support speculative decoding kernel in sgl-kernel

[zhyncs]

Motivation

• support build_tree_kernel build_tree_kernel_efficient tree_speculative_sampling_target_only by @Ying1123
• bump v0.0.3.post2

sglang git:(ying-spec) python3 python/sglang/srt/speculative/build_eagle_tree.py
=========== build tree kernel ==========
position=tensor([ 5,  6,  6,  7,  7,  8,  8,  9, 10, 11, 12, 12, 12, 12, 13, 14],
       device='cuda:0')
retrive_index=tensor([[ 0, -1, -1, -1, -1, -1],
        [ 0,  2,  4,  6, -1, -1],
        [ 0,  1,  3,  5,  7, -1],
        [ 8, -1, -1, -1, -1, -1],
        [ 8,  9, 10, -1, -1, -1],
        [ 8,  9, 12, -1, -1, -1],
        [ 8,  9, 13, -1, -1, -1],
        [ 8,  9, 11, 14, 15, -1]], device='cuda:0')
retrive_cum_len=tensor([0, 3, 8], device='cuda:0', dtype=torch.int32)
draft_tokens=tensor([29974, 29896, 29906, 29889, 29974, 29946, 29896, 29946,    13,    13,
        22550,  4136, 16492,  8439, 29871, 29941], device='cuda:0')
=========== build tree kernel efficient ==========
position=tensor([ 5,  6,  6,  7,  7,  8,  8,  9, 10, 11, 12, 12, 12, 12, 13, 14],
       device='cuda:0')
retrive_index=tensor([[ 0,  1,  2,  3,  4,  5,  6,  7],
        [ 8,  9, 10, 11, 12, 13, 14, 15]], device='cuda:0')
retrive_next_token=tensor([[ 1,  3,  4,  5,  6,  7, -1, -1],
        [ 1,  2, -1,  6, -1, -1,  7, -1]], device='cuda:0')
retrive_next_sibling=tensor([[-1,  2, -1, -1, -1, -1, -1, -1],
        [-1, -1,  3,  4,  5, -1, -1, -1]], device='cuda:0')
draft_tokens=tensor([29974, 29896, 29906, 29889, 29974, 29946, 29896, 29946,    13,    13,
        22550,  4136, 16492,  8439, 29871, 29941], device='cuda:0')

sgl-kernel git:(ying-spec) python3 tests/test_speculative_sampling.py
candidates=tensor([[ 0,  1,  2,  3,  4,  5],
        [ 7,  8,  9, 10, 11, 12]], device='cuda:0', dtype=torch.int32)
retrive_index=tensor([[ 0,  1,  2,  3,  4,  5],
        [ 6,  7,  8,  9, 10, 11]], device='cuda:0', dtype=torch.int32)
retrive_next_token=tensor([[ 1,  2, -1,  4,  5, -1],
        [ 4,  2,  3, -1,  5, -1]], device='cuda:0', dtype=torch.int32)
retrive_next_sibling=tensor([[-1,  3, -1, -1, -1, -1],
        [-1, -1, -1, -1,  1, -1]], device='cuda:0', dtype=torch.int32)
coins=tensor([[0.0536, 0.7639, 0.4346, 0.6656, 0.2928, 0.9630],
        [0.0203, 0.2049, 0.6945, 0.1469, 0.6007, 0.9195]], device='cuda:0')
predicts=tensor([ 3, -1, -1,  4,  5, 18, 11, -1, -1, -1, 12, 18], device='cuda:0',
       dtype=torch.int32)
accept_index=tensor([[ 0,  3,  4,  5],
        [ 6, 10, 11, -1]], device='cuda:0', dtype=torch.int32)
accept_token_num=tensor([3, 2], device='cuda:0', dtype=torch.int32)

Modifications

Checklist

• Format your code according to the Code Formatting with Pre-Commit.
• Add unit tests as outlined in the Running Unit Tests.
• Update documentation / docstrings / example tutorials as needed, according to Writing Documentation.
• Provide throughput / latency benchmark results and accuracy evaluation results as needed, according to Benchmark and Profiling and Accuracy Results.
• For reviewers: If you haven't made any contributions to this PR and are only assisting with merging the main branch, please remove yourself as a co-author when merging the PR.