ipex-llm 2.2.0b20250210__py3-none-win_amd64.whl → 2.2.0b20250211__py3-none-win_amd64.whl

ipex_llm/transformers/convert.py

@@ -1062,6 +1062,11 @@ def _optimize_pre(model, qtype=None):
         from ipex_llm.transformers.models.glm import merge_qkv, split_mlp
         model.apply(merge_qkv)
         model.apply(split_mlp)
+    elif model.config.model_type == "baichuan_m1":
+        from ipex_llm.transformers.models.baichuan_m1 import pre_register_inv_freq
+        model.apply(pre_register_inv_freq)
+    elif model.config.model_type == "multi_modality":
+        pass
 
     return model
 
@@ -1994,5 +1999,21 @@ def _optimize_post(model):
         model.llm.config.rope_scaling = {"rope_type": "default"}
         _optimize_post(model.llm)
         model.llm.config.model_type = "megrezo"
+    elif model.config.model_type == "baichuan_m1":
+        modeling_module_name = model.__class__.__module__
+        module = importlib.import_module(modeling_module_name)
+        from ipex_llm.transformers.models.common import rms_norm_forward
+        from ipex_llm.transformers.models.baichuan_m1 import model_forward
+        from ipex_llm.transformers.models.baichuan_m1 import eager_attention_forward
+        convert_forward(model, module.BaichuanModel, model_forward)
+        convert_forward(model, module.BaichuanRMSNorm, rms_norm_forward)
+        convert_forward(model, module.BaichuanAttention, eager_attention_forward)
+    elif model.config.model_type == "multi_modality":
+        # vision
+        vpm_modeling_module_name = model.vision_model.vision_tower.__class__.__module__
+        vpm_module = importlib.import_module(vpm_modeling_module_name)
+
+        from ipex_llm.transformers.models.janus import vision_attention_forward
+        convert_forward(model.vision_model, vpm_module.Attention, vision_attention_forward)
 
     return model

ipex_llm/transformers/low_bit_linear.py

@@ -699,7 +699,7 @@ class LowBitLinear(nn.Linear):
                 if is_server() and (not is_spr()) and \
                         self.qtype == SYM_INT4 and x_2d.shape[0] >= TORCH_LINEAR_THRESHOLD:
                     x0_fp32 = ggml_int4_convert_fp32(x0, self.weight_shape, self.weight_length)
-                    result = F.linear(x, x0_fp32)
+                    result = F.linear(x.to(dtype=x0_fp32.dtype), x0_fp32)
                 else:
                     # Weight does not need a convert
                     result = ggml_matmul_src1_x_src0_t(x0, x_2d, self.weight_shape, self.qtype)

ipex_llm/transformers/models/baichuan_m1.py

@@ -0,0 +1,240 @@
+#
+# Copyright 2016 The BigDL Authors.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+# This file is adapted from
+# https://huggingface.co/baichuan-inc/Baichuan-M1-14B-Instruct/blob/main/modeling_baichuan.py
+
+
+import math
+import torch
+import torch.nn.functional as F
+
+from typing import Optional, Tuple, Union
+from transformers.cache_utils import Cache
+from transformers.modeling_outputs import BaseModelOutputWithPast
+from ipex_llm.utils.common import invalidInputError
+from ipex_llm.transformers.models.utils import should_use_fuse_rope, repeat_kv
+from ipex_llm.transformers.models.common import attention_softmax
+from ipex_llm.transformers.models.common import scaled_dot_product_attention
+from ipex_llm.transformers.kv import DynamicNormalCache
+
+
+def pre_register_inv_freq(module: torch.nn.Module):
+    if module.__class__.__name__ == "RotaryEmbedding":
+        inv_freq = module.inv_freq
+        del module.inv_freq
+        module.register_buffer("inv_freq", inv_freq, persistent=False)
+
+
+# copied from Baichuan M1
+def custom_convolution(U, K):
