ipex-llm 2.3.0b20250410__py3-none-win_amd64.whl → 2.3.0b20250414__py3-none-win_amd64.whl
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- ipex_llm/libs/bloom-api.dll +0 -0
- ipex_llm/libs/bloom.dll +0 -0
- ipex_llm/libs/gptneox-api.dll +0 -0
- ipex_llm/libs/gptneox.dll +0 -0
- ipex_llm/libs/libbloom_avx.dll +0 -0
- ipex_llm/libs/libbloom_vnni.dll +0 -0
- ipex_llm/libs/libgptneox_avx.dll +0 -0
- ipex_llm/libs/libgptneox_vnni.dll +0 -0
- ipex_llm/libs/libllama_avx.dll +0 -0
- ipex_llm/libs/libllama_vnni.dll +0 -0
- ipex_llm/libs/libstarcoder_avx.dll +0 -0
- ipex_llm/libs/libstarcoder_vnni.dll +0 -0
- ipex_llm/libs/llama-api.dll +0 -0
- ipex_llm/libs/llama.dll +0 -0
- ipex_llm/libs/main-bloom.exe +0 -0
- ipex_llm/libs/main-gptneox.exe +0 -0
- ipex_llm/libs/main-llama.exe +0 -0
- ipex_llm/libs/main-starcoder.exe +0 -0
- ipex_llm/libs/pipeline.dll +0 -0
- ipex_llm/libs/quantize-bloom.exe +0 -0
- ipex_llm/libs/quantize-bloom_vnni.exe +0 -0
- ipex_llm/libs/quantize-gptneox.exe +0 -0
- ipex_llm/libs/quantize-gptneox_vnni.exe +0 -0
- ipex_llm/libs/quantize-llama.exe +0 -0
- ipex_llm/libs/quantize-llama_vnni.exe +0 -0
- ipex_llm/libs/quantize-starcoder.exe +0 -0
- ipex_llm/libs/quantize-starcoder_vnni.exe +0 -0
- ipex_llm/libs/starcoder-api.dll +0 -0
- ipex_llm/libs/starcoder.dll +0 -0
- ipex_llm/transformers/convert.py +9 -0
- ipex_llm/transformers/kv.py +51 -1
- ipex_llm/transformers/models/common.py +3 -3
- ipex_llm/transformers/models/utils.py +43 -0
- {ipex_llm-2.3.0b20250410.dist-info → ipex_llm-2.3.0b20250414.dist-info}/METADATA +20 -20
- {ipex_llm-2.3.0b20250410.dist-info → ipex_llm-2.3.0b20250414.dist-info}/RECORD +41 -41
- {ipex_llm-2.3.0b20250410.data → ipex_llm-2.3.0b20250414.data}/scripts/ipex-llm-init.bat +0 -0
- {ipex_llm-2.3.0b20250410.data → ipex_llm-2.3.0b20250414.data}/scripts/llm-chat.ps1 +0 -0
- {ipex_llm-2.3.0b20250410.data → ipex_llm-2.3.0b20250414.data}/scripts/llm-cli.ps1 +0 -0
- {ipex_llm-2.3.0b20250410.dist-info → ipex_llm-2.3.0b20250414.dist-info}/WHEEL +0 -0
- {ipex_llm-2.3.0b20250410.dist-info → ipex_llm-2.3.0b20250414.dist-info}/entry_points.txt +0 -0
- {ipex_llm-2.3.0b20250410.dist-info → ipex_llm-2.3.0b20250414.dist-info}/top_level.txt +0 -0
ipex_llm/libs/bloom-api.dll
CHANGED
Binary file
|
ipex_llm/libs/bloom.dll
CHANGED
Binary file
|
ipex_llm/libs/gptneox-api.dll
CHANGED
Binary file
|
ipex_llm/libs/gptneox.dll
CHANGED
Binary file
|
ipex_llm/libs/libbloom_avx.dll
CHANGED
Binary file
|
ipex_llm/libs/libbloom_vnni.dll
CHANGED
Binary file
|
ipex_llm/libs/libgptneox_avx.dll
CHANGED
Binary file
|
Binary file
|
ipex_llm/libs/libllama_avx.dll
CHANGED
Binary file
|
ipex_llm/libs/libllama_vnni.dll
CHANGED
Binary file
|
Binary file
|
Binary file
|
ipex_llm/libs/llama-api.dll
CHANGED
Binary file
|
ipex_llm/libs/llama.dll
CHANGED
Binary file
|
ipex_llm/libs/main-bloom.exe
CHANGED
Binary file
|
ipex_llm/libs/main-gptneox.exe
CHANGED
Binary file
|
ipex_llm/libs/main-llama.exe
CHANGED
Binary file
|
ipex_llm/libs/main-starcoder.exe
CHANGED
Binary file
|
ipex_llm/libs/pipeline.dll
CHANGED
Binary file
|
ipex_llm/libs/quantize-bloom.exe
CHANGED
Binary file
|
Binary file
|
Binary file
|
Binary file
|
ipex_llm/libs/quantize-llama.exe
CHANGED
Binary file
|
Binary file
|
Binary file
|
Binary file
|
ipex_llm/libs/starcoder-api.dll
CHANGED
Binary file
|
ipex_llm/libs/starcoder.dll
CHANGED
Binary file
|
ipex_llm/transformers/convert.py
CHANGED
@@ -1089,6 +1089,15 @@ def ggml_convert_low_bit(model, qtype, optimize_model=True,
|
|
1089
1089
|
embedding_qtype=None,
|
1090
1090
|
mixed_precision=False,
|
1091
1091
|
disable_optimize_pre=False):
|
1092
|
+
if qtype == ggml_tensor_qtype["sym_int4"] and torch.__version__ >= "2.6":
|
1093
|
+
logger.warning("sym_int4 is deprecated, use woq_int4 instead, "
|
1094
|
+
"if you are loading saved sym_int4 low bit model, "
|
1095
|
+
"please resaved it with woq_int4")
|
1096
|
+
qtype = ggml_tensor_qtype["woq_int4"]
|
1097
|
+
elif qtype == ggml_tensor_qtype["woq_int4"] and torch.__version__ < "2.6":
|
1098
|
+
logger.warning("woq_int4 is not supported with pytorch<2.6, "
|
1099
|
+
"use sym_int4 instead or use ipex-llm with pytorch>=2.6")
|
1100
|
+
qtype = ggml_tensor_qtype["sym_int4"]
|
1092
1101
|
if qtype in ggml_tensor_qtype.values():
|
1093
1102
|
index = list(ggml_tensor_qtype.values()).index(qtype)
|
1094
1103
|
logger.info(f"Converting the current model to "
|
ipex_llm/transformers/kv.py
CHANGED
@@ -22,7 +22,8 @@ import math
|
|
22
22
|
|
23
23
|
from .models.utils import (
|
24
24
|
init_fp8_kv_cache, append_fp8_kv_cache,
|
25
|
-
init_kv_cache, append_kv_cache, extend_kv_cache
|
25
|
+
init_kv_cache, append_kv_cache, extend_kv_cache,
|
26
|
+
init_unbalanced_fp8_kv_cache, append_unbalanced_fp8_kv_cache,
|
26
27
|
)
|
27
28
|
from typing import Optional, Dict, Tuple, Any, List
|
28
29
|
from transformers.cache_utils import DynamicCache
|
@@ -151,6 +152,55 @@ class DynamicNormalCache(DynamicCache):
|
|
151
152
|
return past_key_values
|
152
153
|
|
153
154
|
|
155
|
+
class DynamicUnbalancedFp8Cache(DynamicCache):
|
156
|
+
def __init__(self, num_hidden_layers: Optional[int] = None) -> None:
|
157
|
+
# ignore num_hidden_layers to fix transformers >= 4.45
|
158
|
+
super().__init__()
|
159
|
+
|
160
|
+
def update(
|
161
|
+
self,
|
162
|
+
key_states: torch.Tensor,
|
163
|
+
value_states: torch.Tensor,
|
164
|
+
layer_idx: int,
|
165
|
+
cache_kwargs: Optional[Dict[str, Any]]=None,
|
166
|
+
) -> Tuple[torch.Tensor, torch.Tensor]:
|
167
|
+
# fix converting empty DynamicCache in transformers >= 4.45
|
168
|
+
if key_states == []:
|
169
|
+
return key_states, value_states
|
170
|
+
|
171
|
+
batch_size, num_heads, seq_len, k_head_dim = key_states.shape
|
172
|
+
_, _, _, v_head_dim = value_states.shape
|
173
|
+
|
174
|
+
if layer_idx == 0:
|
175
|
+
if hasattr(self, "_seen_tokens"):
|
176
|
+
# 4.39 uses `_seen_tokens`
|
177
|
+
self._seen_tokens += seq_len
|
178
|
+
else:
|
179
|
+
# 4.37 uses `seen_tokens`
|
180
|
+
self.seen_tokens += seq_len
|
181
|
+
|
182
|
+
# Update the cache
|
183
|
+
if len(self.key_cache) <= layer_idx:
|
184
|
+
k_cache, v_cache = init_unbalanced_fp8_kv_cache(
|
185
|
+
batch_size, num_heads, seq_len, k_head_dim, v_head_dim,
|
186
|
+
device=key_states.device,
|
187
|
+
)
|
188
|
+
k_cache, v_cache = append_unbalanced_fp8_kv_cache(k_cache, v_cache,
