Google Git
Sign in
chromium/external/github.com/google-ai-edge/LiteRT-LM.git/HEAD
  1. ce0fbc3 fix: remove thinking content handling from Qwen3 chat template by Wai Hon Law · 13 hours ago upstream/main
  2. 372ff7e feat: set filterChannelContentFromKvCache to true by default by Wai Hon Law · 13 hours ago
  3. 7355f6d Add large Gemma 4 models (12B, 26B, 31B) and VRAM warning to LiteRT-LM chat app. by Google AI Edge · 14 hours ago
  4. f35fcbc feat: default filter_channel_content_from_kv_cache to None across CLI and language bindings by Wai Hon Law · 17 hours ago
  5. df23d63 Internal change to enable support for Google Tensor backend by Google AI Edge · 17 hours ago
  6. c74608b Centralize CPU and GPU option configurations across LiteRT-LM executors by Salil Tambe · 17 hours ago
  7. 587d1d6 Add response_format API for constrained decoding (JSON Schema and Regex) in Kotlin. by Wai Hon Law · 20 hours ago
  8. e94d740 Internal change by Wai Hon Law · 20 hours ago
  9. 0e380a5 On Windows, `ERROR` is defined as a macro. When passed as an argument, the preprocessor expands `ERROR` prematurely, resulting in build breaks. by Google AI Edge · 21 hours ago
  10. f947287 Add a flag to set gpu_enable_metal_residency_set by Byungchul Kim · 21 hours ago
  11. 5dc3536 [litertlm] Add SuppressTokensConfig to LiteRT-LM Kotlin API. by Bangcheng Wang · 21 hours ago
  12. 92f3caa Add support for GPU LoRA weight verification for audio inputs and AST testing. by Salil Tambe · 22 hours ago
  13. 2d816fb refactor: extract LiteRT-LM C API internal structs into shared engine_internal.h by Wai Hon Law · 22 hours ago
  14. 21a1eba [litertlm] Add NoRepeatNgramConfig to LiteRtLm Kotlin API. by Bangcheng Wang · 32 hours ago
  15. 70f8523 Set `self.handle = nil` in `Conversation.deinit` and `Engine.deinit` before native deallocation to prevent re-entry or double-free errors during teardown. by Yishuang Pang · 33 hours ago
  16. 36c721f fix: support thinking budget for Qwen models and dynamically select first channel as thinking channel by Wai Hon Law · 34 hours ago
  17. 4f2cae6 Reorder constructor initializer list in LlmLiteRtNpuCompiledModelExecutor to match class field declaration order and prevent compiler reordering warnings (-Wreorder-ctor). by Google AI Edge · 35 hours ago
  18. 33e0b35 fix: support padded vocabulary sizes in ConstrainedDecoder by Wai Hon Law · 2 days ago
  19. 071264d Support response_format API in LiteRT-LM OpenAI handler. by Wai Hon Law · 2 days ago
  20. e5ab46e NPU execution: Update HW mask for sliding window attention by Matt Kreileder · 2 days ago
  21. 2bedb7b Set NPU backend in CompiledModel options by David Massoud · 2 days ago
  22. d75150e Downgrade most ABSL_LOG(INFO) to ABSL_VLOG. by Google AI Edge · 2 days ago
  23. 9643e94 Internal code update by Tenghui Zhu · 2 days ago
  24. 719ddb4 [litertlm] Add RepetitionPenaltyConfig to LiteRtLm Kotlin API. by Bangcheng Wang · 2 days ago
  25. 3f67cd3 Internal code update by Tenghui Zhu · 2 days ago
  26. c7adc1b Add metadata for LiteRT-LM embedding models. by Matthew Chan · 2 days ago
  27. 4698342 Expose op profiling symbols to litert windows DLL. by Google AI Edge · 2 days ago
  28. 325d683 Add GetTFLiteModelSectionOffsetAndSize to ModelResources. by David Massoud · 2 days ago
  29. 28441e2 Enable residency set configuration. by Fengwu Yao · 2 days ago
  30. d116ee4 Internal changes only. by Fengwu Yao · 2 days ago
  31. 1bdd7da fix: update LiteRtLmConversationConfig test struct layout to match C engine header by Wai Hon Law · 2 days ago
  32. e398f12 NPU: Initial implementation for sliding window attention support in HW cache update by Matt Kreileder · 2 days ago
  33. fddcd26 Add residency API setting. by Fengwu Yao · 2 days ago