+    """
+    U: Input matrix, shape (bs, seq, h, d)
+    K: Convolution kernel, shape (w, h)
+    Returns: Output matrix V, shape (bs, seq, h, d)
+    """
+    # h, w = K.shape
+    w = K.size(-1)
+    padding = (w - 1, 0)
+    U_padded = F.pad(U, (0, 0, 0, 0, *padding))  # Shape becomes (bs, seq+w-1, h, d)
+    U_unfolded = U_padded.unfold(1, w, 1)  # Shape becomes (bs, seq+w-1, h, d, w)
+    V_unfolded = U_unfolded * K  # Shape remains (bs, seq, h, d, w)
+    V = V_unfolded.sum(dim=-1)  # Shape becomes (bs, seq, h, d)
+    return V
+
+
+def model_forward(
+    self,
+    input_ids: torch.LongTensor = None,
+    attention_mask: Optional[torch.Tensor] = None,
+    position_ids: Optional[torch.LongTensor] = None,
+    seqlens: Optional[torch.LongTensor] = None,
+    past_key_values: Optional[Cache] = None,
+    inputs_embeds: Optional[torch.FloatTensor] = None,
+    use_cache: Optional[bool] = None,
+    output_attentions: Optional[bool] = None,
+    output_hidden_states: Optional[bool] = None,
+    return_dict: Optional[bool] = None,
+    cache_position: Optional[torch.LongTensor] = None,
+) -> Union[Tuple, BaseModelOutputWithPast]:
+    output_attentions = (
+        output_attentions if output_attentions is not None
+        else self.config.output_attentions
+    )
+    output_hidden_states = (
+        output_hidden_states if output_hidden_states is not None
+        else self.config.output_hidden_states
+    )
+    return_dict = return_dict if return_dict is not None else self.config.use_return_dict
+
+    invalidInputError((input_ids is None) ^ (inputs_embeds is None),
+                      "You cannot specify both input_ids and inputs_embeds at the same time, "
+                      "and must specify either one")
+
+    if inputs_embeds is None:
+        inputs_embeds = self.embed_tokens(input_ids)
+
+    use_cache = use_cache if use_cache is not None else self.config.use_cache
+    use_cache = True if inputs_embeds.device.type == "xpu" else use_cache
+
+    # IPEX-LLM changes start: remove batch multi-pack and use ipex-llm's kv cache
+    # kept for BC (non `Cache` `past_key_values` inputs)
+    if use_cache and not isinstance(past_key_values, DynamicNormalCache):
+        past_key_values = DynamicNormalCache.from_legacy_cache(past_key_values)
+    # IPEX-LLM changes end
+
+    if cache_position is None:
+        past_seen_tokens = past_key_values.get_seq_length() if past_key_values is not None else 0
+        cache_position = torch.arange(
+            past_seen_tokens, past_seen_tokens + inputs_embeds.shape[1],
+            device=inputs_embeds.device
+        )
+    if position_ids is None:
+        position_ids = cache_position.unsqueeze(0)
+
+    causal_mask = self._update_causal_mask(
+        attention_mask, inputs_embeds, cache_position, past_key_values, output_attentions
+    )
+
+    hidden_states = inputs_embeds
+
+    # create position embeddings to be shared across the decoder layers
+    # position_embeddings = self.rotary_emb(hidden_states, position_ids)
+    position_embeddings = None
+
+    # decoder layers
+    all_hidden_states = () if output_hidden_states else None
+    all_self_attns = () if output_attentions else None
+    next_decoder_cache = None
+
+    for decoder_layer in self.layers:
+        if output_hidden_states:
+            all_hidden_states += (hidden_states,)
+
+        layer_outputs = decoder_layer(
+            hidden_states,
+            attention_mask=causal_mask,
+            position_ids=position_ids,
+            seqlens=None,
+            past_key_value=past_key_values,
+            output_attentions=output_attentions,