|
189
|
+
key_states, value_states)
|
190
|
+
|
191
|
+
self.key_cache.append(k_cache)
|
192
|
+
self.value_cache.append(v_cache)
|
193
|
+
else:
|
194
|
+
k_cache = self.key_cache[layer_idx]
|
195
|
+
v_cache = self.value_cache[layer_idx]
|
196
|
+
k_cache, v_cache = append_unbalanced_fp8_kv_cache(k_cache, v_cache,
|
197
|
+
key_states, value_states)
|
198
|
+
self.key_cache[layer_idx] = k_cache
|
199
|
+
self.value_cache[layer_idx] = v_cache
|
200
|
+
|
201
|
+
return self.key_cache[layer_idx], self.value_cache[layer_idx]
|
202
|
+
|
203
|
+
|
154
204
|
# Copied from transformers.models.llama.modeling_llama.repeat_kv
|
155
205
|
def repeat_kv(hidden_states: torch.Tensor, n_rep: int) -> torch.Tensor:
|
156
206
|
"""
|
@@ -273,11 +273,11 @@ def scaled_dot_product_attention(query: torch.Tensor, key: torch.Tensor,
|
|
273
273
|
else:
|
274
274
|
attn_output = xe_addons.sdp_causal(query, key, value, mask, scale)
|
275
275
|
elif seq_length != kv_length and seq_length <= 32:
|
276
|
-
# todo: add scale support
|
276
|
+
# todo: add further scale support
|
277
277
|
if key.dtype == torch.uint8:
|
278
|
-
attn_output = xe_addons.sdp_fp8(query, key, value, mask)
|
278
|
+
attn_output = xe_addons.sdp_fp8(query, key, value, mask, scale)
|
279
279
|
else:
|
280
|
-
attn_output = xe_addons.sdp(query, key, value, mask)
|
280
|
+
attn_output = xe_addons.sdp(query, key, value, mask, scale)
|
281
281
|
else:
|
282
282
|
if key.dtype == torch.uint8:
|
283
283
|
attn_output = xe_addons.sdp_fp8_non_causal(query, key, value, mask, scale)
|
@@ -138,6 +138,49 @@ def append_fp8_kv_cache(k_cache, v_cache, key, value):
|
|
138
138
|
return new_k_cache, new_v_cache
|
139
139
|
|
140
140
|
|
141
|
+
def init_unbalanced_fp8_kv_cache(batch_size, num_heads, current_length,
|
142
|
+
k_head_dim, v_head_dim, device):
|
143
|
+
# for case which k head dim is different from v head dim
|
144
|
+
max_length = current_length + FP8_KV_ALLOC_LENGTH
|
145
|
+
|
146
|
+
k_cache_storage = torch.empty(batch_size, num_heads, max_length, k_head_dim,
|
147
|
+
dtype=torch.uint8, device=device)
|
148
|
+
k_cache = k_cache_storage.as_strided((batch_size, num_heads, 0, k_head_dim),
|
149
|
+
k_cache_storage.stride(), storage_offset=0)
|
150
|
+
|
151
|
+
v_cache_storage = torch.empty(batch_size, num_heads, max_length, v_head_dim,
|
152
|
+
dtype=torch.uint8, device=device)
|
153
|
+
v_cache = v_cache_storage.as_strided((batch_size, num_heads, 0, v_head_dim),
|
154
|
+
v_cache_storage.stride(), storage_offset=0)
|
155
|
+
return k_cache, v_cache
|
156
|
+
|
157
|
+
|
158
|
+
def append_unbalanced_fp8_kv_cache(k_cache, v_cache, key, value):
|
159
|
+
batch_size, num_heads, cur_length, k_head_dim = k_cache.shape
|
160
|
+
_, _, _, v_head_dim = v_cache.shape
|
161
|
+
new_length = cur_length + key.size(2)
|
162
|
+
new_k_size = (batch_size, num_heads, new_length, k_head_dim)
|
163
|
+
new_v_size = (batch_size, num_heads, new_length, v_head_dim)
|
164
|
+
|
165
|
+
if k_cache.stride(1) < new_length * k_cache.size(3):
|
166
|
+
new_k_cache, new_v_cache = init_unbalanced_fp8_kv_cache(batch_size, num_heads, new_length,
|
167
