wyseow
/

InternVL-Chat-V1-5-Int8-OL

Visual Question Answering

feature-extraction

4-bit precision

Model card Files Files and versions Community

wyseow commited on May 27

Commit

be280a1

•

1 Parent(s): 650741b

change to 4-bit quan config

Files changed (1) hide show

config.json +5 -5

config.json CHANGED Viewed

@@ -106,15 +106,15 @@
   "pad2square": false,
   "ps_version": "v2",
   "quantization_config": {
-    "bnb_4bit_compute_dtype": "float32",
-    "bnb_4bit_quant_type": "fp4",
-    "bnb_4bit_use_double_quant": false,
     "llm_int8_enable_fp32_cpu_offload": true,
     "llm_int8_has_fp16_weight": false,
     "llm_int8_skip_modules": null,
     "llm_int8_threshold": 6.0,
-    "load_in_4bit": false,
-    "load_in_8bit": true,
     "quant_method": "bitsandbytes"
   },
   "select_layer": -1,

   "pad2square": false,
   "ps_version": "v2",
   "quantization_config": {
+    "bnb_4bit_compute_dtype": "bfloat16",
+    "bnb_4bit_quant_type": "nf4",
+    "bnb_4bit_use_double_quant": true,
     "llm_int8_enable_fp32_cpu_offload": true,
     "llm_int8_has_fp16_weight": false,
     "llm_int8_skip_modules": null,
     "llm_int8_threshold": 6.0,
+    "load_in_4bit": true,
+    "load_in_8bit": false,
     "quant_method": "bitsandbytes"
   },
   "select_layer": -1,