improve SQ mpt

Signed-off-by: changwangss <[email protected]>

intel_extension_for_transformers/transformers/modeling/modeling_auto.py

@@ -840,6 +840,14 @@ def forward(self, input: torch.Tensor) -> tuple[torch.Tensor, None]:
                 or device_map == torch.device("cpu")
             ) and model.config.model_type == "chatglm":
                 model = model.float()
+            if (
+                not torch.cuda.is_available()
+                or device_map == "cpu"
+                or device_map == torch.device("cpu")
+            ) and model.config.model_type == "mpt":
+                config = AutoConfig.from_pretrained("mosaicml_mpt-7b_config.json",
+                                                    torchscript=True)
+                model.config = config
             model.eval()
             model_type = model.config.model_type.replace("_", "-")
 
@@ -1077,6 +1085,7 @@ def calib_func(model):
                 recipes=quantization_config.recipes,
                 example_inputs=example_inputs,
             )
+
             model = quantization.fit(
                 model,
                 conf,

intel_extension_for_transformers/transformers/modeling/mosaicml_mpt-7b_config.json

@@ -0,0 +1,48 @@
+{
+    "architectures": [
+        "MptForCausalLM"
+    ],
+    "attn_config": {
+        "alibi": true,
+        "alibi_bias_max": 8,
+        "attn_impl": "torch",
+        "attn_pdrop": 0,
+        "attn_type": "multihead_attention",
+        "attn_uses_sequence_id": false,
+        "clip_qkv": null,
+        "prefix_lm": false,
+        "qk_ln": false,
+        "softmax_scale": null
+    },
+    "d_model": 4096,
+    "emb_pdrop": 0,
+    "embedding_fraction": 1.0,
+    "expansion_ratio": 4,
+    "init_config": {
+        "emb_init_std": null,
+        "emb_init_uniform_lim": null,
+        "fan_mode": "fan_in",
+        "init_div_is_residual": true,
+        "init_gain": 0,
+        "init_nonlinearity": "relu",
+        "init_std": 0.02,
+        "name": "kaiming_normal_",
+        "verbose": 0
+    },
+    "init_device": "cpu",
+    "learned_pos_emb": true,
+    "logit_scale": null,
+    "max_seq_len": 2048,
+    "model_type": "mpt",
+    "n_heads": 32,
+    "n_layers": 32,
+    "no_bias": true,
+    "norm_type": "low_precision_layernorm",
+    "resid_pdrop": 0,
+    "tokenizer_name": "EleutherAI/gpt-neox-20b",
+    "torch_dtype": "bfloat16",
+    "transformers_version": "4.28.1",
+    "use_cache": false,
+    "verbose": 0,
+    "vocab_size": 50432
+}