+            use_cache=use_cache,
+            cache_position=cache_position,
+            position_embeddings=position_embeddings,
+        )
+
+        hidden_states = layer_outputs[0]
+        if use_cache:
+            next_decoder_cache = layer_outputs[2 if output_attentions else 1]
+
+        if output_attentions:
+            all_self_attns += (layer_outputs[1],)
+
+    hidden_states = self.norm(hidden_states)
+
+    # add hidden states from the last decoder layer
+    if output_hidden_states:
+        all_hidden_states += (hidden_states,)
+
+    next_cache = next_decoder_cache if use_cache else None
+    if not return_dict:
+        return tuple(v for v in [hidden_states, next_cache, all_hidden_states, all_self_attns]
+                     if v is not None)
+    return BaseModelOutputWithPast(
+        last_hidden_state=hidden_states,
+        past_key_values=next_cache,
+        hidden_states=all_hidden_states,
+        attentions=all_self_attns,
+    )
+
+
+def eager_attention_forward(
+    self,
+    hidden_states: torch.Tensor,
+    attention_mask: Optional[torch.Tensor] = None,
+    position_ids: Optional[torch.LongTensor] = None,
+    seqlens: Optional[torch.LongTensor] = None,
+    past_key_value: Optional[Cache] = None,
+    output_attentions: bool = False,
+    use_cache: bool = False,
+    cache_position: Optional[torch.LongTensor] = None,
+    position_embeddings: Optional[Tuple[torch.Tensor, torch.Tensor]]=None,
+):
+    invalidInputError(seqlens is None, "`seq_lens` must be None")
+
+    bsz, q_len, _ = hidden_states.size()
+    qkv = self.W_pack(hidden_states)
+    qkv = qkv.view(bsz, q_len, self.num_heads + 2 * self.num_key_value_heads, self.head_dim)
+    query_states, key_states, value_states = qkv.split([self.num_heads,
+                                                        self.num_key_value_heads,
+                                                        self.num_key_value_heads], dim=2)
+    # q, k, v: [bsz, seq_len, num_heads, head_dim]
+
+    if past_key_value is None or past_key_value.get_seq_length(self.layer_idx) == 0:    # prefill
+        self.last_k = key_states[:, -1:]
+        self.last_v = value_states[:, -1:]
+
+        key_states = custom_convolution(key_states, self.conv_k)
+        value_states = custom_convolution(value_states, self.conv_v)
+    else:
+        new_key_states = (self.conv_k[0, 0, :, 0, :1] * self.last_k +
+                          self.conv_k[0, 0, :, 0, 1:] * key_states)
+        self.last_k = key_states
+        key_states = new_key_states
+
+        new_value_states = (self.conv_v[0, 0, :, 0, : 1] * self.last_v +
+                            self.conv_v[0, 0, :, 0, 1:] * value_states)
+        self.last_v = value_states
+        value_states = new_value_states
+
+    query_states = query_states.transpose(1, 2)
+    key_states = key_states.transpose(1, 2)
+    value_states = value_states.transpose(1, 2)
+    # q, k, v: [bsz, num_heads, seq_len, head_dim]
+
+    invalidInputError(should_use_fuse_rope(hidden_states, position_ids, self.training),
+                      "fuse rope must be used")
+    import xe_addons
+    xe_addons.rotary_half_inplaced(self.rotary_emb.inv_freq, position_ids,
+                                   query_states, key_states)
+
+    # ignore sliding window
+    key_states, value_states = past_key_value.update(key_states, value_states,
+                                                     self.layer_idx, None)
+    if self.head_dim <= 128:
+        attn_weights = None
+        attn_output = scaled_dot_product_attention(
+            query_states, key_states, value_states,
