update quant_mllm shell (#2959)

Author: Jintao-Huang

examples/export/quantize/gptq.sh

@@ -1,4 +1,4 @@
-# OMP_NUM_THREADS=14 please Check issue:https://github.com/AutoGPTQ/AutoGPTQ/issues/439
+# OMP_NUM_THREADS=14 please Check issue: https://github.com/AutoGPTQ/AutoGPTQ/issues/439
 OMP_NUM_THREADS=14 \
 CUDA_VISIBLE_DEVICES=0 \
 swift export \

examples/export/quantize/mllm/awq.sh

@@ -0,0 +1,11 @@
+CUDA_VISIBLE_DEVICES=0 \
+swift export \
+    --model Qwen/Qwen2-VL-2B-Instruct \
+    --dataset 'AI-ModelScope/alpaca-gpt4-data-zh#500' \
+              'AI-ModelScope/alpaca-gpt4-data-en#500' \
+    --quant_n_samples 256 \
+    --quant_batch_size -1 \
+    --max_length 2048 \
+    --quant_method awq \
+    --quant_bits 4 \
+    --output_dir Qwen/Qwen2-VL-2B-Instruct-AWQ

examples/export/quantize/mllm/gptq.sh

@@ -0,0 +1,18 @@
+# OMP_NUM_THREADS=14 please Check issue: https://github.com/AutoGPTQ/AutoGPTQ/issues/439
+OMP_NUM_THREADS=14 \
+CUDA_VISIBLE_DEVICES=0 \
+MAX_PIXELS=1003520 \
+VIDEO_MAX_PIXELS=50176 \
+FPS_MAX_FRAMES=12 \
+swift export \
+    --model Qwen/Qwen2-VL-2B-Instruct \
+    --dataset 'AI-ModelScope/alpaca-gpt4-data-zh#500' \
+              'AI-ModelScope/alpaca-gpt4-data-en#500' \
+              'modelscope/coco_2014_caption:validation#500' \
+              'swift/VideoChatGPT:Generic#500' \
+    --quant_n_samples 256 \
+    --quant_batch_size 1 \
+    --max_length 2048 \
+    --quant_method gptq \
+    --quant_bits 4 \
+    --output_dir Qwen/Qwen2-VL-2B-Instruct-GPTQ-Int4

swift/llm/argument/export_args.py

@@ -75,6 +75,8 @@ def _init_output_dir(self):
         assert not os.path.exists(self.output_dir), f'args.output_dir: {self.output_dir} already exists.'
 
     def __post_init__(self):
+        if self.quant_batch_size == -1:
+            self.quant_batch_size = None
         if self.quant_bits and self.quant_method is None:
             raise ValueError('Please specify the quantization method using `--quant_method awq/gptq/bnb`.')
         if self.quant_method and self.quant_bits is None: