Add adapter_quant field to LoraConfig#20503
Conversation
Summary: Split out from D105388759. Adds a per-adapter quantization/precision field `adapter_quant` to `LoraConfig` so each LoRA adapter can independently select `int8`/`fp16`/`fp32`, overriding the global `--lora_precision` flag for that adapter only. This is a standalone, behavior-preserving dataclass field addition (defaults to `None`) that the per-adapter quant logic in `export_llm_backbone.py` builds on. Landed separately so the config schema change can be reviewed and merged independently. Changes are mirrored in both the `fbcode` and `xplat` copies of `llm_config.py`. This diff was authored with Claude Code. Differential Revision: D109642451
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/20503
Note: Links to docs will display an error until the docs builds have been completed. ❗ 2 Active SEVsThere are 2 currently active SEVs. If your PR is affected, please view them below:
❌ 1 New Failure, 2 Unrelated FailuresAs of commit 86e9963 with merge base 45a14b9 ( NEW FAILURE - The following job has failed:
BROKEN TRUNK - The following jobs failed but were present on the merge base:👉 Rebase onto the `viable/strict` branch to avoid these failures
This comment was automatically generated by Dr. CI and updates every 15 minutes. |
meta-cla
Bot
added
the
CLA Signed
|
|
@billmguo has exported this pull request. If you are a Meta employee, you can view the originating Diff in D109642451. |
This PR needs a
|
1 participant
Summary:
Split out from D105388759. Adds a per-adapter quantization/precision field
adapter_quanttoLoraConfigso each LoRA adapter can independently selectint8/fp16/fp32, overriding the global--lora_precisionflag for that adapter only.This is a standalone, behavior-preserving dataclass field addition (defaults to
None) that the per-adapter quant logic inexport_llm_backbone.pybuilds on. Landed separately so the config schema change can be reviewed and merged independently. Changes are mirrored in both thefbcodeandxplatcopies ofllm_config.py.This diff was authored with Claude Code.
Differential Revision: D109642451