+            attention_mask, q_len == key_states.size(2)
+        )
+    else:
+        n_rep = self.num_heads // self.num_key_value_heads
+        key_states = repeat_kv(key_states, n_rep)
+        value_states = repeat_kv(value_states, n_rep)
+        attn_weights = torch.matmul(query_states,
+                                    key_states.transpose(2, 3)) / math.sqrt(self.head_dim)
+        if attention_mask is not None:
+            attn_weights = attn_weights + attention_mask
+        attn_weights = attention_softmax(attn_weights)
+        attn_output = torch.matmul(attn_weights, value_states)
+
+    attn_output = attn_output.transpose(1, 2).contiguous()
+    attn_output = attn_output.reshape(bsz, q_len, self.hidden_size)
+
+    attn_output = self.o_proj(attn_output)
+
+    if not output_attentions:
+        attn_weights = None
+    return attn_output, attn_weights, past_key_value

{ipex_llm-2.2.0b20250210.dist-info → ipex_llm-2.2.0b20250211.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: ipex-llm
-Version: 2.2.0b20250210
+Version: 2.2.0b20250211
 Summary: Large Language Model Develop Toolkit
 Home-page: https://github.com/intel-analytics/ipex-llm
 Author: BigDL Authors
@@ -27,7 +27,7 @@ Requires-Dist: intel-openmp ; (platform_machine == "x86_64" or platform_machine
 Requires-Dist: torch ==2.1.2+cpu ; (platform_system == "Linux") and extra == 'all'
 Requires-Dist: torch ==2.1.2 ; (platform_system == "Windows") and extra == 'all'
 Provides-Extra: cpp
-Requires-Dist: bigdl-core-cpp ==2.6.0b20250210 ; extra == 'cpp'
+Requires-Dist: bigdl-core-cpp ==2.6.0b20250211 ; extra == 'cpp'
 Requires-Dist: setuptools ; extra == 'cpp'
 Requires-Dist: onednn-devel ==2025.0.1 ; (platform_system == "Windows") and extra == 'cpp'
 Requires-Dist: onednn ==2025.0.1 ; (platform_system == "Windows") and extra == 'cpp'
@@ -60,7 +60,7 @@ Requires-Dist: transformers ==4.40.0 ; extra == 'npu'
 Requires-Dist: intel-openmp ; (platform_machine == "x86_64" or platform_machine == "AMD64") and extra == 'npu'
 Requires-Dist: torch ==2.1.2+cpu ; (platform_system == "Linux") and extra == 'npu'
 Requires-Dist: torch ==2.1.2 ; (platform_system == "Windows") and extra == 'npu'
-Requires-Dist: bigdl-core-npu ==2.6.0b20250210 ; (platform_system == "Windows") and extra == 'npu'
+Requires-Dist: bigdl-core-npu ==2.6.0b20250211 ; (platform_system == "Windows") and extra == 'npu'
 Provides-Extra: serving
 Requires-Dist: py-cpuinfo ; extra == 'serving'
 Requires-Dist: fschat[model_worker,webui] ==0.2.36 ; extra == 'serving'
@@ -80,9 +80,9 @@ Requires-Dist: setuptools <70.0.0 ; extra == 'xpu'
 Requires-Dist: torch ==2.1.0a0 ; extra == 'xpu'
 Requires-Dist: torchvision ==0.16.0a0 ; extra == 'xpu'
 Requires-Dist: intel-extension-for-pytorch ==2.1.10+xpu ; extra == 'xpu'
-Requires-Dist: bigdl-core-xe-21 ==2.6.0b20250210 ; extra == 'xpu'
-Requires-Dist: bigdl-core-xe-batch-21 ==2.6.0b20250210 ; extra == 'xpu'
-Requires-Dist: bigdl-core-xe-addons-21 ==2.6.0b20250210 ; extra == 'xpu'
+Requires-Dist: bigdl-core-xe-21 ==2.6.0b20250211 ; extra == 'xpu'
+Requires-Dist: bigdl-core-xe-batch-21 ==2.6.0b20250211 ; extra == 'xpu'
+Requires-Dist: bigdl-core-xe-addons-21 ==2.6.0b20250211 ; extra == 'xpu'
 Provides-Extra: xpu-2-1
 Requires-Dist: py-cpuinfo ; extra == 'xpu-2-1'
 Requires-Dist: protobuf ; extra == 'xpu-2-1'
@@ -97,9 +97,9 @@ Requires-Dist: setuptools <70.0.0 ; extra == 'xpu-2-1'
 Requires-Dist: torch ==2.1.0a0 ; extra == 'xpu-2-1'
 Requires-Dist: torchvision ==0.16.0a0 ; extra == 'xpu-2-1'
 Requires-Dist: intel-extension-for-pytorch ==2.1.10+xpu ; extra == 'xpu-2-1'
-Requires-Dist: bigdl-core-xe-21 ==2.6.0b20250210 ; extra == 'xpu-2-1'
