression/qwen2.5-vl-3b_fastv_0.5_L8.json +38 -0
  76. trio_core-0.3.0/research/eval-results/regression/qwen2.5-vl-3b_mlxvlm_native.json +36 -0
  77. trio_core-0.3.0/research/eval-results/regression/qwen2.5-vl-3b_tome_r1.json +36 -0
  78. trio_core-0.3.0/research/eval-results/regression/qwen2.5-vl-3b_tome_r2.json +36 -0
  79. trio_core-0.3.0/research/eval-results/regression/qwen2.5-vl-3b_tome_r4.json +36 -0
  80. trio_core-0.3.0/research/eval-results/regression/qwen2.5-vl-3b_tome_r4_adaptive.json +38 -0
  81. trio_core-0.3.0/research/eval-results/regression/qwen2.5-vl-3b_tome_r4_fastv_0.5.json +38 -0
  82. trio_core-0.3.0/research/eval-results/regression/qwen3-vl-4b_baseline.json +36 -0
  83. trio_core-0.3.0/research/eval-results/regression/qwen3-vl-4b_baseline_latest.json +36 -0
  84. trio_core-0.3.0/research/eval-results/regression/qwen3-vl-4b_mlxvlm_native.json +36 -0
  85. trio_core-0.3.0/research/eval-results/regression/qwen3-vl-4b_tome_r4.json +36 -0
  86. trio_core-0.3.0/research/eval-results/regression/qwen3-vl-4b_tome_r4_fastv_0.5.json +33 -0
  87. trio_core-0.3.0/research/eval-results/regression/qwen3.5-0.8b-mlx_baseline.json +36 -0
  88. trio_core-0.3.0/research/eval-results/regression/qwen3.5-0.8b-mlx_baseline_latest.json +36 -0
  89. trio_core-0.3.0/research/eval-results/regression/qwen3.5-0.8b-mlx_mlxvlm_native.json +36 -0
  90. trio_core-0.3.0/research/eval-results/regression/qwen3.5-0.8b-mlx_tome_r4.json +36 -0
  91. trio_core-0.3.0/research/eval-results/speculative-decode-benchmark.md +119 -0
  92. trio_core-0.3.0/research/eval-results/tier1/eval_qwen2.5-vl-3b_480p_baseline.json +322 -0
  93. trio_core-0.3.0/research/eval-results/tier1/eval_qwen2.5-vl-7b_480p_baseline.json +322 -0
  94. trio_core-0.3.0/research/eval-results/tier1/eval_qwen3-vl-2b_480p_baseline.json +322 -0
  95. trio_core-0.3.0/research/eval-results/tier1/eval_qwen3-vl-4b_480p_baseline.json +322 -0
  96. trio_core-0.3.0/research/eval-results/tier1/eval_qwen3-vl-8b_480p_baseline.json +322 -0
  97. trio_core-0.3.0/research/eval-results/tier1/eval_qwen3.5-0.8b_480p_baseline.json +322 -0
  98. trio_core-0.3.0/research/eval-results/tier1/eval_qwen3.5-2b_480p_baseline.json +322 -0
  99. trio_core-0.3.0/research/eval-results/tier1/eval_qwen3.5-4b_480p_baseline.json +322 -0
  100. trio_core-0.3.0/research/eval-results/tier1/eval_qwen3.5-9b_480p_baseline.json +322 -0
  101. trio_core-0.3.0/research/eval-results/tier1/kv_reuse_internvl3-1b.json +11 -0
  102. trio_core-0.3.0/research/eval-results/tier1/kv_reuse_internvl3-2b.json +11 -0
  103. trio_core-0.3.0/research/eval-results/tier1/kv_reuse_nanollava.json +11 -0
  104. trio_core-0.3.0/research/eval-results/tier1/non_qwen_benchmark_summary.json +150 -0
  105. trio_core-0.3.0/research/eval-results/tier1/perf_internvl3-1b_baseline.json +53 -0
  106. trio_core-0.3.0/research/eval-results/tier1/perf_internvl3-1b_compressed_50.json +53 -0
  107. trio_core-0.3.0/research/eval-results/tier1/perf_internvl3-2b_baseline.json +53 -0
  108. trio_core-0.3.0/research/eval-results/tier1/perf_internvl3-2b_compressed_50.json +53 -0
  109. trio_core-0.3.0/research/eval-results/tier1/perf_nanollava_baseline.json +53 -0
