Update configuration.py

Author: nikita-malininn

optimum/intel/openvino/configuration.py

@@ -33,25 +33,30 @@
 logger = logging.getLogger(__name__)
 
 _DEFAULT_4BIT_CONFIGS = {
-    "databricks/dolly-v2-3b": {"bits": 4, "sym": False, "group_size": 32, "ratio": 0.5},
+    "databricks/dolly-v2-3b": {"bits": 4, "sym": False, "group_size": 128, "ratio": 0.8},
     "EleutherAI/gpt-j-6b": {"bits": 4, "sym": False, "group_size": 64},
     "facebook/opt-6.7b": {"bits": 4, "sym": False, "group_size": 64, "ratio": 0.8},
     "bigscience/bloomz-7b1": {"bits": 4, "sym": False, "group_size": 32, "ratio": 0.6},
     "togethercomputer/RedPajama-INCITE-7B-Instruct": {"bits": 4, "sym": False, "group_size": 128},
     "HuggingFaceH4/zephyr-7b-beta": {"bits": 4, "sym": True, "group_size": 64, "ratio": 0.6},
-    "meta-llama/Llama-2-7b": {"bits": 4, "sym": True, "group_size": 128, "ratio": 0.6},
+    "meta-llama/Llama-2-7b": {"bits": 4, "sym": True, "group_size": 128, "ratio": 0.7},
     "meta-llama/Llama-2-7b-chat": {"bits": 4, "sym": True, "group_size": 128, "ratio": 0.8},
     "meta-llama/Llama-2-13b-chat": {"bits": 4, "sym": True, "group_size": 64, "ratio": 0.8},
     "stabilityai/stablelm-3b-4e1t": {"bits": 4, "sym": True, "group_size": 64, "ratio": 0.8},
     "stablelm-epoch-3b-preview": {"bits": 4, "sym": True, "group_size": 64, "ratio": 0.8},
-    "stable-zephyr-3b-dpo": {"bits": 4, "sym": False, "group_size": 64, "ratio": 0.8},
+    "stabilityai/stable-zephyr-3b-dpo": {"bits": 4, "sym": False, "group_size": 128, "ratio": 1.0},
     "pansophic/rocket-3B": {"bits": 4, "sym": True, "group_size": 128, "ratio": 0.8},
     "THUDM/chatglm2-6b": {"bits": 4, "sym": True, "group_size": 128, "ratio": 0.72},
     "Qwen/Qwen-7B-Chat": {"bits": 4, "sym": True, "group_size": 128, "ratio": 0.6},
     "openlm-research/open_llama_3b": {"bits": 4, "sym": True, "group_size": 64, "all_layers": True},
     "tiiuae/falcon-7b": {"bits": 4, "sym": True, "group_size": 64, "all_layers": True},
     "psmathur/orca_mini_3b": {"bits": 4, "sym": True, "group_size": 64, "all_layers": True},
     "mistralai/Mixtral-8x7B-v0.1": {"bits": 4, "sym": True, "group_size": 128, "ratio": 0.8},
+    "facebook/opt-2.7b": {"bits": 4, "sym": True, "group_size": 128, "ratio": 0.7},
+    "togethercomputer/RedPajama-INCITE-Chat-3B-v1": {"bits": 4, "sym": False, "group_size": 128, "ratio": 0.8},
+    "lmsys/vicuna-7b-v1.5": {"bits": 4, "sym": False, "group_size": 128, "ratio": 1.0},
+    "stabilityai/stablelm-tuned-alpha-3b": {"bits": 4, "sym": False, "group_size": 128, "ratio": 0.8},
+    "mistralai/Mistral-7B-v0.1": {"bits": 4, "sym": True, "group_size": 128, "ratio": 0.9},
 }