|
+
k_head_dim, v_head_dim, key.device)
|
168
|
+
new_k_cache = new_k_cache.as_strided(new_k_size, new_k_cache.stride(), storage_offset=0)
|
169
|
+
new_v_cache = new_v_cache.as_strided(new_v_size, new_v_cache.stride(), storage_offset=0)
|
170
|
+
new_k_cache[:, :, :cur_length, :] = k_cache
|
171
|
+
new_v_cache[:, :, :cur_length, :] = v_cache
|
172
|
+
else:
|
173
|
+
new_k_cache = k_cache.as_strided(new_k_size, k_cache.stride(), storage_offset=0)
|
174
|
+
new_v_cache = v_cache.as_strided(new_v_size, v_cache.stride(), storage_offset=0)
|
175
|
+
|
176
|
+
import xe_addons
|
177
|
+
xe_addons.quantize_key_value(key, value,
|
178
|
+
new_k_cache[:, :, cur_length:new_length, :],
|
179
|
+
new_v_cache[:, :, cur_length:new_length, :])
|
180
|
+
|
181
|
+
return new_k_cache, new_v_cache
|
182
|
+
|
183
|
+
|
141
184
|
def restore_fp8_kv_cache(k_cache, v_cache, dtype):
|
142
185
|
key_states = torch.empty(k_cache.shape, device=k_cache.device, dtype=dtype)
|
143
186
|
value_states = torch.empty(v_cache.shape, device=v_cache.device, dtype=dtype)
|
@@ -1,6 +1,6 @@
|
|
1
1
|
Metadata-Version: 2.1
|
2
2
|
Name: ipex-llm
|
3
|
-
Version: 2.3.
|
3
|
+
Version: 2.3.0b20250414
|
4
4
|
Summary: Large Language Model Develop Toolkit
|
5
5
|
Home-page: https://github.com/intel-analytics/ipex-llm
|
6
6
|
Author: BigDL Authors
|
@@ -27,7 +27,7 @@ Requires-Dist: intel-openmp ; (platform_machine == "x86_64" or platform_machine
|
|
27
27
|
Requires-Dist: torch ==2.1.2+cpu ; (platform_system == "Linux") and extra == 'all'
|
28
28
|
Requires-Dist: torch ==2.1.2 ; (platform_system == "Windows") and extra == 'all'
|
29
29
|
Provides-Extra: cpp
|
30
|
-
Requires-Dist: bigdl-core-cpp ==2.7.
|
30
|
+
Requires-Dist: bigdl-core-cpp ==2.7.0b20250414 ; extra == 'cpp'
|
31
31
|
Requires-Dist: setuptools ; extra == 'cpp'
|
32
32
|
Requires-Dist: onednn-devel ==2025.0.1 ; (platform_system == "Windows") and extra == 'cpp'
|
33
33
|
Requires-Dist: onednn ==2025.0.1 ; (platform_system == "Windows") and extra == 'cpp'
|
@@ -60,7 +60,7 @@ Requires-Dist: transformers ==4.40.0 ; extra == 'npu'
|
|
60
60
|
Requires-Dist: intel-openmp ; (platform_machine == "x86_64" or platform_machine == "AMD64") and extra == 'npu'
|
61
61
|
Requires-Dist: torch ==2.1.2+cpu ; (platform_system == "Linux") and extra == 'npu'
|
62
62
|
Requires-Dist: torch ==2.1.2 ; (platform_system == "Windows") and extra == 'npu'
|
63
|
-
Requires-Dist: bigdl-core-npu ==2.7.
|
63
|
+
Requires-Dist: bigdl-core-npu ==2.7.0b20250414 ; (platform_system == "Windows") and extra == 'npu'
|
64
64
|
Provides-Extra: serving
|
65
65
|
Requires-Dist: py-cpuinfo ; extra == 'serving'
|
66
66
|
Requires-Dist: fschat[model_worker,webui] ==0.2.36 ; extra == 'serving'
|
@@ -80,9 +80,9 @@ Requires-Dist: setuptools <70.0.0 ; extra == 'xpu'
|
|
80
80
|
Requires-Dist: torch ==2.1.0a0 ; extra == 'xpu'
|
81
81
|
Requires-Dist: torchvision ==0.16.0a0 ; extra == 'xpu'
|
82
82
|
Requires-Dist: intel-extension-for-pytorch ==2.1.10+xpu ; extra == 'xpu'
|
83
|
-
Requires-Dist: bigdl-core-xe-21 ==2.7.
|
84
|
-
Requires-Dist: bigdl-core-xe-batch-21 ==2.7.
|
85
|
-
Requires-Dist: bigdl-core-xe-addons-21 ==2.7.