  110. trio_core-0.3.0/research/eval-results/tier1/pope_internvl3-1b_baseline.json +519 -0
  111. trio_core-0.3.0/research/eval-results/tier1/pope_internvl3-1b_compressed_50.json +519 -0
  112. trio_core-0.3.0/research/eval-results/tier1/pope_internvl3-1b_fastv_0.5.json +519 -0
  113. trio_core-0.3.0/research/eval-results/tier1/pope_internvl3-2b_baseline.json +519 -0
  114. trio_core-0.3.0/research/eval-results/tier1/pope_internvl3-2b_compressed_50.json +519 -0
  115. trio_core-0.3.0/research/eval-results/tier1/pope_internvl3-2b_fastv_0.5.json +519 -0
  116. trio_core-0.3.0/research/eval-results/tier1/pope_nanollava_baseline.json +519 -0
  117. trio_core-0.3.0/research/eval-results/tier1/pope_qwen2.5-vl-3b_baseline.json +469 -0
  118. trio_core-0.3.0/research/eval-results/tier1/pope_qwen2.5-vl-7b_baseline.json +469 -0
  119. trio_core-0.3.0/research/eval-results/tier1/pope_qwen3-vl-2b_baseline.json +469 -0
  120. trio_core-0.3.0/research/eval-results/tier1/pope_qwen3-vl-4b_baseline.json +469 -0
  121. trio_core-0.3.0/research/eval-results/tier1/pope_qwen3-vl-8b_baseline.json +469 -0
  122. trio_core-0.3.0/research/eval-results/tier1/pope_qwen3.5-0.8b_baseline.json +469 -0
  123. trio_core-0.3.0/research/eval-results/tier1/pope_qwen3.5-2b_baseline.json +469 -0
  124. trio_core-0.3.0/research/eval-results/tier1/pope_qwen3.5-4b_baseline.json +469 -0
  125. trio_core-0.3.0/research/eval-results/tier1/pope_qwen3.5-9b_baseline.json +469 -0
  126. trio_core-0.3.0/research/eval-results/tier1/streammem_nanollava.json +80 -0
  127. trio_core-0.3.0/research/eval-results/tier1/tier1_combined_baseline.json +8 -0
  128. trio_core-0.3.0/research/eval-strategy.md +168 -0
  129. trio_core-0.3.0/research/native-engine-plan.md +201 -0
  130. trio_core-0.3.0/research/phase1-custom-generate.md +215 -0
  131. trio_core-0.3.0/research/poc_attention_analysis.py +291 -0
  132. trio_core-0.3.0/research/tome-implementation-plan.md +278 -0
  133. trio_core-0.3.0/research/visual-token-compression.md +324 -0
  134. trio_core-0.3.0/src/trio_core/__init__.py +32 -0
  135. trio_core-0.3.0/src/trio_core/api/__init__.py +0 -0
  136. trio_core-0.3.0/src/trio_core/api/models.py +170 -0
  137. trio_core-0.3.0/src/trio_core/api/server.py +451 -0
  138. trio_core-0.3.0/src/trio_core/backends.py +746 -0
  139. trio_core-0.3.0/src/trio_core/callbacks.py +80 -0
  140. trio_core-0.3.0/src/trio_core/cli.py +172 -0
  141. trio_core-0.3.0/src/trio_core/compressed_backend.py +264 -0
  142. trio_core-0.3.0/src/trio_core/config.py +103 -0
  143. trio_core-0.3.0/src/trio_core/device.py +188 -0
  144. trio_core-0.3.0/src/trio_core/engine.py +400 -0
  145. trio_core-0.3.0/src/trio_core/eval.py +406 -0
  146. trio_core-0.3.0/src/trio_core/eval_benchmarks.py +1083 -0
  147. trio_core-0.3.0/src/trio_core/fastv_backend.py +593 -0
  148. trio_core-0.3.0/src/trio_core/generate.py +896 -0
  149. trio_core-0.3.0/src/trio_core/model_adapter.py +511 -0
  150. trio_core-0.3.0/src/trio_core/models/__init__.py +1 -0
  151. trio_core-0.3.0/src/trio_core/models/base.py +231 -0
  152. trio_core-0.3.0/src/trio_core/models/loader.py +170 -0
  153. trio_core-0.3.0/src/trio_core/models/qwen2_5_vl/__init__.py +4 -0
