tpu-inference 0.0.1rc1__py3-none-any.whl → 0.11.1.dev202511130813__py3-none-any.whl

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.

Potentially problematic release.

This version of tpu-inference might be problematic. Click here for more details.

Files changed (67) hide show
  1. tests/kernels/fused_moe_v1_test.py +34 -303
  2. tests/kernels/ragged_paged_attention_kernel_v3_hd64_test.py +2 -2
  3. tests/lora/test_layers.py +6 -0
  4. tests/lora/utils.py +8 -0
  5. tests/test_utils.py +16 -24
  6. tpu_inference/__init__.py +3 -22
  7. tpu_inference/core/core_tpu.py +9 -17
  8. tpu_inference/core/disagg_utils.py +8 -6
  9. tpu_inference/distributed/tpu_connector.py +4 -3
  10. tpu_inference/distributed/utils.py +2 -3
  11. tpu_inference/envs.py +8 -61
  12. tpu_inference/executors/ray_distributed_executor.py +11 -31
  13. tpu_inference/kernels/fused_moe/v1/kernel.py +110 -641
  14. tpu_inference/kernels/ragged_paged_attention/v3/kernel.py +54 -77
  15. tpu_inference/kernels/ragged_paged_attention/v3/kernel_hd64.py +143 -287
  16. tpu_inference/kernels/ragged_paged_attention/v3/tuned_block_sizes.py +0 -7
  17. tpu_inference/layers/jax/attention/attention.py +1 -1
  18. tpu_inference/layers/{common → jax}/attention_interface.py +2 -8
  19. tpu_inference/layers/jax/sample/rejection_sampler.py +1 -1
  20. tpu_inference/layers/jax/sample/sampling.py +2 -2
  21. tpu_inference/layers/{common → jax}/sharding.py +5 -5
  22. tpu_inference/layers/vllm/attention.py +1 -1
  23. tpu_inference/layers/vllm/fused_moe.py +208 -170
  24. tpu_inference/layers/vllm/quantization/__init__.py +3 -7
  25. tpu_inference/layers/vllm/quantization/awq.py +3 -4
  26. tpu_inference/layers/vllm/quantization/common.py +1 -6
  27. tpu_inference/layers/vllm/quantization/compressed_tensors/compressed_tensors.py +2 -4
  28. tpu_inference/layers/vllm/quantization/unquantized.py +67 -62
  29. tpu_inference/layers/vllm/sharding.py +2 -2
  30. tpu_inference/lora/torch_punica_tpu.py +2 -1
  31. tpu_inference/mock/__init__.py +0 -0
  32. tpu_inference/mock/vllm_config_utils.py +28 -0
  33. tpu_inference/mock/vllm_envs.py +1219 -0
  34. tpu_inference/mock/vllm_logger.py +212 -0
  35. tpu_inference/mock/vllm_logging_utils.py +15 -0
  36. tpu_inference/models/common/model_loader.py +12 -46
  37. tpu_inference/models/jax/llama3.py +3 -4
  38. tpu_inference/models/jax/llama_eagle3.py +5 -8
  39. tpu_inference/models/jax/phi3.py +376 -0
  40. tpu_inference/models/jax/qwen2.py +2 -3
  41. tpu_inference/models/jax/qwen2_5_vl.py +50 -165
  42. tpu_inference/models/jax/qwen3.py +2 -3
  43. tpu_inference/models/jax/utils/quantization/quantization_utils.py +6 -3
  44. tpu_inference/models/jax/utils/weight_utils.py +143 -198
  45. tpu_inference/models/vllm/vllm_model_wrapper.py +14 -32
  46. tpu_inference/platforms/tpu_platform.py +34 -47
  47. tpu_inference/runner/compilation_manager.py +60 -145
  48. tpu_inference/runner/kv_cache.py +2 -2
  49. tpu_inference/runner/kv_cache_manager.py +18 -17
  50. tpu_inference/runner/persistent_batch_manager.py +2 -40
  51. tpu_inference/runner/structured_decoding_manager.py +3 -2
  52. tpu_inference/runner/tpu_runner.py +135 -283
  53. tpu_inference/runner/utils.py +2 -2
  54. tpu_inference/spec_decode/jax/eagle3.py +21 -71
  55. tpu_inference/tpu_info.py +3 -4
  56. tpu_inference/utils.py +15 -38
  57. tpu_inference/worker/tpu_worker.py +26 -163
  58. {tpu_inference-0.0.1rc1.dist-info → tpu_inference-0.11.1.dev202511130813.dist-info}/METADATA +3 -4
  59. {tpu_inference-0.0.1rc1.dist-info → tpu_inference-0.11.1.dev202511130813.dist-info}/RECORD +63 -61
  60. tests/test_envs.py +0 -203
  61. tpu_inference/layers/common/quant_methods.py +0 -8
  62. tpu_inference/layers/vllm/quantization/mxfp4.py +0 -331
  63. tpu_inference/models/jax/llama_guard_4.py +0 -361
  64. /tpu_inference/layers/{common → jax}/binary_search.py +0 -0
  65. {tpu_inference-0.0.1rc1.dist-info → tpu_inference-0.11.1.dev202511130813.dist-info}/WHEEL +0 -0
  66. {tpu_inference-0.0.1rc1.dist-info → tpu_inference-0.11.1.dev202511130813.dist-info}/licenses/LICENSE +0 -0
  67. {tpu_inference-0.0.1rc1.dist-info → tpu_inference-0.11.1.dev202511130813.dist-info}/top_level.txt +0 -0