Add 8bit quantized model

config.json

@@ -1,5 +1,5 @@
 {
-  "_name_or_path": "beomi/KoAlpaca-Polyglot-5.8B",
+  "_name_or_path": "./polyglot-5.8b-koalpaca-v1.1b",
   "architectures": [
     "GPTNeoXForCausalLM"
   ],
@@ -15,6 +15,12 @@
   "num_attention_heads": 16,
   "num_hidden_layers": 28,
   "num_steps": "global_step320000",
+  "quantization_config": {
+    "llm_int8_enable_fp32_cpu_offload": false,
+    "llm_int8_skip_modules": null,
+    "llm_int8_threshold": 6.0,
+    "load_in_8bit": true
+  },
   "rotary_emb_base": 10000,
   "rotary_pct": 0.25,
   "tie_word_embeddings": false,

pytorch_model-00001-of-00002.bin

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:b088704796dab521146f1309a6e6eb2e936cc255f01f2101fbf36dee9f94be45
-size 10017673915

pytorch_model-00002-of-00002.bin → pytorch_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:5f8bfb30e258bc8ceb43b9e8f5feafcc8c00b802df7a3120448794e6d8e963cf
-size 1870040575
+oid sha256:0e9e227fdfe4b23600ef74a1420248ee839de65ab25a058b7d4432859e55923c
+size 6254727565