  34. 4771083 Fix Windows model download path resolution in api_test.py using shutil.which. by Google AI Edge · 2 days ago
  35. 2a51e94 fix(cli): allow additional properties in config.json schema by Wai Hon Law · 2 days ago
  36. 5b92c80 Rename verify_api.py to api_test.py, configure GCS model downloads, fix Kokoro apt lock timeouts. by Google AI Edge · 2 days ago
  37. ef16b76 This is an internal change by Google AI Edge · 3 days ago
  38. 5ae2b32 Internal change by Google AI Edge · 3 days ago
  39. ef12fc2 feat(python): add response format API by Wai Hon Law · 3 days ago
  40. 13df7ab This is an internal change by Google AI Edge · 3 days ago
  41. 98eba9b Skip macOS gated model download and pytest on unauthenticated pull requests. by Salil Tambe · 3 days ago
  42. 2a568b8 feat(cli): add --chat-template option to litert-lm run by Wai Hon Law · 3 days ago
  43. 3852882 Implement single-turn prompt template rendering for GenericDataProcessor by delegating to RenderSingleTurnTemplateCommon. Also add unit tests verifying RenderSingleTurnTemplate for single-turn rendering and appending message flows. by Google AI Edge · 3 days ago
  44. 4991db6 feat(cli): add 9 new config options to config.json and validate via JSON Schema by Wai Hon Law · 3 days ago
  45. bea5562 [litertlm] Add SuppressTokensConfig to LiteRT LM C API and Python bindings. by Bangcheng Wang · 3 days ago
  46. 6340ece Fix Hugging Face S3 redirect 403 authorization failures on macOS and Linux CI workflows by Salil Tambe · 3 days ago
  47. 8cafae3 Update GetOptimizedPrefillWorkGroups to use cautious greedy algorithm by Marissa Ikonomidis · 3 days ago
  48. f871d8d [litertlm] Add NoRepeatNgramConfig to LiteRT LM C and Python APIs. by Bangcheng Wang · 3 days ago
  49. 4790839 Prepare only 1 step ahead of command buffer for single KV cache case by Byungchul Kim · 3 days ago
  50. 5c17868 Optimize MTP rejection sampling verifier batch locking by Salil Tambe · 3 days ago
  51. e5f9ba7 feat(cli): add pack and unpack commands and litertlm_builder.pack API by Wai Hon Law · 3 days ago
  52. 69cb398 feat(cli): add --speculative-decoding option and deprecate --enable-speculative-decoding; update --thinking options by Wai Hon Law · 3 days ago
  53. 44d2a22 Update CI build to download model from GCS instead of Hugging Face. by Google AI Edge · 3 days ago
  54. da33e30 feat(cli): support configuration via config.json and validation via JSON Schema by Wai Hon Law · 4 days ago
  55. db900bd Reset current context and executor if only session is released by Byungchul Kim · 4 days ago
  56. 020a3c0 Guard HuggingFaceTokenizer include with a build flag by Google AI Edge · 4 days ago
  57. 83a91b5 Update dependencies of litert_lm by Byungchul Kim · 6 days ago
  58. 156e8d8 feat(cli): use blue for all channel contents and print [/channel-name] at channel end by Wai Hon Law · 7 days ago
  59. 840fe9e Correct wrong condition to enable sampler handling input by Byungchul Kim · 7 days ago
  60. 056d1ba Internal changes only by Bangcheng Wang · 7 days ago
  61. 1d3024f Update checksums for LiteRT-LM v0.14.0 binary targets. by Google AI Edge · 7 days ago
  62. ad53ed1 Allow configurable NPU KV cache initialization value via model metadata. by Andrew Zhang · 7 days ago upstream/v0.15.0-alpha0
  63. 36af951 ci: add -c opt and unify bazel disk and repository cache paths in macOS CI workflow by Wai Hon Law · 8 days ago
  64. 5c1cd36 [litertlm] Add RepetitionPenaltyConfig to LiteRT LM C and Python APIs. by Bangcheng Wang · 8 days ago
  65. 60c408e style(cli): format non-streaming OpenAI handler JSON responses with 2-space indentation and trailing newline by Wai Hon Law · 8 days ago