|
83
|
+
Requires-Dist: bigdl-core-xe-21 ==2.7.0b20250414 ; extra == 'xpu'
|
84
|
+
Requires-Dist: bigdl-core-xe-batch-21 ==2.7.0b20250414 ; extra == 'xpu'
|
85
|
+
Requires-Dist: bigdl-core-xe-addons-21 ==2.7.0b20250414 ; extra == 'xpu'
|
86
86
|
Provides-Extra: xpu-2-1
|
87
87
|
Requires-Dist: py-cpuinfo ; extra == 'xpu-2-1'
|
88
88
|
Requires-Dist: protobuf ; extra == 'xpu-2-1'
|
@@ -97,9 +97,9 @@ Requires-Dist: setuptools <70.0.0 ; extra == 'xpu-2-1'
|
|
97
97
|
Requires-Dist: torch ==2.1.0a0 ; extra == 'xpu-2-1'
|
98
98
|
Requires-Dist: torchvision ==0.16.0a0 ; extra == 'xpu-2-1'
|
99
99
|
Requires-Dist: intel-extension-for-pytorch ==2.1.10+xpu ; extra == 'xpu-2-1'
|
100
|
-
Requires-Dist: bigdl-core-xe-21 ==2.7.
|
101
|
-
Requires-Dist: bigdl-core-xe-batch-21 ==2.7.
|
102
|
-
Requires-Dist: bigdl-core-xe-addons-21 ==2.7.
|
100
|
+
Requires-Dist: bigdl-core-xe-21 ==2.7.0b20250414 ; extra == 'xpu-2-1'
|
101
|
+
Requires-Dist: bigdl-core-xe-batch-21 ==2.7.0b20250414 ; extra == 'xpu-2-1'
|
102
|
+
Requires-Dist: bigdl-core-xe-addons-21 ==2.7.0b20250414 ; extra == 'xpu-2-1'
|
103
103
|
Requires-Dist: intel-openmp ; (platform_machine == "x86_64" or platform_machine == "AMD64") and extra == 'xpu-2-1'
|
104
104
|
Requires-Dist: dpcpp-cpp-rt ==2024.0.2 ; (platform_system == "Windows") and extra == 'xpu-2-1'
|
105
105
|
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'
|
|
117
117
|
Requires-Dist: torch ==2.6.0+xpu ; extra == 'xpu-2-6'
|
118
118
|
Requires-Dist: torchvision ==0.21.0+xpu ; extra == 'xpu-2-6'
|
119
119
|
Requires-Dist: torchaudio ==2.6.0+xpu ; extra == 'xpu-2-6'
|
120
|
-
Requires-Dist: bigdl-core-xe-all ==2.7.
|
120
|
+
Requires-Dist: bigdl-core-xe-all ==2.7.0b20250414 ; extra == 'xpu-2-6'
|
121
121
|
Requires-Dist: onednn-devel ==2025.0.1 ; extra == 'xpu-2-6'
|
122
122
|
Requires-Dist: onednn ==2025.0.1 ; extra == 'xpu-2-6'
|
123
123
|
Requires-Dist: dpcpp-cpp-rt ==2025.0.2 ; extra == 'xpu-2-6'
|
@@ -132,7 +132,7 @@ Requires-Dist: tokenizers ==0.15.2 ; extra == 'xpu-2-6-arl'
|
|
132
132
|
Requires-Dist: accelerate ==0.23.0 ; extra == 'xpu-2-6-arl'
|
133
133
|
Requires-Dist: tabulate ; extra == 'xpu-2-6-arl'
|
134
134
|
Requires-Dist: setuptools ; extra == 'xpu-2-6-arl'
|
135
|
-
Requires-Dist: bigdl-core-xe-all ==2.7.
|
135
|
+
Requires-Dist: bigdl-core-xe-all ==2.7.0b20250414 ; extra == 'xpu-2-6-arl'
|
136
136
|
Requires-Dist: onednn-devel ==2025.0.1 ; extra == 'xpu-2-6-arl'
|
137
137
|
Requires-Dist: onednn ==2025.0.1 ; extra == 'xpu-2-6-arl'
|
138
138
|
Requires-Dist: dpcpp-cpp-rt ==2025.0.2 ; extra == 'xpu-2-6-arl'
|
@@ -155,9 +155,9 @@ Requires-Dist: tokenizers ==0.15.2 ; extra == 'xpu-arc'
|
|
155
155
|
Requires-Dist: accelerate ==0.23.0 ; extra == 'xpu-arc'
|
156
156
|
Requires-Dist: tabulate ; extra == 'xpu-arc'
|
157
157
|
Requires-Dist: setuptools ; extra == 'xpu-arc'
|
158
|
-
Requires-Dist: bigdl-core-xe-23 ==2.7.