  154. trio_core-0.3.0/src/trio_core/models/qwen2_5_vl/config.py +90 -0
  155. trio_core-0.3.0/src/trio_core/models/qwen2_5_vl/language.py +461 -0
  156. trio_core-0.3.0/src/trio_core/models/qwen2_5_vl/qwen2_5_vl.py +132 -0
  157. trio_core-0.3.0/src/trio_core/models/qwen2_5_vl/vision.py +342 -0
  158. trio_core-0.3.0/src/trio_core/models/qwen3_5/__init__.py +2 -0
  159. trio_core-0.3.0/src/trio_core/models/qwen3_5/config.py +96 -0
  160. trio_core-0.3.0/src/trio_core/models/qwen3_5/language.py +667 -0
  161. trio_core-0.3.0/src/trio_core/models/qwen3_5/qwen3_5.py +150 -0
  162. trio_core-0.3.0/src/trio_core/models/qwen3_5/vision.py +5 -0
  163. trio_core-0.3.0/src/trio_core/models/qwen3_vl/__init__.py +4 -0
  164. trio_core-0.3.0/src/trio_core/models/qwen3_vl/config.py +104 -0
  165. trio_core-0.3.0/src/trio_core/models/qwen3_vl/language.py +554 -0
  166. trio_core-0.3.0/src/trio_core/models/qwen3_vl/qwen3_vl.py +161 -0
  167. trio_core-0.3.0/src/trio_core/models/qwen3_vl/vision.py +419 -0
  168. trio_core-0.3.0/src/trio_core/native_vision.py +485 -0
  169. trio_core-0.3.0/src/trio_core/native_vision_standard.py +194 -0
  170. trio_core-0.3.0/src/trio_core/profiles.py +649 -0
  171. trio_core-0.3.0/src/trio_core/streaming_memory.py +319 -0
  172. trio_core-0.3.0/src/trio_core/token_compression.py +206 -0
  173. trio_core-0.3.0/src/trio_core/tome.py +223 -0
  174. trio_core-0.3.0/src/trio_core/tome_backend.py +386 -0
  175. trio_core-0.3.0/src/trio_core/tome_vision.py +393 -0
  176. trio_core-0.3.0/src/trio_core/utils.py +94 -0
  177. trio_core-0.3.0/src/trio_core/video.py +553 -0
  178. trio_core-0.3.0/tests/__init__.py +0 -0
  179. trio_core-0.3.0/tests/test_api.py +297 -0
  180. trio_core-0.3.0/tests/test_backends.py +89 -0
  181. trio_core-0.3.0/tests/test_callbacks.py +59 -0
  182. trio_core-0.3.0/tests/test_device.py +50 -0
  183. trio_core-0.3.0/tests/test_engine.py +139 -0
  184. trio_core-0.3.0/tests/test_fastv.py +279 -0
  185. trio_core-0.3.0/tests/test_model_adapter.py +284 -0
  186. trio_core-0.3.0/tests/test_native_vision.py +286 -0
  187. trio_core-0.3.0/tests/test_prefix_cache.py +350 -0
  188. trio_core-0.3.0/tests/test_profiles.py +133 -0
  189. trio_core-0.3.0/tests/test_stream_capture.py +99 -0
  190. trio_core-0.3.0/tests/test_streaming_memory.py +339 -0
  191. trio_core-0.3.0/tests/test_tome.py +275 -0
  192. trio_core-0.3.0/tests/test_tome_integration.py +366 -0
  193. trio_core-0.3.0/tests/test_video.py +93 -0
@@ -0,0 +1,46 @@
1
+ ---
2
+ name: Bug Report
3
+ about: Report a bug in TrioCore
4
+ title: "[Bug] "
5
+ labels: bug
6
+ assignees: ''
7
+ ---
8
+
9
+ ## Description
10
+
11
+ A clear description of the bug.
12
+
13
+ ## Environment
14
+
15
+ - **OS**: macOS / Linux / Windows
16
+ - **Hardware**: Apple M3 Pro 18GB / RTX 4090 / CPU only
17
+ - **Python**: 3.12
18
+ - **trio-core version**: 0.2.1
19
+ - **Backend**: mlx / transformers
20
+
21
+ ## Model
22
+
23
+ - **Model ID**: `mlx-community/Qwen2.5-VL-3B-Instruct-4bit`
24
+
25
+ ## Steps to Reproduce
26
+
27
+ ```python
28
+ from trio_core import TrioCore
29
+ engine = TrioCore()
30
+ engine.load()
31
+ # ...
