Upload ms-swift/examples/export/quantize/reward_model/gptq.sh with huggingface_hub
Browse files
ms-swift/examples/export/quantize/reward_model/gptq.sh
ADDED
|
@@ -0,0 +1,13 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# gptq quantize
|
| 2 |
+
CUDA_VISIBLE_DEVICES=0 swift export \
|
| 3 |
+
--model Shanghai_AI_Laboratory/internlm2-1_8b-reward \
|
| 4 |
+
--output_dir output/internlm2-1_8b-reward-gptq-int4 \
|
| 5 |
+
--quant_bits 4 \
|
| 6 |
+
--quant_method gptq \
|
| 7 |
+
--dataset 'AI-ModelScope/alpaca-gpt4-data-zh#1000' 'AI-ModelScope/alpaca-gpt4-data-en#1000'
|
| 8 |
+
|
| 9 |
+
# infer
|
| 10 |
+
CUDA_VISIBLE_DEVICES=0 swift infer \
|
| 11 |
+
--model output/internlm2-1_8b-reward-gptq-int4 \
|
| 12 |
+
--val_dataset 'AI-ModelScope/alpaca-gpt4-data-zh#1000' \
|
| 13 |
+
--max_batch_size 16
|