|
159
|
-
Requires-Dist: bigdl-core-xe-batch-23 ==2.7.
|
160
|
-
Requires-Dist: bigdl-core-xe-addons-23 ==2.7.
|
158
|
+
Requires-Dist: bigdl-core-xe-23 ==2.7.0b20250414 ; extra == 'xpu-arc'
|
159
|
+
Requires-Dist: bigdl-core-xe-batch-23 ==2.7.0b20250414 ; extra == 'xpu-arc'
|
160
|
+
Requires-Dist: bigdl-core-xe-addons-23 ==2.7.0b20250414 ; extra == 'xpu-arc'
|
161
161
|
Requires-Dist: intel-openmp ; (platform_machine == "x86_64" or platform_machine == "AMD64") and extra == 'xpu-arc'
|
162
162
|
Requires-Dist: torch ==2.3.1+cxx11.abi ; (platform_system == "Linux") and extra == 'xpu-arc'
|
163
163
|
Requires-Dist: torchvision ==0.18.1+cxx11.abi ; (platform_system == "Linux") and extra == 'xpu-arc'
|
@@ -178,9 +178,9 @@ Requires-Dist: tokenizers ==0.15.2 ; extra == 'xpu-arl'
|
|
178
178
|
Requires-Dist: accelerate ==0.23.0 ; extra == 'xpu-arl'
|
179
179
|
Requires-Dist: tabulate ; extra == 'xpu-arl'
|
180
180
|
Requires-Dist: setuptools ; extra == 'xpu-arl'
|
181
|
-
Requires-Dist: bigdl-core-xe-23 ==2.7.
|
182
|
-
Requires-Dist: bigdl-core-xe-batch-23 ==2.7.
|
183
|
-
Requires-Dist: bigdl-core-xe-addons-23 ==2.7.
|
181
|
+
Requires-Dist: bigdl-core-xe-23 ==2.7.0b20250414 ; extra == 'xpu-arl'
|
182
|
+
Requires-Dist: bigdl-core-xe-batch-23 ==2.7.0b20250414 ; extra == 'xpu-arl'
|
183
|
+
Requires-Dist: bigdl-core-xe-addons-23 ==2.7.0b20250414 ; extra == 'xpu-arl'
|
184
184
|
Requires-Dist: intel-openmp ; (platform_machine == "x86_64" or platform_machine == "AMD64") and extra == 'xpu-arl'
|
185
185
|
Requires-Dist: torch ==2.3.1+cxx11.abi ; (platform_system == "Linux") and extra == 'xpu-arl'
|
186
186
|
Requires-Dist: torchvision ==0.18.1+cxx11.abi ; (platform_system == "Linux") and extra == 'xpu-arl'
|
@@ -201,9 +201,9 @@ Requires-Dist: tokenizers ==0.15.2 ; extra == 'xpu-lnl'
|
|
201
201
|
Requires-Dist: accelerate ==0.23.0 ; extra == 'xpu-lnl'
|
202
202
|
Requires-Dist: tabulate ; extra == 'xpu-lnl'
|
203
203
|
Requires-Dist: setuptools ; extra == 'xpu-lnl'
|
204
|
-
Requires-Dist: bigdl-core-xe-23 ==2.7.
|
205
|
-
Requires-Dist: bigdl-core-xe-batch-23 ==2.7.
|
206
|
-
Requires-Dist: bigdl-core-xe-addons-23 ==2.7.