32
+ ```
33
+
34
+ ## Expected Behavior
35
+
36
+ What you expected to happen.
37
+
38
+ ## Actual Behavior
39
+
40
+ What actually happened. Include error messages and tracebacks.
41
+
42
+ ## Additional Context
43
+
44
+ - [ ] ToMe enabled? If so, what `r` value?
45
+ - [ ] Streaming or batch inference?
46
+ - [ ] Happens with all models or specific ones?
@@ -0,0 +1,25 @@
1
+ ---
2
+ name: Feature Request
3
+ about: Suggest a new feature or improvement
4
+ title: "[Feature] "
5
+ labels: enhancement
6
+ assignees: ''
7
+ ---
8
+
9
+ ## Problem
10
+
11
+ What problem does this feature solve? What's the use case?
12
+
13
+ ## Proposed Solution
14
+
15
+ Describe how you'd like it to work.
16
+
17
+ ## Alternatives Considered
18
+
19
+ Other approaches you've thought about.
20
+
21
+ ## Additional Context
22
+
23
+ - Related papers or prior art
24
+ - Benchmark data if applicable
25
+ - Impact on existing functionality
@@ -0,0 +1,30 @@
1
+ ---
2
+ name: Model Support Request
3
+ about: Request support for a new VLM model
4
+ title: "[Model] "
5
+ labels: model-support
6
+ assignees: ''
7
+ ---
8
+
9
+ ## Model
10
+
11
+ - **Name**: e.g., Qwen2.5-VL-72B
12
+ - **HuggingFace ID**: e.g., `Qwen/Qwen2.5-VL-72B-Instruct`
13
+ - **MLX variant** (if available): e.g., `mlx-community/Qwen2.5-VL-72B-Instruct-4bit`
14
+
15
+ ## Architecture Details
16
+
17
+ - **Family**: Qwen2.5-VL / Qwen3-VL / Gemma 3 / SmolVLM / Other
18
+ - **Parameter size**: 72B
19
+ - **Vision encoder**: ViT patch size, merge factor
20
+ - **Context window**: 128K
21
+ - **Quantized size**: ~40GB (4-bit)
22
+
23
+ ## Why This Model?
24
+
25
+ What makes this model a good fit for TrioCore?
26
+
27
+ ## Testing
28
+
29
+ - [ ] I have tested basic inference with this model
30
+ - [ ] I can provide benchmark results
@@ -0,0 +1,28 @@
1
+ ## Summary
2
+
3
+ Brief description of what this PR does.
4
+
5
+ ## Changes
6
+
7
+ - Change 1
8
+ - Change 2
9
+
10
+ ## Test Plan
11
+
12
+ - [ ] All existing tests pass (`python -m pytest tests/`)
13
+ - [ ] Regression test passes (`python examples/run_regression.py`) — required for inference changes
14
+ - [ ] New tests added for new functionality
15
+
16
+ ## Benchmarks
17
+
18
+ If this PR affects inference performance or accuracy, include before/after numbers:
19
+
20
+ | Metric | Before | After |
21
+ |--------|--------|-------|
22
+ | POPE accuracy | | |
23
+ | Prefill latency | | |
24
+ | Visual tokens | | |
25
+
26
+ ## Related Issues
27
+
28
+ Closes #
@@ -0,0 +1,28 @@
1
+ name: Publish to PyPI
2
+
3
+ on:
4
+ release:
5
+ types: [published]
6
+
7
+ jobs:
8
+ publish:
9
+ runs-on: ubuntu-latest
10
+ permissions:
11
+ id-token: write # trusted publishing
12
+
13
+ steps:
14
+ - uses: actions/checkout@v4
15
+
16
+ - name: Set up Python
17
+ uses: actions/setup-python@v5
18
+ with:
19
+ python-version: "3.12"
20
+
21
+ - name: Install build tools
22
+ run: pip install hatchling
23
+
24
+ - name: Build wheel
25
+ run: python -m hatchling build -t wheel
26
+
27
+ - name: Publish to PyPI
28
+ uses: pypa/gh-action-pypi-publish@release/v1
@@ -0,0 +1,28 @@
1
+ name: Tests
2
+
3
+ on:
4
+ push:
5
+ branches: [main]
6
+ pull_request:
7
+ branches: [main]
8
+
9
+ jobs:
10
+ test:
11
+ runs-on: ubuntu-latest
12
+ strategy:
13
+ matrix:
14
+ python-version: ["3.10", "3.11", "3.12"]
15
+
16
+ steps:
17
+ - uses: actions/checkout@v4
18
+
19
+ - name: Set up Python ${{ matrix.python-version }}
20
+ uses: actions/setup-python@v5
21
+ with:
22
+ python-version: ${{ matrix.python-version }}
23
+
24
+ - name: Install dependencies
25
+ run: pip install -e '.[dev]'
26
+
27
+ - name: Run tests
28
+ run: python -m pytest tests/ -v
@@ -0,0 +1,16 @@
1
+ __pycache__/
2
+ *.py[cod]
3
+ *$py.class
4
+ *.egg-info/
5
+ dist/
6
+ build/
7
+ .eggs/
8
+ *.egg
9
+ .venv/
10
+ venv/
11
+ .env
12
+ .pytest_cache/
13
+ .mypy_cache/
14
+ *.so
15
+ .DS_Store
16
+ /eval_*.json
@@ -0,0 +1,61 @@
1
+ # Changelog
2
+
3
+ All notable changes to this project will be documented in this file.