-Requires-Dist: bigdl-core-xe-batch-21 ==2.6.0b20250210 ; extra == 'xpu-2-1'
-Requires-Dist: bigdl-core-xe-addons-21 ==2.6.0b20250210 ; extra == 'xpu-2-1'
+Requires-Dist: bigdl-core-xe-21 ==2.6.0b20250211 ; extra == 'xpu-2-1'
+Requires-Dist: bigdl-core-xe-batch-21 ==2.6.0b20250211 ; extra == 'xpu-2-1'
+Requires-Dist: bigdl-core-xe-addons-21 ==2.6.0b20250211 ; extra == 'xpu-2-1'
 Requires-Dist: intel-openmp ; (platform_machine == "x86_64" or platform_machine == "AMD64") and extra == 'xpu-2-1'
 Requires-Dist: dpcpp-cpp-rt ==2024.0.2 ; (platform_system == "Windows") and extra == 'xpu-2-1'
 Requires-Dist: mkl-dpcpp ==2024.0.0 ; (platform_system == "Windows") and extra == 'xpu-2-1'
@@ -117,7 +117,7 @@ Requires-Dist: setuptools ; extra == 'xpu-2-6'
 Requires-Dist: torch ==2.6.0+xpu ; extra == 'xpu-2-6'
 Requires-Dist: torchvision ==0.21.0+xpu ; extra == 'xpu-2-6'
 Requires-Dist: torchaudio ==2.6.0+xpu ; extra == 'xpu-2-6'
-Requires-Dist: bigdl-core-xe-all ==2.6.0b20250210 ; extra == 'xpu-2-6'
+Requires-Dist: bigdl-core-xe-all ==2.6.0b20250211 ; extra == 'xpu-2-6'
 Requires-Dist: onednn-devel ==2025.0.1 ; extra == 'xpu-2-6'
 Requires-Dist: onednn ==2025.0.1 ; extra == 'xpu-2-6'
 Requires-Dist: dpcpp-cpp-rt ==2025.0.2 ; extra == 'xpu-2-6'
@@ -133,9 +133,9 @@ Requires-Dist: tokenizers ==0.15.2 ; extra == 'xpu-arc'
 Requires-Dist: accelerate ==0.23.0 ; extra == 'xpu-arc'
 Requires-Dist: tabulate ; extra == 'xpu-arc'
 Requires-Dist: setuptools ; extra == 'xpu-arc'
-Requires-Dist: bigdl-core-xe-23 ==2.6.0b20250210 ; extra == 'xpu-arc'
-Requires-Dist: bigdl-core-xe-batch-23 ==2.6.0b20250210 ; extra == 'xpu-arc'
-Requires-Dist: bigdl-core-xe-addons-23 ==2.6.0b20250210 ; extra == 'xpu-arc'
+Requires-Dist: bigdl-core-xe-23 ==2.6.0b20250211 ; extra == 'xpu-arc'
+Requires-Dist: bigdl-core-xe-batch-23 ==2.6.0b20250211 ; extra == 'xpu-arc'
+Requires-Dist: bigdl-core-xe-addons-23 ==2.6.0b20250211 ; extra == 'xpu-arc'
 Requires-Dist: intel-openmp ; (platform_machine == "x86_64" or platform_machine == "AMD64") and extra == 'xpu-arc'
 Requires-Dist: torch ==2.3.1+cxx11.abi ; (platform_system == "Linux") and extra == 'xpu-arc'
 Requires-Dist: torchvision ==0.18.1+cxx11.abi ; (platform_system == "Linux") and extra == 'xpu-arc'
@@ -156,9 +156,9 @@ Requires-Dist: tokenizers ==0.15.2 ; extra == 'xpu-arl'
 Requires-Dist: accelerate ==0.23.0 ; extra == 'xpu-arl'
 Requires-Dist: tabulate ; extra == 'xpu-arl'
 Requires-Dist: setuptools ; extra == 'xpu-arl'
-Requires-Dist: bigdl-core-xe-23 ==2.6.0b20250210 ; extra == 'xpu-arl'
-Requires-Dist: bigdl-core-xe-batch-23 ==2.6.0b20250210 ; extra == 'xpu-arl'
-Requires-Dist: bigdl-core-xe-addons-23 ==2.6.0b20250210 ; extra == 'xpu-arl'
+Requires-Dist: bigdl-core-xe-23 ==2.6.0b20250211 ; extra == 'xpu-arl'
+Requires-Dist: bigdl-core-xe-batch-23 ==2.6.0b20250211 ; extra == 'xpu-arl'
+Requires-Dist: bigdl-core-xe-addons-23 ==2.6.0b20250211 ; extra == 'xpu-arl'
 Requires-Dist: intel-openmp ; (platform_machine == "x86_64" or platform_machine == "AMD64") and extra == 'xpu-arl'
 Requires-Dist: torch ==2.3.1+cxx11.abi ; (platform_system == "Linux") and extra == 'xpu-arl'
 Requires-Dist: torchvision ==0.18.1+cxx11.abi ; (platform_system == "Linux") and extra == 'xpu-arl'
@@ -179,9 +179,9 @@ Requires-Dist: tokenizers ==0.15.2 ; extra == 'xpu-lnl'
 Requires-Dist: accelerate ==0.23.0 ; extra == 'xpu-lnl'
 Requires-Dist: tabulate ; extra == 'xpu-lnl'
 Requires-Dist: setuptools ; extra == 'xpu-lnl'
-Requires-Dist: bigdl-core-xe-23 ==2.6.0b20250210 ; extra == 'xpu-lnl'
-Requires-Dist: bigdl-core-xe-batch-23 ==2.6.0b20250210 ; extra == 'xpu-lnl'