|
204
|
+
Requires-Dist: bigdl-core-xe-23 ==2.7.0b20250414 ; extra == 'xpu-lnl'
|
205
|
+
Requires-Dist: bigdl-core-xe-batch-23 ==2.7.0b20250414 ; extra == 'xpu-lnl'
|
206
|
+
Requires-Dist: bigdl-core-xe-addons-23 ==2.7.0b20250414 ; extra == 'xpu-lnl'
|
207
207
|
Requires-Dist: intel-openmp ; (platform_machine == "x86_64" or platform_machine == "AMD64") and extra == 'xpu-lnl'
|
208
208
|
Requires-Dist: torch ==2.3.1+cxx11.abi ; (platform_system == "Linux") and extra == 'xpu-lnl'
|
209
209
|
Requires-Dist: torchvision ==0.18.1+cxx11.abi ; (platform_system == "Linux") and extra == 'xpu-lnl'
|
@@ -41,35 +41,35 @@ ipex_llm/langchain/llms/transformerspipelinellm.py,sha256=vm522YPPwWxxAPVvQBtxRf
|
|
41
41
|
ipex_llm/langchain/vllm/__init__.py,sha256=T-EbRT6GJ_8RCu-iLmSzcftOimXSPQf2d5X72AUAy2Y,874
|
42
42
|
ipex_llm/langchain/vllm/vllm.py,sha256=6dxc-ZISZQrJilEa_HA827l75Dv9rcHpY_G6FdJ8BVs,7793
|
43
43
|
ipex_llm/libs/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
|
44
|
-
ipex_llm/libs/bloom-api.dll,sha256=
|
45
|
-
ipex_llm/libs/bloom.dll,sha256=
|
46
|
-
ipex_llm/libs/gptneox-api.dll,sha256=
|
47
|
-
ipex_llm/libs/gptneox.dll,sha256=
|
48
|
-
ipex_llm/libs/libbloom_avx.dll,sha256=
|
49
|
-
ipex_llm/libs/libbloom_vnni.dll,sha256=
|
50
|
-
ipex_llm/libs/libgptneox_avx.dll,sha256=
|
51
|
-
ipex_llm/libs/libgptneox_vnni.dll,sha256=
|
52
|
-
ipex_llm/libs/libllama_avx.dll,sha256=
|
53
|
-
ipex_llm/libs/libllama_vnni.dll,sha256=
|
54
|
-
ipex_llm/libs/libstarcoder_avx.dll,sha256=
|
55
|
-
ipex_llm/libs/libstarcoder_vnni.dll,sha256=
|
56
|
-
ipex_llm/libs/llama-api.dll,sha256=
|
57
|
-
ipex_llm/libs/llama.dll,sha256=
|
58
|
-
ipex_llm/libs/main-bloom.exe,sha256
|
59
|
-
ipex_llm/libs/main-gptneox.exe,sha256=
|
60
|
-
ipex_llm/libs/main-llama.exe,sha256=
|
61
|
-
ipex_llm/libs/main-starcoder.exe,sha256=
|
62
|
-
ipex_llm/libs/pipeline.dll,sha256=
|
63
|
-
ipex_llm/libs/quantize-bloom.exe,sha256=
|
64
|
-
ipex_llm/libs/quantize-bloom_vnni.exe,sha256=
|
65
|
-
ipex_llm/libs/quantize-gptneox.exe,sha256=
|
66
|
-
ipex_llm/libs/quantize-gptneox_vnni.exe,sha256=
|
67
|
-
ipex_llm/libs/quantize-llama.exe,sha256=
|
68
|
-
ipex_llm/libs/quantize-llama_vnni.exe,sha256=
|
69
|
-
ipex_llm/libs/quantize-starcoder.exe,sha256=
|
70
|
-
ipex_llm/libs/quantize-starcoder_vnni.exe,sha256=
|
71
|
-
ipex_llm/libs/starcoder-api.dll,sha256=
|
72
|
-
ipex_llm/libs/starcoder.dll,sha256=
|
44
|
+
ipex_llm/libs/bloom-api.dll,sha256=RQH35fsCjbpJAqZ6vqpzIvurExHg77Dm_j9SF1eL8DQ,36352
|
45
|
+
ipex_llm/libs/bloom.dll,sha256=riUIPG6kaX1CeCirSMMuPjnadxQrm89t3s8MCjW9VDg,507904
|
46
|
+
ipex_llm/libs/gptneox-api.dll,sha256=3Z21zemTxGTPtZyLrJQlwaphhjnEUH_nD2_azE4GHjU,24576
|
47
|
+
ipex_llm/libs/gptneox.dll,sha256=Y91HtjmiOHXOgttQod7KG9f9KkCdU9Jbu_V_2Ncz1Ic,568320
|
48
|
+
ipex_llm/libs/libbloom_avx.dll,sha256=siYYEY3keJNXxD1q1qe2EcX0Cz6ezyr4E1jD4kQ8DMI,536576