4
+
5
+ The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.1.0/),
6
+ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
7
+
8
+ ## [Unreleased]
9
+
10
+ ### Added
11
+ - Webcam GUI example with live video preview and VLM text overlay
12
+ - Accuracy regression test suite (POPE + TextVQA gate)
13
+ - Native engine plan and inference pipeline documentation
14
+
15
+ ## [0.2.1] - 2026-03-06
16
+
17
+ ### Added
18
+ - Qwen3.5 support (0.8B / 4B / 9B profiles, verified ToMe compatibility)
19
+ - Gemma 3 model profiles (4B, 12B) with SigLIP vision encoder parameters
20
+ - SmolVLM model profiles (256M, 500M, 2.2B) for ultra-lightweight inference
21
+ - Pattern matching for HuggingFace model IDs of all supported families
22
+
23
+ ### Changed
24
+ - Generalized Transformers backend: replaced `Qwen2_5_VLForConditionalGeneration` with `AutoModelForVision2Seq` for universal VLM support
25
+ - Added generic PIL-based input path for non-Qwen models (Gemma 3, SmolVLM, etc.)
26
+
27
+ ## [0.2.0] - 2026-03-04
28
+
29
+ ### Added
30
+ - ToMe (Token Merging) visual token compression inside ViT blocks
31
+ - Windowed-attention-aware merging for Qwen2.5-VL
32
+ - Qwen3-VL support with deepstack feature handling
33
+ - Compressed position encoding after token merging
34
+ - POPE and TextVQA benchmark framework
35
+ - Synthetic eval framework (prefill, decode, memory profiling)
36
+ - Benchmark CLI with A/B comparison (`run_benchmark.py`, `run_eval.py`)
37
+
38
+ ### Results
39
+ - Qwen2.5-VL-3B at 1080p: **73% prefill speedup**, 68% token reduction with ToMe r=4
40
+ - Qwen3-VL-4B POPE: **zero quality loss** (91% accuracy) with 31% prefill speedup
41
+
42
+ ## [0.1.0] - 2026-03-01
43
+
44
+ ### Added
45
+ - Core inference engine with three-phase pipeline (preprocess, inference, postprocess)
46
+ - MLX backend for Apple Silicon (M1-M4) via mlx-vlm
47
+ - Transformers backend for NVIDIA GPU and CPU
48
+ - Hardware auto-detection and model recommendation
49
+ - StreamCapture for live video streams (RTSP, YouTube, webcam)
50
+ - Temporal deduplication (normalized L2 on 64x64 downscale)
51
+ - Motion gating (frame differencing + EMA background)
52
+ - Model profiles with architecture-aware parameters
53
+ - FastAPI server with OpenAI-compatible chat endpoint
54
+ - CLI: `trio-core serve`, `trio-core analyze`, `trio-core device`
55
+ - Callback system with 10 lifecycle events
56
+ - 120 unit tests
57
+
58
+ [Unreleased]: https://github.com/machinefi/trio-core/compare/v0.2.1...HEAD
59
+ [0.2.1]: https://github.com/machinefi/trio-core/compare/v0.2.0...v0.2.1
60
+ [0.2.0]: https://github.com/machinefi/trio-core/compare/v0.1.0...v0.2.0
61
+ [0.1.0]: https://github.com/machinefi/trio-core/releases/tag/v0.1.0
@@ -0,0 +1,102 @@
1
+ # Contributing to TrioCore
2
+
3
+ Thank you for your interest in contributing to TrioCore! This guide will help you get started.