-Requires-Dist: bigdl-core-xe-addons-23 ==2.6.0b20250210 ; extra == 'xpu-lnl'
+Requires-Dist: bigdl-core-xe-23 ==2.6.0b20250211 ; extra == 'xpu-lnl'
+Requires-Dist: bigdl-core-xe-batch-23 ==2.6.0b20250211 ; extra == 'xpu-lnl'
+Requires-Dist: bigdl-core-xe-addons-23 ==2.6.0b20250211 ; extra == 'xpu-lnl'
 Requires-Dist: intel-openmp ; (platform_machine == "x86_64" or platform_machine == "AMD64") and extra == 'xpu-lnl'
 Requires-Dist: torch ==2.3.1+cxx11.abi ; (platform_system == "Linux") and extra == 'xpu-lnl'
 Requires-Dist: torchvision ==0.18.1+cxx11.abi ; (platform_system == "Linux") and extra == 'xpu-lnl'

{ipex_llm-2.2.0b20250210.dist-info → ipex_llm-2.2.0b20250211.dist-info}/RECORD

@@ -41,35 +41,35 @@ ipex_llm/langchain/llms/transformerspipelinellm.py,sha256=vm522YPPwWxxAPVvQBtxRf
 ipex_llm/langchain/vllm/__init__.py,sha256=T-EbRT6GJ_8RCu-iLmSzcftOimXSPQf2d5X72AUAy2Y,874
 ipex_llm/langchain/vllm/vllm.py,sha256=6dxc-ZISZQrJilEa_HA827l75Dv9rcHpY_G6FdJ8BVs,7793
 ipex_llm/libs/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
-ipex_llm/libs/bloom-api.dll,sha256=lqfplzgoUUcMRerQS4Dnj9_EPY3AqYLKfOYED27YIAs,36352
-ipex_llm/libs/bloom.dll,sha256=kevOYhURxmjZ4uK7JxHchlYg8NOTbOo_iaKIokKpIKo,507904
-ipex_llm/libs/gptneox-api.dll,sha256=DdmNMxtkgh8sRsGFTYjSHm_W_BgDAIiqS3ooWEtMMgs,24576
-ipex_llm/libs/gptneox.dll,sha256=ce9pad_x--9MNHLVZx76amHcD-FLdHUUX_3J4XoPzl4,568320
-ipex_llm/libs/libbloom_avx.dll,sha256=z-a8kKkkKUMf9aBIeJlEM2y_sRMPKEf3ZXZZICl5J6Q,536576
-ipex_llm/libs/libbloom_vnni.dll,sha256=F6S_Yvx_ZBibwz-ggGKX_qX7g-Rxf7Auj3lF7MdFBmU,508416
-ipex_llm/libs/libgptneox_avx.dll,sha256=78wwK0DXFVL6LmUWXKYdpCDkDTOFMXfce6R1iL-SZO0,596992
-ipex_llm/libs/libgptneox_vnni.dll,sha256=tQD2cQPx9dtZsTJ1Ud1OtYOWq136x8mY7VzMfs8VVH4,568832
-ipex_llm/libs/libllama_avx.dll,sha256=ZwfgXj4gjx8zNwI-jdyMOXqY2ZCnDVWqiYr8XCnQ8KY,591360
-ipex_llm/libs/libllama_vnni.dll,sha256=GzUixqZsWDEAQ6_wQdoQfg7uetKNhNokdge934WUxZ4,563200
-ipex_llm/libs/libstarcoder_avx.dll,sha256=mlnT4q1xb5CQzkqwvNQE1mvjAib6o6v9J4eyObYn5KE,627712
-ipex_llm/libs/libstarcoder_vnni.dll,sha256=8GctyMyR5X6ienHOJ6vAoSb9AOG6iIXaSchDs3rH95g,599552
-ipex_llm/libs/llama-api.dll,sha256=Kes9mqsYLyHZ1u09ztljvdlgkyS6cuqQN3zIiQ2eRRQ,25600
-ipex_llm/libs/llama.dll,sha256=lEioZzQgKIY8KJxvSkktyEnLcncLChlUPjuzkNigMIs,562688
-ipex_llm/libs/main-bloom.exe,sha256=OsQCjZrYp2bZvqNSx8SF-5Du9i8rzkVFAnyryNYfTW0,103424
-ipex_llm/libs/main-gptneox.exe,sha256=28dbFg81AFpqUxOs0fUNKa-D0gtasqZcElWqOW1AczY,98816
-ipex_llm/libs/main-llama.exe,sha256=B3SyXbPj0bmtCa2jMIpixXe1cbNYSYwTDosFx4c9BDA,99840
-ipex_llm/libs/main-starcoder.exe,sha256=UBv9JjyhqVRi4vFHIxz7ZT5am0NF_wCrordhxKu2slo,157696
-ipex_llm/libs/pipeline.dll,sha256=MBRq9mjVP1ixvdlikvUaawdkv3DyYAEWN4-zTtXj42M,73216
-ipex_llm/libs/quantize-bloom.exe,sha256=WyhdSu43U22MUsMaPAHHrxNRuWyha7kiK_qVo562NhQ,126464
-ipex_llm/libs/quantize-bloom_vnni.exe,sha256=LaQyFrkPOS87-1DAuzFkgVBJSoxulwQqcSWcb5O3IWY,128000
-ipex_llm/libs/quantize-gptneox.exe,sha256=IhJcljCZ5buuyPNHXesadudvXGA1jg45YsBTqPcaEnE,104448
-ipex_llm/libs/quantize-gptneox_vnni.exe,sha256=Jdo_A4hOOdKK_9Rle4FUDD_grbYoNJGBMz0EP8Pp3Qo,104960
-ipex_llm/libs/quantize-llama.exe,sha256=aNUQ_mkgFQPi4lBFpDzMg3RjIQWgypZ3NjnIrr3GkqE,110080
-ipex_llm/libs/quantize-llama_vnni.exe,sha256=hnCSMssU4NRXK0IXuCJWfhqegiJvnrH8e9TUX50fXzo,110592
-ipex_llm/libs/quantize-starcoder.exe,sha256=QhqhYRKafiOpgc48YBTzF5MxvREguASQ6_wXjIvNiZo,127488
-ipex_llm/libs/quantize-starcoder_vnni.exe,sha256=PB2mturnl4F6OppCrb0cY3vHS7LmrEP9EWtWamczUpg,128512
-ipex_llm/libs/starcoder-api.dll,sha256=UKfEaI5pADFdvWkWT2D-V6SG60yy2uHjQ2PPJvjxkxo,21504
-ipex_llm/libs/starcoder.dll,sha256=f34bCkWEt5CdvLIbVYfQydoH_qnodFZ8oJIrCbHLdOA,599040
