PR #7887 [SP] add SP deny list instead of allow

add SP deny list instead of allow

kashif committed 159 days ago

hub-kernel lazy registration before validation and tests

kashif committed 156 days ago

Merge branch 'master' into sp_attn_deny

kashif committed 156 days ago

Merge branch 'master' into sp_attn_deny

tohtana committed 154 days ago

Merge branch 'master' into sp_attn_deny

kashif committed 151 days ago

Merge branch 'master' into sp_attn_deny

kashif committed 150 days ago

position_ids generation and flex_attention BlockMask

kashif committed 150 days ago

refactor

kashif committed 150 days ago

update comments

kashif committed 150 days ago

do not check for has_packed_samples

kashif committed 148 days ago

raise error instead of warning

kashif committed 147 days ago

cache BlockMask

kashif committed 147 days ago

Merge branch 'master' into sp_attn_deny

loadams committed 141 days ago

Merge branch 'master' into sp_attn_deny

sfc-gh-truwase committed 137 days ago

flex_attention: lazy imports to fix pickle issue and add test

kashif committed 137 days ago

fix flex_attention BlockMask guard and add position_ids warning in forward()

kashif committed 137 days ago

flex_attention test: use local tiny model, fix _compile gradient issue

kashif committed 137 days ago

don't compile create_block_mask inside forward()

kashif committed 137 days ago

assert position_ids in forward() instead of warning

kashif committed 136 days ago

Merge branch 'master' into sp_attn_deny

sfc-gh-truwase committed 135 days ago

Merge branch 'master' into sp_attn_deny

stas00 committed 133 days ago

Merge branch 'master' into sp_attn_deny

stas00 committed 132 days ago

DeepSpeed [SP] add SP deny list instead of allow #7887 Merged