4
+
5
+ ## Development Setup
6
+
7
+ ```bash
8
+ # Clone
9
+ git clone https://github.com/machinefi/trio-core.git
10
+ cd trio-core
11
+
12
+ # Create virtual environment
13
+ python -m venv .venv
14
+ source .venv/bin/activate
15
+
16
+ # Install in development mode
17
+ pip install -e '.[dev]'
18
+
19
+ # Apple Silicon: also install MLX backend
20
+ pip install -e '.[mlx,dev]'
21
+
22
+ # NVIDIA GPU: also install Transformers backend
23
+ pip install -e '.[transformers,dev]'
24
+ ```
25
+
26
+ ## Running Tests
27
+
28
+ ```bash
29
+ # Run all tests
30
+ python -m pytest tests/ -v
31
+
32
+ # Run specific test file
33
+ python -m pytest tests/test_tome.py -v
34
+
35
+ # Run with coverage
36
+ python -m pytest tests/ --cov=trio_core
37
+ ```
38
+
39
+ ## Running Benchmarks
40
+
41
+ ```bash
42
+ # Quick accuracy check (50 samples, ~2 min)
43
+ python examples/run_regression.py --save-baseline
44
+
45
+ # Full POPE benchmark
46
+ python examples/run_benchmark.py --bench pope --samples 100
47
+
48
+ # Performance eval
49
+ python examples/run_eval.py --resolution 1080p --runs 3
50
+ ```
51
+
52
+ ## Code Style
53
+
54
+ - Python 3.10+ with type hints
55
+ - Use `from __future__ import annotations` for forward references
56
+ - Keep functions focused and short
57
+ - No unnecessary abstractions — three similar lines > premature helper
58
+ - Comments only where logic isn't self-evident
59
+
60
+ ## Pull Request Process
61
+
62
+ 1. **Fork** the repository and create a feature branch from `main`
63
+ 2. **Write tests** for any new functionality
64
+ 3. **Run the test suite** — all 120+ tests must pass
65
+ 4. **Run regression tests** if changing inference code:
66
+ ```bash
67
+ python examples/run_regression.py
68
+ ```
69
+ 5. **Keep PRs focused** — one feature or fix per PR
70
+ 6. **Write a clear description** of what and why
71
+
72
+ ## What to Contribute
73
+
74
+ ### High Impact
75
+ - Bug fixes in inference pipeline
76
+ - New model support (profiles + backend testing)
77
+ - Performance optimizations with benchmarks
78
+ - Accuracy improvements on POPE/TextVQA
79
+
80
+ ### Welcome
81
+ - Documentation improvements
82
+ - New examples and usage patterns
83
+ - CI/CD improvements
84
+ - Test coverage for edge cases
85
+
86
+ ### Please Discuss First
87
+ - Architecture changes (open an issue)
88
+ - New dependencies
89
+ - API breaking changes
90
+
91
+ ## Reporting Issues
92
+
93
+ Use the [issue templates](.github/ISSUE_TEMPLATE/) to report bugs or request features. Include:
94
+
95
+ - Python version, OS, hardware (M1/M2/M3/M4, GPU model)
96
+ - Model being used
97
+ - Minimal reproduction steps
98
+ - Expected vs actual behavior
99
+
100
+ ## License
101
+
102
+ By contributing, you agree that your contributions will be licensed under the Apache License 2.0.
@@ -0,0 +1,190 @@
1
+ Apache License
2
+ Version 2.0, January 2004
3
+ http://www.apache.org/licenses/
4
+
5
+ TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
6
+
7
+ 1. Definitions.
8
+
9
+ "License" shall mean the terms and conditions for use, reproduction,
10
+ and distribution as defined by Sections 1 through 9 of this document.
11
+
12
+ "Licensor" shall mean the copyright owner or entity authorized by
13
+ the copyright owner that is granting the License.
14
+
15
+ "Legal Entity" shall mean the union of the acting entity and all
16
+ other entities that control, are controlled by, or are under common
17
+ control with that entity. For the purposes of this definition,
18
+ "control" means (i) the power, direct or indirect, to cause the
19
+ direction or management of such entity, whether by contract or
20
+ otherwise, or (ii) ownership of fifty percent (50%) or more of the
21
+ outstanding shares, or (iii) beneficial ownership of such entity.
