Update README.md (#11)
Browse files- Update README.md (530aa354f48dd13bd6e5795e1cd98ee7516cb9b9)
Co-authored-by: haoyanli <[email protected]>
README.md
CHANGED
|
@@ -24,16 +24,18 @@ python3 quantize_quark.py \
|
|
| 24 |
--output_dir c4ai-command-r-plus-FP8-KV \
|
| 25 |
--quant_scheme w_fp8_a_fp8 \
|
| 26 |
--kv_cache_dtype fp8 \
|
| 27 |
-
--num_calib_data 128
|
| 28 |
-
--model_export quark_safetensors
|
|
|
|
| 29 |
# If model size is too large for single GPU, please use multi GPU instead.
|
| 30 |
python3 quantize_quark.py \
|
| 31 |
--model_dir $MODEL_DIR \
|
| 32 |
--output_dir c4ai-command-r-plus-FP8-KV \
|
| 33 |
--quant_scheme w_fp8_a_fp8 \
|
| 34 |
--kv_cache_dtype fp8 \
|
| 35 |
-
--num_calib_data 128
|
| 36 |
--model_export quark_safetensors \
|
|
|
|
| 37 |
--multi_gpu
|
| 38 |
```
|
| 39 |
## Deployment
|
|
|
|
| 24 |
--output_dir c4ai-command-r-plus-FP8-KV \
|
| 25 |
--quant_scheme w_fp8_a_fp8 \
|
| 26 |
--kv_cache_dtype fp8 \
|
| 27 |
+
--num_calib_data 128 \
|
| 28 |
+
--model_export quark_safetensors \
|
| 29 |
+
--no_weight_matrix_merge
|
| 30 |
# If model size is too large for single GPU, please use multi GPU instead.
|
| 31 |
python3 quantize_quark.py \
|
| 32 |
--model_dir $MODEL_DIR \
|
| 33 |
--output_dir c4ai-command-r-plus-FP8-KV \
|
| 34 |
--quant_scheme w_fp8_a_fp8 \
|
| 35 |
--kv_cache_dtype fp8 \
|
| 36 |
+
--num_calib_data 128 \
|
| 37 |
--model_export quark_safetensors \
|
| 38 |
+
--no_weight_matrix_merge \
|
| 39 |
--multi_gpu
|
| 40 |
```
|
| 41 |
## Deployment
|