| .. |
|
01-distributed_communication_op.py.diff
|
0d8166de95
port DCP2 onto our tree (22-file superset wrapper + a2a recipe); A/B: 162/176, 5.34M-token KV @512K, -5 vs b12x — the high-KV lane
|
1 月之前 |
|
02-model_executor_kernels_attention_dsa_candidate_blocks.py.diff
|
0d8166de95
port DCP2 onto our tree (22-file superset wrapper + a2a recipe); A/B: 162/176, 5.34M-token KV @512K, -5 vs b12x — the high-KV lane
|
1 月之前 |
|
03-model_executor_layers_fused_moe_b12x.py.diff
|
0d8166de95
port DCP2 onto our tree (22-file superset wrapper + a2a recipe); A/B: 162/176, 5.34M-token KV @512K, -5 vs b12x — the high-KV lane
|
1 月之前 |
|
04-model_executor_layers_sparse_attn_indexer.py.diff
|
0d8166de95
port DCP2 onto our tree (22-file superset wrapper + a2a recipe); A/B: 162/176, 5.34M-token KV @512K, -5 vs b12x — the high-KV lane
|
1 月之前 |
|
06-models_deepseek_v4_1_attention.py.diff
|
0d8166de95
port DCP2 onto our tree (22-file superset wrapper + a2a recipe); A/B: 162/176, 5.34M-token KV @512K, -5 vs b12x — the high-KV lane
|
1 月之前 |
|
07-models_deepseek_v4_1_common_engram.py.diff
|
0d8166de95
port DCP2 onto our tree (22-file superset wrapper + a2a recipe); A/B: 162/176, 5.34M-token KV @512K, -5 vs b12x — the high-KV lane
|
1 月之前 |
|
08-models_deepseek_v4_1_compressor.py.diff
|
0d8166de95
port DCP2 onto our tree (22-file superset wrapper + a2a recipe); A/B: 162/176, 5.34M-token KV @512K, -5 vs b12x — the high-KV lane
|
1 月之前 |
|
09-models_deepseek_v4_1_nvidia_flashinfer_sparse.py.diff
|
0d8166de95
port DCP2 onto our tree (22-file superset wrapper + a2a recipe); A/B: 162/176, 5.34M-token KV @512K, -5 vs b12x — the high-KV lane
|
1 月之前 |
|
10-models_deepseek_v4_1_nvidia_model_state.py.diff
|
0d8166de95
port DCP2 onto our tree (22-file superset wrapper + a2a recipe); A/B: 162/176, 5.34M-token KV @512K, -5 vs b12x — the high-KV lane
|
1 月之前 |
|
11-models_deepseek_v4_1_sparse_mla.py.diff
|
0d8166de95
port DCP2 onto our tree (22-file superset wrapper + a2a recipe); A/B: 162/176, 5.34M-token KV @512K, -5 vs b12x — the high-KV lane
|
1 月之前 |
|
12-v1_attention_backend.py.diff
|
0d8166de95
port DCP2 onto our tree (22-file superset wrapper + a2a recipe); A/B: 162/176, 5.34M-token KV @512K, -5 vs b12x — the high-KV lane
|
1 月之前 |
|
13-v1_attention_backends_mla_compressor_utils.py.diff
|
0d8166de95
port DCP2 onto our tree (22-file superset wrapper + a2a recipe); A/B: 162/176, 5.34M-token KV @512K, -5 vs b12x — the high-KV lane
|
1 月之前 |
|
14-v1_attention_backends_mla_indexer.py.diff
|
0d8166de95
port DCP2 onto our tree (22-file superset wrapper + a2a recipe); A/B: 162/176, 5.34M-token KV @512K, -5 vs b12x — the high-KV lane
|
1 月之前 |
|
15-v1_attention_backends_mla_sparse_swa.py.diff
|
0d8166de95
port DCP2 onto our tree (22-file superset wrapper + a2a recipe); A/B: 162/176, 5.34M-token KV @512K, -5 vs b12x — the high-KV lane
|
1 月之前 |
|
16-v1_core_kv_cache_coordinator.py.diff
|
0d8166de95
port DCP2 onto our tree (22-file superset wrapper + a2a recipe); A/B: 162/176, 5.34M-token KV @512K, -5 vs b12x — the high-KV lane
|
1 月之前 |
|
17-v1_core_kv_cache_utils.py.diff
|
0d8166de95
port DCP2 onto our tree (22-file superset wrapper + a2a recipe); A/B: 162/176, 5.34M-token KV @512K, -5 vs b12x — the high-KV lane
|
1 月之前 |
|
18-v1_kv_cache_interface.py.diff
|
0d8166de95
port DCP2 onto our tree (22-file superset wrapper + a2a recipe); A/B: 162/176, 5.34M-token KV @512K, -5 vs b12x — the high-KV lane
|
1 月之前 |
|
19-v1_worker_cp_utils.py.diff
|
0d8166de95
port DCP2 onto our tree (22-file superset wrapper + a2a recipe); A/B: 162/176, 5.34M-token KV @512K, -5 vs b12x — the high-KV lane
|
1 月之前 |
|
20-v1_worker_gpu_attn_utils.py.diff
|
0d8166de95
port DCP2 onto our tree (22-file superset wrapper + a2a recipe); A/B: 162/176, 5.34M-token KV @512K, -5 vs b12x — the high-KV lane
|
1 月之前 |
|
21-v1_worker_gpu_block_table.py.diff
|
0d8166de95
port DCP2 onto our tree (22-file superset wrapper + a2a recipe); A/B: 162/176, 5.34M-token KV @512K, -5 vs b12x — the high-KV lane
|
1 月之前 |
|
22-v1_worker_gpu_model_runner.py.diff
|
0d8166de95
port DCP2 onto our tree (22-file superset wrapper + a2a recipe); A/B: 162/176, 5.34M-token KV @512K, -5 vs b12x — the high-KV lane
|
1 月之前 |
|
23-v1_worker_gpu_spec_decode_dflash_speculator.py.diff
|
0d8166de95
port DCP2 onto our tree (22-file superset wrapper + a2a recipe); A/B: 162/176, 5.34M-token KV @512K, -5 vs b12x — the high-KV lane
|
1 月之前 |
