Update config.json

config.json

@@ -20,27 +20,27 @@
   "pad_token_id": 128255,
   "pretraining_tp": 1,
   "quantization_config": {
-    "_load_in_4bit": false,
-    "_load_in_8bit": true,
-    "bnb_4bit_compute_dtype": "float32",  // Changed to float32 for CPU compatibility
+    "_load_in_4bit": true,
+    "_load_in_8bit": false,
+    "bnb_4bit_compute_dtype": "bfloat16",
     "bnb_4bit_quant_storage": "uint8",
     "bnb_4bit_quant_type": "nf4",
-    "bnb_4bit_use_double_quant": false,  // Changed to false
-    "llm_int8_enable_fp32_cpu_offload": true,  // Enabled for optimized CPU offload
+    "bnb_4bit_use_double_quant": true,
+    "llm_int8_enable_fp32_cpu_offload": false,
     "llm_int8_has_fp16_weight": false,
     "llm_int8_skip_modules": null,
     "llm_int8_threshold": 6.0,
-    "load_in_4bit": false,  // Disabled for CPU
-    "load_in_8bit": true,  // Disabled for CPU
+    "load_in_4bit": true,
+    "load_in_8bit": false,
     "quant_method": "bitsandbytes"
   },
   "rms_norm_eps": 1e-05,
   "rope_scaling": null,
   "rope_theta": 500000.0,
   "tie_word_embeddings": false,
-  "torch_dtype": "float32",  // Changed to float32 for CPU compatibility
+  "torch_dtype": "bfloat16",
   "transformers_version": "4.44.2",
   "unsloth_version": "2024.9",
   "use_cache": true,
   "vocab_size": 128256
-}
+}