tracker: generate
compatibility with torch.compile
#28981
Labels
WIP
Label your PR/Issue with WIP for some long outstanding Issues/PRs that are work in progress
generate
馃 馃torch.compile
This issue is a tracker of the compatibility between
.generate
andtorch.compile
(intro docs by pytorch). The goal is to enablefullgraph=True
compilation on the maingenerate
use cases.generate
use case not covered by this tracker? Check if it was requested below and upvote it if it was. Otherwise, add a comment. We will consider expanding the selection below on widely requested use cases 馃Decoding Strategies
greedy_search
/sample
are compatible (Generate: end-to-end compilation聽#30788)beam_search
/beam_sample
are compatible, depends on the step aboveassisted_decoding
(aka speculative decoding) is compatible, depends on the steps aboveGenerate Flags and Options
LogitsProcessor
classes were checked for compatibility (and the appropriate exceptions are raised when not compatible)StoppingCriteria
classes were checked for compatibility (and the appropriate exceptions are raised when not compatible)Models
(models tagged as "important models" in our CI + popular models)
Decoder-only:
Core generation
] Adds support for static KV cache聽#27931)gemma
] Adds support for Gemma 馃拵聽#29167)torch.compile
implementation聽#29891)Encoder-decoder:
Quantization
Others
The text was updated successfully, but these errors were encountered: