use 8-bit weights

config.json

@@ -1,5 +1,5 @@
 {
-  "_name_or_path": "bart-large-cnn",
+  "_name_or_path": "booksouls/bart-large-cnn",
   "_num_labels": 3,
   "activation_dropout": 0.0,
   "activation_function": "gelu",
@@ -53,9 +53,9 @@
   "pad_token_id": 1,
   "prefix": " ",
   "quantization_config": {
-    "_load_in_4bit": true,
-    "_load_in_8bit": false,
-    "bnb_4bit_compute_dtype": "bfloat16",
+    "_load_in_4bit": false,
+    "_load_in_8bit": true,
+    "bnb_4bit_compute_dtype": "float32",
     "bnb_4bit_quant_storage": "uint8",
     "bnb_4bit_quant_type": "fp4",
     "bnb_4bit_use_double_quant": false,
@@ -63,8 +63,8 @@
     "llm_int8_has_fp16_weight": false,
     "llm_int8_skip_modules": null,
     "llm_int8_threshold": 6.0,
-    "load_in_4bit": true,
-    "load_in_8bit": false,
+    "load_in_4bit": false,
+    "load_in_8bit": true,
     "quant_method": "bitsandbytes"
   },
   "scale_embedding": false,
@@ -79,7 +79,7 @@
     }
   },
   "torch_dtype": "float16",
-  "transformers_version": "4.41.2",
+  "transformers_version": "4.40.2",
   "use_cache": true,
   "vocab_size": 50264
 }

generation_config.json

@@ -12,5 +12,5 @@
   "no_repeat_ngram_size": 3,
   "num_beams": 4,
   "pad_token_id": 1,
-  "transformers_version": "4.41.2"
+  "transformers_version": "4.40.2"
 }

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:dadcc379c74fc48b0b2395f42794326901c92f98902ced7ebb9c7d369d486db5
-size 306379320
+oid sha256:b840ef94392ee418f91f8bb77983b81f1c55bb9a0599f856bb5416bd5002c222
+size 461542680