add quan config like InternVL-Chat-V1-5-Int8-OL
Browse files- config.json +7 -0
config.json
CHANGED
|
@@ -1,4 +1,11 @@
|
|
| 1 |
{
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 2 |
"_commit_hash": null,
|
| 3 |
"_name_or_path": "OpenGVLab/InternVL-Chat-V1-5",
|
| 4 |
"architectures": [
|
|
|
|
| 1 |
{
|
| 2 |
+
"quantization_config": {
|
| 3 |
+
"bnb_4bit_compute_dtype": "bfloat16",
|
| 4 |
+
"bnb_4bit_quant_type": "nf4",
|
| 5 |
+
"bnb_4bit_use_double_quant": true,
|
| 6 |
+
"load_in_4bit": true,
|
| 7 |
+
"quant_method": "bitsandbytes"
|
| 8 |
+
},
|
| 9 |
"_commit_hash": null,
|
| 10 |
"_name_or_path": "OpenGVLab/InternVL-Chat-V1-5",
|
| 11 |
"architectures": [
|