22
+
23
+ "You" (or "Your") shall mean an individual or Legal Entity
24
+ exercising permissions granted by this License.
25
+
26
+ "Source" form shall mean the preferred form for making modifications,
27
+ including but not limited to software source code, documentation
28
+ source, and configuration files.
29
+
30
+ "Object" form shall mean any form resulting from mechanical
31
+ transformation or translation of a Source form, including but
32
+ not limited to compiled object code, generated documentation,
33
+ and conversions to other media types.
34
+
35
+ "Work" shall mean the work of authorship, whether in Source or
36
+ Object form, made available under the License, as indicated by a
37
+ copyright notice that is included in or attached to the work
38
+ (an example is provided in the Appendix below).
39
+
40
+ "Derivative Works" shall mean any work, whether in Source or Object
41
+ form, that is based on (or derived from) the Work and for which the
42
+ editorial revisions, annotations, elaborations, or other modifications
43
+ represent, as a whole, an original work of authorship. For the purposes
44
+ of this License, Derivative Works shall not include works that remain
45
+ separable from, or merely link (or bind by name) to the interfaces of,
46
+ the Work and Derivative Works thereof.
47
+
48
+ "Contribution" shall mean any work of authorship, including
49
+ the original version of the Work and any modifications or additions
50
+ to that Work or Derivative Works thereof, that is intentionally
51
+ submitted to the Licensor for inclusion in the Work by the copyright owner
52
+ or by an individual or Legal Entity authorized to submit on behalf of
53
+ the copyright owner. For the purposes of this definition, "submitted"
54
+ means any form of electronic, verbal, or written communication sent
55
+ to the Licensor or its representatives, including but not limited to
56
+ communication on electronic mailing lists, source code control systems,
57
+ and issue tracking systems that are managed by, or on behalf of, the
58
+ Licensor for the purpose of discussing and improving the Work, but
59
+ excluding communication that is conspicuously marked or otherwise
60
+ designated in writing by the copyright owner as "Not a Contribution."
61
+
62
+ "Contributor" shall mean Licensor and any individual or Legal Entity
63
+ on behalf of whom a Contribution has been received by the Licensor and
64
+ subsequently incorporated within the Work.
65
+
66
+ 2. Grant of Copyright License. Subject to the terms and conditions of
67
+ this License, each Contributor hereby grants to You a perpetual,
68
+ worldwide, non-exclusive, no-charge, royalty-free, irrevocable
69
+ copyright license to reproduce, prepare Derivative Works of,
70
+ publicly display, publicly perform, sublicense, and distribute the
71
+ Work and such Derivative Works in Source or Object form.
72
+
73
+ 3. Grant of Patent License. Subject to the terms and conditions of
74
+ this License, each Contributor hereby grants to You a perpetual,
75
+ worldwide, non-exclusive, no-charge, royalty-free, irrevocable
76
+ (except as stated in this section) patent license to make, have made,
77
+ use, offer to sell, sell, import, and otherwise transfer the Work,
78
+ where such license applies only to those patent claims licensable
79
+ by such Contributor that are necessarily infringed by their
80
+ Contribution(s) alone or by combination of their Contribution(s)
81
+ with the Work to which such Contribution(s) was submitted. If You
82
+ institute patent litigation against any entity (including a
83
+ cross-claim or counterclaim in a lawsuit) alleging that the Work
84
+ or a Contribution incorporated within the Work constitutes direct
85
+ or contributory patent infringement, then any patent licenses
86
+ granted to You under this License for that Work shall terminate
87
+ as of the date such litigation is filed.
