Skip to content

PTQ-Level Efficiency, QAT-Level Accuracy | 训练后量化的效率,量化感知训练的精度

fix bug of backend (#294)

wenhuach214f228717ad02
diff --git a/README.md b/README.md
index ca6abfa34..a7597134f 100644
--- a/README.md
+++ b/README.md
@@ -57,14 +57,14 @@ pip install auto-round
### Basic Usage (Gaudi2/CPU/GPU)
A user guide detailing the full list of supported arguments is provided by calling ```auto-round -h``` on the terminal.
-Alternatively, you can use ```auto_round``` instead of ```auto-round```.
+Alternatively, you can use ```auto_round``` instead of ```auto-round```. Set the format you want in `format` and multiple formats exporting has been supported.
```bash
CUDA_VISIBLE_DEVICES=0 auto-round \
--model facebook/opt-125m \
--bits 4 \
--group_size 128 \
- --format auto_round \
+ --format "auto_round,auto_gptq" \
--disable_eval \
--output_dir ./tmp_autoround
```
diff --git a/auto_round/backend.py b/auto_round/backend.py
index ea79ac850..6afbf1ef0 100644
--- a/auto_round/backend.py
+++ b/auto_round/backend.py
@@ -137,7 +137,7 @@ BackendInfos['auto_round:qbits'] = BackendInfo(device=["cpu"], sym=[True, False]
requirements=["intel-extension-for-transformers"])
BackendInfos['auto_round:qbits_zp'] = BackendInfo(device=["cpu"], sym=[True, False],
- packing_format="qbits",
+ packing_format="qbits_zp+-1",
bits=[2, 4, 8], group_size=None,
priority=0 if "intel" in get_cpu_manufacturer() else 5,
feature_checks=[],