[FP8][Kernel] Dynamic kv cache scaling factors computation (#11906)

Signed-off-by: Gregory Shtrasberg <Gregory.Shtrasberg@amd.com>
Co-authored-by: Micah Williamson <micah.williamson@amd.com>
This commit is contained in:
Gregory Shtrasberg
2025-01-23 18:04:03 +00:00
committed by GitHub
co-authored by Micah Williamson
parent 6e650f56a1
commit e97f802b2d
60 changed files with 276 additions and 1365 deletions
+5
View File
@@ -190,6 +190,7 @@ class TPUModelRunner(ModelRunnerBase[ModelInputForTPU]):
num_decode_tokens=0,
slot_mapping=slot_mapping,
multi_modal_placeholder_index_maps=None,
enable_kv_scales_calculation=False,
block_tables=None,
context_lens=None,
effective_query_lens=None,
@@ -208,6 +209,7 @@ class TPUModelRunner(ModelRunnerBase[ModelInputForTPU]):
num_decode_tokens=0,
slot_mapping=slot_mapping,
multi_modal_placeholder_index_maps=None,
enable_kv_scales_calculation=False,
block_tables=block_tables,
context_lens=context_lens,
effective_query_lens=effective_query_lens,
@@ -239,6 +241,7 @@ class TPUModelRunner(ModelRunnerBase[ModelInputForTPU]):
num_decode_tokens=batch_size * seq_len,
slot_mapping=slot_mapping,
multi_modal_placeholder_index_maps=None,
enable_kv_scales_calculation=False,
block_tables=block_tables,
context_lens=context_lens,
)
@@ -425,6 +428,7 @@ class TPUModelRunner(ModelRunnerBase[ModelInputForTPU]):
num_decode_tokens=0,
slot_mapping=slot_mapping,
multi_modal_placeholder_index_maps=None,
enable_kv_scales_calculation=False,
block_tables=block_tables,
context_lens=context_lens,
effective_query_lens=prompt_lens,
@@ -496,6 +500,7 @@ class TPUModelRunner(ModelRunnerBase[ModelInputForTPU]):
num_decode_tokens=batch_size,
slot_mapping=slot_mapping,
multi_modal_placeholder_index_maps=None,
enable_kv_scales_calculation=False,
block_tables=block_tables,
context_lens=context_lens,
)