Add minicpm-2b in L0 pipeline (#12308)

python/llm/example/NPU/HF-Transformers-AutoModels/LLM/Pipeline-Models/README.md

@@ -11,7 +11,7 @@ In this directory, you will find examples on how to directly run HuggingFace `tr
 | Qwen2 | [Qwen/Qwen2-1.5B-Instruct](https://huggingface.co/Qwen/Qwen2-1.5B-Instruct) |
 | Qwen2.5 | [Qwen/Qwen2.5-7b-Instruct](https://huggingface.co/Qwen/Qwen2.5-7b-Instruct) |
 | Baichuan2 | [baichuan-inc/Baichuan2-7B-Chat](https://huggingface.co/baichuan-inc/Baichuan-7B-Chat) |
-| MiniCPM | [openbmb/MiniCPM-1B-sft-bf16](https://huggingface.co/openbmb/MiniCPM-1B-sft-bf16) |
+| MiniCPM | [openbmb/MiniCPM-1B-sft-bf16](https://huggingface.co/openbmb/MiniCPM-1B-sft-bf16), [openbmb/MiniCPM-2B-sft-bf16](https://huggingface.co/openbmb/MiniCPM-2B-sft-bf16) |
 
 ## 0. Requirements
 To run these examples with IPEX-LLM on Intel NPUs, make sure to install the newest driver version of Intel NPU.
@@ -59,6 +59,9 @@ python baichuan2.py
 