88
+
89
+ 4. Redistribution. You may reproduce and distribute copies of the
90
+ Work or Derivative Works thereof in any medium, with or without
91
+ modifications, and in Source or Object form, provided that You
92
+ meet the following conditions:
93
+
94
+ (a) You must give any other recipients of the Work or
95
+ Derivative Works a copy of this License; and
96
+
97
+ (b) You must cause any modified files to carry prominent notices
98
+ stating that You changed the files; and
99
+
100
+ (c) You must retain, in the Source form of any Derivative Works
101
+ that You distribute, all copyright, patent, trademark, and
102
+ attribution notices from the Source form of the Work,
103
+ excluding those notices that do not pertain to any part of
104
+ the Derivative Works; and
105
+
106
+ (d) If the Work includes a "NOTICE" text file as part of its
107
+ distribution, then any Derivative Works that You distribute must
108
+ include a readable copy of the attribution notices contained
109
+ within such NOTICE file, excluding any notices that do not
110
+ pertain to any part of the Derivative Works, in at least one
111
+ of the following places: within a NOTICE text file distributed
112
+ as part of the Derivative Works; within the Source form or
113
+ documentation, if provided along with the Derivative Works; or,
114
+ within a display generated by the Derivative Works, if and
115
+ wherever such third-party notices normally appear. The contents
116
+ of the NOTICE file are for informational purposes only and
117
+ do not modify the License. You may add Your own attribution
118
+ notices within Derivative Works that You distribute, alongside
119
+ or as an addendum to the NOTICE text from the Work, provided
120
+ that such additional attribution notices cannot be construed
121
+ as modifying the License.
122
+
123
+ You may add Your own copyright statement to Your modifications and
124
+ may provide additional or different license terms and conditions
125
+ for use, reproduction, or distribution of Your modifications, or
126
+ for any such Derivative Works as a whole, provided Your use,
127
+ reproduction, and distribution of the Work otherwise complies with
128
+ the conditions stated in this License.
129
+
130
+ 5. Submission of Contributions. Unless You explicitly state otherwise,
131
+ any Contribution intentionally submitted for inclusion in the Work
132
+ by You to the Licensor shall be under the terms and conditions of
133
+ this License, without any additional terms or conditions.
134
+ Notwithstanding the above, nothing herein shall supersede or modify
135
+ the terms of any separate license agreement you may have executed
136
+ with Licensor regarding such Contributions.
137
+
138
+ 6. Trademarks. This License does not grant permission to use the trade
139
+ names, trademarks, service marks, or product names of the Licensor,
140
+ except as required for reasonable and customary use in describing the
141
+ origin of the Work and reproducing the content of the NOTICE file.
142
+
143
+ 7. Disclaimer of Warranty. Unless required by applicable law or
144
+ agreed to in writing, Licensor provides the Work (and each
145
+ Contributor provides its Contributions) on an "AS IS" BASIS,
146
+ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
147
+ implied, including, without limitation, any warranties or conditions
148
+ of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
149
+ PARTICULAR PURPOSE. You are solely responsible for determining the
150
+ appropriateness of using or redistributing the Work and assume any
151
+ risks associated with Your exercise of permissions under this License.
152
+
153
+ 8. Limitation of Liability. In no event and under no legal theory,
154
+ whether in tort (including negligence), contract, or otherwise,
155
+ unless required by applicable law (such as deliberate and grossly
156
+ negligent acts) or agreed to in writing, shall any Contributor be
157
+ liable to You for damages, including any direct, indirect, special,
158
+ incidental, or consequential damages of any character arising as a
159
+ result of this License or out of the use or inability to use the
160
+ Work (including but not limited to damages for loss of goodwill,
161
+ work stoppage, computer failure or malfunction, or any and all
162
+ other commercial damages or losses), even if such Contributor
163
+ has been advised of the possibility of such damages.
164
+
165
+ 9. Accepting Warranty or Additional Liability. While redistributing
166
+ the Work or Derivative Works thereof, You may choose to offer,
167
+ and charge a fee for, acceptance of support, warranty, indemnity,
168
+ or other liability obligations and/or rights consistent with this
169
+ License. However, in accepting such obligations, You may act only
170
+ on Your own behalf and on Your sole responsibility, not on behalf
171
+ of any other Contributor, and only if You agree to indemnify,
172
+ defend, and hold each Contributor harmless for any liability
173
+ incurred by, or claims asserted against, such Contributor by reason
174
+ of your accepting any such warranty or additional liability.
175
+
176
+ END OF TERMS AND CONDITIONS
177
+
178
+ Copyright 2025 Trio Team
179
+
180
+ Licensed under the Apache License, Version 2.0 (the "License");
181
+ you may not use this file except in compliance with the License.
182
+ You may obtain a copy of the License at
183
+
184
+ http://www.apache.org/licenses/LICENSE-2.0
185
+
186
+ Unless required by applicable law or agreed to in writing, software
187
+ distributed under the License is distributed on an "AS IS" BASIS,
188
+ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
189
+ See the License for the specific language governing permissions and
190
+ limitations under the License.