  66. 105dcba Support varying audio chunk size for streaming audio encoder models by Yi-Chun Kuo · 8 days ago
  67. b0ac2f9 Reverts df914b949e728bced8059506d45736024720a51c by Fengwu Yao · 8 days ago
  68. 4c48743 Create test_lm_multi_prefill.litertlm for testing multi-prefill. by Matthew Chan · 8 days ago
  69. 8a4a12b Support varying audio chunk size for streaming audio encoder models by Google AI Edge · 8 days ago
  70. 862d3c0 feat(cli): add reasoning_tokens to usage in OpenAI chat completions endpoint by Wai Hon Law · 8 days ago
  71. eeb4802 Enable LFM on-device testing for Exynos NPU by Salil Tambe · 8 days ago
  72. ba9a470 feat(swift): expose ThinkingConfig and thinking token budget in Swift API by Wai Hon Law · 8 days ago
  73. df914b9 Update dependencies of litert_lm by Fengwu Yao · 8 days ago
  74. ca4dadc Automated Code Change by Google AI Edge · 8 days ago
  75. c8bc396 Support varying audio chunk size for streaming audio encoder models by Yi-Chun Kuo · 8 days ago
  76. 5dd1326 Fix premature callback invocation and empty prefill hang in Conversation. by Google AI Edge · 8 days ago
  77. 910fb26 [litertlm] Support completion_usage object in OpenAI server handler by Wai Hon Law · 9 days ago
  78. e62cbd2 feat(cli): support reasoning_effort parameter in OpenAI server handler by Wai Hon Law · 9 days ago
  79. 4174318 Bump version to 0.15.0 by Google AI Edge · 9 days ago
  80. fa7268c Fix msan failure by initializing memory of audio buffer in cached_session_test. by Matthew Chan · 9 days ago
  81. c27a4a8 Add LaTeX rendering to LiteRT-LM web chat app by Google AI Edge · 9 days ago
  82. 5893966 feat(kotlin): expose ThinkingConfig and thinking token budget in Kotlin API by Wai Hon Law · 9 days ago
  83. 2e4796b feat(cli): group thinking related options under the --thinking prefix by Wai Hon Law · 9 days ago
  84. d2b8a38 feat(kotlin): add maxOutputToken API to ConversationConfig and sendMessage by Wai Hon Law · 9 days ago
  85. dd411af fix(python): provide distinct error messages for different NPU initialization failures by Wai Hon Law · 9 days ago
  86. 440c37a ci: add --repository_cache flag and paths across Linux, Android, and macOS workflows by Wai Hon Law · 9 days ago
  87. 305324b ci(win): optimize execution time of Windows ML-CI workflow by ~47% by Wai Hon Law · 9 days ago
  88. 7ba7ac9 LiteRT-LM NPU: Support models with sliding window attention by Matt Kreileder · 9 days ago
  89. a8eb6fd [litertlm] Add --no_repeat_ngram_size` and `--no_repeat_ngram_window_size flags to LiteRT LM. by Bangcheng Wang · 9 days ago
  90. 089d44e [litertlm] Integrate NoRepeatNgramConfig into the decoding pipeline. by Bangcheng Wang · 9 days ago
  91. c2ab9ab feat(python): add C and Python APIs for thinking token budget and ThinkingConfig by Wai Hon Law · 9 days ago
  92. aece6ae [litertlm] Refactor RepetitionPenaltyConfig and SuppressTokensConfig handling. by Bangcheng Wang · 9 days ago
  93. 3c0d244 Add Web Demo link to README. by Google AI Edge · 10 days ago
  94. cc6660e Use filesystem API for local models in Web demo (if available) by Google AI Edge · 10 days ago
  95. 2fd31fb Add option to upload your own model to litertlm web chat demo by Google AI Edge · 10 days ago
  96. aaf44be Lazy-load `per_layer_embedding_lookup_manager_` in NPU executor by Salil Tambe · 10 days ago
  97. 0d4272f refactor: use strongly typed C enums for log severity and activation data type by Wai Hon Law · 10 days ago
  98. eaccbaf Expose session and conversation cloning to the LiteRT-LM JS API. by Google AI Edge · 10 days ago
  99. cf6b506 Ensure LiteRT-LM builder and CLI reject writes to CNS paths by Google AI Edge · 10 days ago
  100. 6ee714b Update LiteRT-LM binary targets to v0.14.0. by Google AI Edge · 10 days ago