+ipex_llm/libs/bloom-api.dll,sha256=isRVYs9jZHXj_OATH3LASq7rAtiuZF1EhRon0kLa1fQ,36352
+ipex_llm/libs/bloom.dll,sha256=55-F9cCHSlfDSrARFN8Mg_SM-wC7qSGvNaD24cR_R-4,507904
+ipex_llm/libs/gptneox-api.dll,sha256=4VKDx3KffmcU8T638sMe6nIG4p3gfSG6AVE9OwRnXXU,24576
+ipex_llm/libs/gptneox.dll,sha256=dGkabZ_ANnynUkBVsALka40oUXiex0nTVhALqrXyiG0,568320
+ipex_llm/libs/libbloom_avx.dll,sha256=RiKXeDK2UqahcWL3s6WOfQdYNT7luAl0rI2jGi6jHV4,536576
+ipex_llm/libs/libbloom_vnni.dll,sha256=deQqdDxqaUJWekAQ_CJMLnZhDbDn1WqlWCZ3QGfPFY4,508416
+ipex_llm/libs/libgptneox_avx.dll,sha256=C1R1hD1XsK-xnTBeI7NJFtilDtZn-Mu16cHqPGyo4W4,596992
+ipex_llm/libs/libgptneox_vnni.dll,sha256=S5ErqE9n8PEZumuxXgesQcfhSRnfdcVBCpPJXAF_7qM,568832
+ipex_llm/libs/libllama_avx.dll,sha256=8XuQpQ9g8EPccaBhime786x4l-KlQGJAhpDKi33z1L0,591360
+ipex_llm/libs/libllama_vnni.dll,sha256=NYsBYek3QBLAZjRAIEFiY4OUM78lByDzt8-c7vyyARw,563200
+ipex_llm/libs/libstarcoder_avx.dll,sha256=4sxz29nT-YHKtIt2XubEhasFloDkyQgPnZsw78YKRcs,627712
+ipex_llm/libs/libstarcoder_vnni.dll,sha256=fCm2zNDmCEE8vT2dnDFo0o7Dg1t-LYIfVZqoJlJF_24,599552
+ipex_llm/libs/llama-api.dll,sha256=7QeY4YePf7aF6yYcRQdCfm_NNgdMhnc5mL3LBoTJACg,25600
+ipex_llm/libs/llama.dll,sha256=TMVwGScsZ4SiaEqQ4jWrwUdZGxxeM5-rBHubjMNVsOs,562688
+ipex_llm/libs/main-bloom.exe,sha256=Ci0vgb1q4ArlQJ4PvDoZrT5koeKGAkpmcHgFOEBKrnk,103424
+ipex_llm/libs/main-gptneox.exe,sha256=3xWuAqEfCxSlHEK6cgUtOI1gFqOKxI5FlTTWuZf9Jlw,98816
+ipex_llm/libs/main-llama.exe,sha256=55ryHK75UQJjxdfZlm4GQGEHe7LfyxBkDhLR_l_bh10,99840
+ipex_llm/libs/main-starcoder.exe,sha256=zpqmu07KnGgR0QW69dSZYHuT_sm5DfAV8xeRspMpm_0,157696
+ipex_llm/libs/pipeline.dll,sha256=ssy7Oi2s4atVipwbA97C4IrOzlcfXyzPY3LSAQV5jGs,73216
+ipex_llm/libs/quantize-bloom.exe,sha256=-bKBNkqsVKv3v2ykATIp8hu3Ih4tvBAKUmAhKG0K9iQ,126464
+ipex_llm/libs/quantize-bloom_vnni.exe,sha256=eoiP-B9w4C4Z6NQH6mV8WD1e7IvrOFIqcZ-PU8kTqdA,128000
+ipex_llm/libs/quantize-gptneox.exe,sha256=srPfQVhfeU9wRwEATzmxMRb0HxTeKzxSJ-_MlsUjs4g,104448
+ipex_llm/libs/quantize-gptneox_vnni.exe,sha256=POmpcftpGvGnmieDpGzSEWG_ihFZRdILdu5tp7ovmc0,104960
+ipex_llm/libs/quantize-llama.exe,sha256=R1arXrmiSZUkJR62tOg6lAgrwPjqulvzJFW6fXFROxk,110080
+ipex_llm/libs/quantize-llama_vnni.exe,sha256=MKiyUKvjO29hMnFqjI2odtWRtcj12DdtwYuCme2DAlg,110592
+ipex_llm/libs/quantize-starcoder.exe,sha256=0KGcTuDuygIbRWkKpVgCTts3Oz4iJJ0qVgU0S4zmTHg,127488
+ipex_llm/libs/quantize-starcoder_vnni.exe,sha256=19EvwpYsucRzqardslcUcABpgTRliz63iGLaUC0bGE0,128512
+ipex_llm/libs/starcoder-api.dll,sha256=dl_63rrUlJ65_iitS008gxs57LyDIRB69DDkrNdgIvU,21504
+ipex_llm/libs/starcoder.dll,sha256=Y8X0huLZBfzjCCJ1ddjzkPRxzKCvLoK5h48JFX705E4,599040
 ipex_llm/llamaindex/__init__.py,sha256=T-EbRT6GJ_8RCu-iLmSzcftOimXSPQf2d5X72AUAy2Y,874
 ipex_llm/llamaindex/llms/__init__.py,sha256=KP1lEdGqDuxPoxL1ZSH25Pm2kKMPJBWUTLR0ckSLMIU,1139
 ipex_llm/llamaindex/llms/bigdlllm.py,sha256=FQBzq1KOjfc6uofTXAha3O7TqpJkNfOFepXQmOVlbnI,26314
@@ -87,14 +87,14 @@ ipex_llm/serving/fastchat/tgi_api_protocol.py,sha256=brT3k3-V0NJrU4fRqUwWjC0O3iO
 ipex_llm/serving/fastchat/tgi_api_server.py,sha256=agNTAEiZPSuj3dEdIdYKwkoY0cXOUDX06DiM9VP2knQ,24418
 ipex_llm/serving/fastchat/vllm_worker.py,sha256=ZLz2Q9GxJO6r_LOiP6epgCRjBGk-K4EB1SNEWSJp5DA,11091
 ipex_llm/transformers/__init__.py,sha256=pJHs6GZXHIObVE4BUCuej-6BKBZZg9pYWKPrkhWSfB4,1192
-ipex_llm/transformers/convert.py,sha256=Ss4q1bsMgGMU2hauZJxUsEA2noPByR-fALrJGPN1sEk,99520
+ipex_llm/transformers/convert.py,sha256=42qHApc3hoL38ldQXTv9BkDVQ0Zb-xGJYA2DWYFzwMg,100795
 ipex_llm/transformers/convert_ipex.py,sha256=_nSnUTQy-yfkKaqGdqnBdWztZf3NGmnbZ0TKaDrF4X4,14617
