b1tg 0fbc551622 train bert with fp8 (#13874) ... * fp8 train * clean * lint * test fix from #13439 * skip first/last layer * rm __init__, restore unroll <=32 check * tests * clean test, remove unused * multi-gpu test, clean quantize_to_fp8 * remove bert contiguous * run script * test: better check * run script search * add seed in bert data shuffle * move script to mi350x folder --------- Co-authored-by: chenyu <chenyu@fastmail.com>		2026-01-09 09:21:59 -05:00
..
__init__.py	fix test extra issue (#3159)	2024-01-17 11:58:08 -08:00
test_bench_log.py	break swizzle into three chunks [pr] (#11153)	2025-07-09 15:30:34 -07:00
test_cfg_viz.py	visualize SQTT with the same cfg infrastructure (#13870)	2026-01-06 14:53:20 +09:00
test_export_model.py	Use matching JS TypedArray for buffer dtype (#8080)	2024-12-06 14:52:23 +01:00
test_f16_decompress.py	only run test_u32_to_f16 if it supports fp16 (#10277)	2025-05-13 11:16:14 -04:00
test_fp8_linear.py	train bert with fp8 (#13874)	2026-01-09 09:21:59 -05:00
test_hevc.py	test_hevc: do not download the whole file (#13531)	2025-12-02 21:31:28 +03:00
test_lr_scheduler.py	fix some slow tests [pr] (#8979)	2025-02-09 15:57:04 +08:00
test_mockgpu.py	move mockgpu to tests [pr] (#8396)	2024-12-24 23:48:02 +03:00
test_tk.py	tk: fix grouped load store (#14035)	2026-01-07 22:38:02 -08:00