 :: to run MiniCPM-1B-sft-bf16
 python minicpm.py
+
+:: to run MiniCPM-2B-sft-bf16
+python minicpm.py --repo-id-or-model-path "openbmb/MiniCPM-2B-sft-bf16"
 ```
 
 Arguments info:

python/llm/example/NPU/HF-Transformers-AutoModels/LLM/Pipeline-Models/minicpm.py

@@ -32,7 +32,7 @@ if __name__ == "__main__":
     parser.add_argument(
         "--repo-id-or-model-path",
         type=str,
-        default="openbmb/MiniCPM-1B-sft-bf16",
+        default="openbmb/MiniCPM-1B-sft-bf16", # or "openbmb/MiniCPM-2B-sft-bf16"
         help="The huggingface repo id for the MiniCPM model to be downloaded"
         ", or the path to the huggingface checkpoint folder",
     )

python/llm/src/ipex_llm/transformers/npu_pipeline_model/convert_pipeline.py

@@ -362,7 +362,7 @@ def convert_llm(model: torch.nn.Module,
         invalidInputError(False, "Now we only support Llama2 / Llama3 / Baichuan2 / "
                                  "Qwen2 / Qwen2.5 / Minicpm for pipeline running.")
 
-    if isinstance(model.lm_head, SlicedLMHead):
+    if hasattr(model, "lm_head") and isinstance(model.lm_head, SlicedLMHead):
         model.lm_head.get_fused_lm_head()
 
     # patch generate function

python/llm/src/ipex_llm/transformers/npu_pipeline_model/minicpm.py

@@ -18,8 +18,10 @@
 import torch
 import numpy as np
 import os
-from .common import update_names_of_IR_and_export_blob, LowBitLLMLMHead
+from .common import update_names_of_IR_and_export_blob
 from intel_npu_acceleration_library.backend.factory import NNFactory
+from ipex_llm.transformers.npu_models.mp_models_base import LLMBaseNNFactory
+from typing import Sequence
 
 
 class MiniCPMEmbedding(NNFactory):
@@ -65,6 +67,68 @@ class MiniCPMEmbedding(NNFactory):
         self.compile()
 
 
+class MiniCPMLMHead(LLMBaseNNFactory):
+    def __init__(
+        self,
+        hidden_shape: Sequence[int],
+        num_heads: int,
+        rms_norm_eps: float,
+        model_norm_weight,
+        vocab_size: int,
+        mode: str = "decode",
+        dtype: np.dtype = np.int8,
+        max_seq_len: int = 1024,
+        transpose_value: bool = False,
+        profile: bool = False,
+        device: str = "NPU",
+    ):
+        super().__init__(max_seq_len=max_seq_len,
+                         transpose_value=transpose_value,
+                         dtype=dtype,
+                         profile=profile,
+                         device=device)
+        self.max_seq_len = max_seq_len
+        self.dtype = dtype
+        self.batch_size, self.seq_len, self.hidden_size = hidden_shape
+        self.mode = mode
+        self.rms_norm_eps = rms_norm_eps
+        self.transpose_value = transpose_value
+        self.vocab_size = vocab_size
+
+        self.num_heads = num_heads
+        self.head_dim = self.hidden_size // self.num_heads
+
+        # define input, the order self.parameter matters
+        input = self.create_input_op((self.batch_size, self.seq_len, self.hidden_size))
+
+        hidden_states = input
+
+        # model norm and lm head
+        model_norm_weight = self.constant(model_norm_weight)
+        hidden_states = self.layer_norm(hidden_states, model_norm_weight)
+        if vocab_size == 122753:
+            # for MiniCPM-2B-sft-bf16
+            hidden_states_1 = self.linear(
+                hidden_states, 73440, self.hidden_size, bias=False, wt_dtype=self.dtype
+            )
+            hidden_states_2 = self.linear(
+                hidden_states, 73440, self.hidden_size, bias=False, wt_dtype=self.dtype
+            )
+            hidden_states_2 = self.slice(hidden_states_2, begin=[0, 0, 0], end=[1, 1, 49313])
+            hidden_states = self.concat(hidden_states_1, hidden_states_2, axis=2)
+        else:
+            # for MiniCPM-1B-sft-bf16
+            hidden_states = self.linear(
+                hidden_states, self.vocab_size, self.hidden_size, bias=False, wt_dtype=self.dtype
+            )
+
+        # define outputs
+        hidden_states = self.convert_to_fp32(hidden_states)
+
+        print("start compiling")
+        self.compile()
+
+
 def convert_lm_head_and_embedding(model, n_splits_linear, temp_dir, weight_dir):
     num_heads = model.model.layers[0].self_attn.num_heads
     num_key_value_heads = model.model.layers[0].self_attn.num_key_value_heads
@@ -72,24 +136,23 @@ def convert_lm_head_and_embedding(model, n_splits_linear, temp_dir, weight_dir):
     rms_norm_eps = model.config.rms_norm_eps
     vocab_size = model.config.vocab_size
     model_norm = model.model.norm
-    lm_head = model.lm_head
     if n_splits_linear == 1:
-        weights = [(lm_head.weight, lm_head.scale)]
+        if vocab_size == 122753:
+            # for MiniCPM-2B-sft-bf16
+            weights = [(model.lm_head_0.weight, model.lm_head_0.scale),
+                       (model.lm_head_1.weight, model.lm_head_1.scale)]
+        else:
+            # for MiniCPM-1B-sft-bf16
+            weights = [(model.lm_head.weight, model.lm_head.scale)]
     else:
-        lm_heads = lm_head.lm_heads
+        # TODO
-        lm_head_weights = []
+        pass
-        scales = []
-        for i in range(n_splits_linear):
-            lm_head_weights.append(lm_heads[i].weight)
-            scales.append(lm_heads[i].scale)
-        weights = [(torch.stack(lm_head_weights, axis=0),
-                    torch.stack(scales, axis=0))]
     if isinstance(weights[0], tuple):
         np_dtype = np.int8 if weights[0][0].dtype == torch.int8 else np.uint8
     else:  # FP16 Linear
         np_dtype = np.float16
 
-    new_lm_head = LowBitLLMLMHead(
+    new_lm_head = MiniCPMLMHead(
         [1, 1, num_heads * head_dim],
         num_heads=num_heads,
         max_seq_len=1,
@@ -99,17 +162,21 @@ def convert_lm_head_and_embedding(model, n_splits_linear, temp_dir, weight_dir):
         dtype=np_dtype,
         model_norm_weight=model_norm.weight.to(torch.float16),
         vocab_size=vocab_size,
-        n_splits=n_splits_linear
     )
     last_blob_path = update_names_of_IR_and_export_blob(new_lm_head, "lm_head", temp_dir)
 
     # save weights bins files
     if n_splits_linear == 1:
-        weight_numpy = [
+        if vocab_size == 122753:
-            lm_head.weight.data.numpy(), lm_head.scale.data.numpy(),
+            weight_numpy = [model.lm_head_0.weight.data.numpy(),
-        ]
+                            model.lm_head_0.scale.data.numpy(),
+                            model.lm_head_1.weight.data.numpy(),
+                            model.lm_head_1.scale.data.numpy(), ]
+        else:
+            weight_numpy = [model.lm_head.weight.data.numpy(), model.lm_head.scale.data.numpy(), ]
     else:
-        weight_numpy = [v.numpy() for v in weights[0]]
+        # TODO
+        pass
 
     for idx, weight in enumerate(weight_numpy):
         bin_file = os.path.join(weight_dir, f"model_lm_head_input_{1+idx}.bin")