 ipex_llm/transformers/embedding.py,sha256=bdgk59DvD4ZZyxRzewXOR7g56nThgO6uhIwk8QL7f-s,9299
 ipex_llm/transformers/kv.py,sha256=k4TU18LlA-Sbq9WNNQnfuzu3RSFBwFhmaV3BcGN5bAo,19191
 ipex_llm/transformers/lisa.py,sha256=F5WxbtXQ7RdKulj83h_2DnEIgKiKGZf7zvOmg6QBl2s,3289
 ipex_llm/transformers/loader.py,sha256=AwjV5RpI2t2bedlv7ZhLm8cfd-QJZm5hny-XyjIvdnk,6876
 ipex_llm/transformers/lookup.py,sha256=b6OlZ9OV10R9qeWw8mVryVpDxszkjwLkldvi7GPMJY8,19614
-ipex_llm/transformers/low_bit_linear.py,sha256=3EtbiCAq5HU_r2pGJ9beSDK4NPTN8Jj-aHMqm1jqX18,39177
+ipex_llm/transformers/low_bit_linear.py,sha256=1S8H684odAx5ZVDx_qNMv4FFrkJekZduao8datZYiqw,39201
 ipex_llm/transformers/model.py,sha256=FyHrEQhkHxG3FbGkhTjVOP2rgFMjc3AXcjDwvvB0HqU,40798
 ipex_llm/transformers/modelling_bigdl.py,sha256=7JpNVMuyq_OmtNUaMFMXdxPWZp2q0QHC02QeA-VTPOw,6709
 ipex_llm/transformers/npu_model.py,sha256=zgXOiLIJ-3p-1Kejgv4jUFK8OiBZbezMZrRyn0_6_8c,40306
@@ -138,6 +138,7 @@ ipex_llm/transformers/gguf/models/model_implement/yuan2/yuan_hf_model.py,sha256=
 ipex_llm/transformers/models/__init__.py,sha256=tp2DcVkKg1-QvdYk7DY7rZvQWCDQ4ZjU8NAQ7Fclrpg,584
 ipex_llm/transformers/models/aquila.py,sha256=VZb5Drpo_fTxwcExZ397LygnsNPX2sVbie9_JeFudZI,5252
 ipex_llm/transformers/models/baichuan.py,sha256=8b43mBRZJEf_xLNoodhA4r9x1anqwC3Wt8awWel-aUo,18306
+ipex_llm/transformers/models/baichuan_m1.py,sha256=l6BD9jbA8TQ5Q0fcngHNRw73SxqvC126ErENs7Zh-do,9513
 ipex_llm/transformers/models/bert.py,sha256=0Mm9jkvkzBxtc_z_GE1TcZoPz-HOg2Z2973ZEWgSwJk,5601
 ipex_llm/transformers/models/bloom.py,sha256=PxfzyYT-nFn3K5rZhTQjmcEjUUzAhUFzxIN4kzRlCuc,8103
 ipex_llm/transformers/models/chatglm.py,sha256=UHai1t2AUtGmF765_eHF8LUMVQzp_oCBx8TJB21WrHk,12597
@@ -248,11 +249,11 @@ ipex_llm/vllm/xpu/engine/__init__.py,sha256=pY_CpyuZd72fr6s32ejeKHKFW0K4vUU2rzZj
 ipex_llm/vllm/xpu/engine/engine.py,sha256=k4-D27WS_Gk3mA--w3HWAjPjb4Aiu043MVPi0ZoAUBc,5984
 ipex_llm/vllm/xpu/entrypoints/openai/api_server.py,sha256=GshTZFB8e4PWvqckfbmTOU6b0oLkNn7A-vzLuG9--j8,21544
 ipex_llm/vllm/xpu/entrypoints/openai/cli_args.py,sha256=2rENA2ucynMaIjiZBEh2ez1o5vR32GaP514t39CD7KM,8676
-ipex_llm-2.2.0b20250210.data/scripts/ipex-llm-init.bat,sha256=HPtCYuDYwEatq7dAwOvdfVcHYCpAVdbj75K1qh0vQek,2578
-ipex_llm-2.2.0b20250210.data/scripts/llm-chat.ps1,sha256=6qrs-hGVAV8IKh7Jx8nq_XrnZcjd7qGU5wndArM7Yag,2769
-ipex_llm-2.2.0b20250210.data/scripts/llm-cli.ps1,sha256=3qBtTLs_EjYDnM8YyCpJhzLnGCKTEGssu9UNqfkjVXs,3009
-ipex_llm-2.2.0b20250210.dist-info/METADATA,sha256=4qPoDhuk8GOrYMBA3s4y5uNItLQm0QLLbP_B_0kYaWM,12369
-ipex_llm-2.2.0b20250210.dist-info/WHEEL,sha256=6iYPr8vTHsyDK75jr9X0V3I9wPSVmtwr_8fdATBciGk,98
-ipex_llm-2.2.0b20250210.dist-info/entry_points.txt,sha256=TiUyBB2MRmfF3ko-pyAEzqeBCRnyhu27bNOAsWPp3e8,61
-ipex_llm-2.2.0b20250210.dist-info/top_level.txt,sha256=CGCMHM-SyqUabU4h8RqJ2KTYckQUO3LvIWwmUQ6Qbzw,9
-ipex_llm-2.2.0b20250210.dist-info/RECORD,,
+ipex_llm-2.2.0b20250211.data/scripts/ipex-llm-init.bat,sha256=HPtCYuDYwEatq7dAwOvdfVcHYCpAVdbj75K1qh0vQek,2578
+ipex_llm-2.2.0b20250211.data/scripts/llm-chat.ps1,sha256=6qrs-hGVAV8IKh7Jx8nq_XrnZcjd7qGU5wndArM7Yag,2769
+ipex_llm-2.2.0b20250211.data/scripts/llm-cli.ps1,sha256=3qBtTLs_EjYDnM8YyCpJhzLnGCKTEGssu9UNqfkjVXs,3009
+ipex_llm-2.2.0b20250211.dist-info/METADATA,sha256=bXwvxJIBVIayYIeP1lVUyFJXiMqzVppGcvwttejTvOk,12369
+ipex_llm-2.2.0b20250211.dist-info/WHEEL,sha256=6iYPr8vTHsyDK75jr9X0V3I9wPSVmtwr_8fdATBciGk,98
+ipex_llm-2.2.0b20250211.dist-info/entry_points.txt,sha256=TiUyBB2MRmfF3ko-pyAEzqeBCRnyhu27bNOAsWPp3e8,61
+ipex_llm-2.2.0b20250211.dist-info/top_level.txt,sha256=CGCMHM-SyqUabU4h8RqJ2KTYckQUO3LvIWwmUQ6Qbzw,9
+ipex_llm-2.2.0b20250211.dist-info/RECORD,,