6 Commits

Author	SHA1	Message	Date
root	d0585b4a17	fix perforence	2026-07-14 13:11:21 +00:00
Lu Xinlong	c84151eef9	fix issues	2026-06-26 12:55:02 +08:00
Lu Xinlong	c2de1c83b0	Utilize chunked prefill + K-tiling techniques to ensure 100K context	2026-06-05 17:00:41 +08:00
Lu Xinlong	2d1ef50992	chunked prefill support and memory opts	2026-06-05 16:03:34 +08:00
Lu Xinlong	8c047a70ea	some modifications to ensure 50K context input	2026-06-04 17:56:29 +08:00
Lu Xinlong	1c33ef1355	add paged_attn	2026-05-29 16:53:39 +08:00