|
49
|
+
ipex_llm/libs/libbloom_vnni.dll,sha256=Q0Wtp6u-xcTW_TTsUHqn0X7353W_ere_k0GwsI0QVhE,508416
|
50
|
+
ipex_llm/libs/libgptneox_avx.dll,sha256=AtpJopCdHu0RAVPyunHsQ9qQWLiJMIE7ba4gme7nC_k,596992
|
51
|
+
ipex_llm/libs/libgptneox_vnni.dll,sha256=Gt4tFPro-zNdU4S22k96Tds-tVCwRgccuu3RRwWCsJ4,568832
|
52
|
+
ipex_llm/libs/libllama_avx.dll,sha256=BrqmZm6ASzZ5hPiu7cjJKLxs2pnTUDS8vlaVwXtglKc,591360
|
53
|
+
ipex_llm/libs/libllama_vnni.dll,sha256=44HGXEJgJYJ6kS_dngqYR9VBFXRiYEooxo4K_xLLvyA,563200
|
54
|
+
ipex_llm/libs/libstarcoder_avx.dll,sha256=XkCQzKH7Yg3n1hLbYhNaCgRGa_vEjT4G048i7FqiPFI,627712
|
55
|
+
ipex_llm/libs/libstarcoder_vnni.dll,sha256=hhggl9l81TYcEGpZymdRkAtGeyrgSUqdHR3Xn-yfARE,599552
|
56
|
+
ipex_llm/libs/llama-api.dll,sha256=k50ow33IgEjCfmyhaeLEK77JRVhu8UacJGiVCdj56lI,25600
|
57
|
+
ipex_llm/libs/llama.dll,sha256=c81XUKbRl5RRpvad8qji61I-TW2N2WIA7SPWTmJ6c7I,562688
|
58
|
+
ipex_llm/libs/main-bloom.exe,sha256=-Pg9G61YLFlBuMloBWPZ0BwuyizTdP9Gmp9H8lBmlIE,103424
|
59
|
+
ipex_llm/libs/main-gptneox.exe,sha256=FqJFdSuRLzHzISAS3Fq-hpNaD5AzByRhidVJGvx8XsM,98816
|
60
|
+
ipex_llm/libs/main-llama.exe,sha256=gCMvDvxMGMcTITQVB7JwzDQl8hWiS7cMedpd5qH1Yso,99840
|
61
|
+
ipex_llm/libs/main-starcoder.exe,sha256=Hp8bVAB3gkleaTILUFP373bj_-FfZBhgY98ifF4Seg8,157696
|
62
|
+
ipex_llm/libs/pipeline.dll,sha256=klcRphRZj-Ver9bb-3YarFz1Aofx2eH_9ZFVj6005IY,73216
|
63
|
+
ipex_llm/libs/quantize-bloom.exe,sha256=1RCyEGDHqMgEV3-Z5whNWeoPH2pbmIuoNdFxwwWVzBM,126464
|
64
|
+
ipex_llm/libs/quantize-bloom_vnni.exe,sha256=9B3eXy6FuOxdw4R7226GWwZdYRkA_0wrrdoEB4iRvL4,128000
|
65
|
+
ipex_llm/libs/quantize-gptneox.exe,sha256=ho-9GbDDXLe6OCZUQXLLKQBaxndRs0uoIp7blm7mkX4,104448
|
66
|
+
ipex_llm/libs/quantize-gptneox_vnni.exe,sha256=YxBxpYZGZASjRE0AvfhWPFN5y07J7b_p7uQ6KBfcgoA,104960
|
67
|
+
ipex_llm/libs/quantize-llama.exe,sha256=N4t0zFvqDydwEF-ziXUfiWtoBUU_0InDyBgufaUGXzU,110080
|
68
|
+
ipex_llm/libs/quantize-llama_vnni.exe,sha256=23i9bbF_ezU4WLc2vSjdzPP2aTMdPyxiQMFnSAaLk1A,110592
|
69
|
+
ipex_llm/libs/quantize-starcoder.exe,sha256=DYm5NLxvHgpeldjdK8e0LcvgnLauLo5gz78CGVEKQTI,127488
|
70
|
+
ipex_llm/libs/quantize-starcoder_vnni.exe,sha256=uoMxJlhpV-BzhANKO6IJls6iIP7KG03HIq8KxZXp-A8,128512
|
71
|
+
ipex_llm/libs/starcoder-api.dll,sha256=wKmuMWuK2rlCEPrSLIaaK_ylaECQEiU730vSyurxBG8,21504
|
72
|
+
ipex_llm/libs/starcoder.dll,sha256=rWkn029EPDSsEjlRu18jLq_0fC4b6axEeHjgoezGSEM,599040
|
73
73
|
ipex_llm/llamaindex/__init__.py,sha256=T-EbRT6GJ_8RCu-iLmSzcftOimXSPQf2d5X72AUAy2Y,874
|
74
74
|
ipex_llm/llamaindex/llms/__init__.py,sha256=KP1lEdGqDuxPoxL1ZSH25Pm2kKMPJBWUTLR0ckSLMIU,1139
|
75
75
|
ipex_llm/llamaindex/llms/bigdlllm.py,sha256=FQBzq1KOjfc6uofTXAha3O7TqpJkNfOFepXQmOVlbnI,26314
|
@@ -87,10 +87,10 @@ ipex_llm/serving/fastchat/tgi_api_protocol.py,sha256=brT3k3-V0NJrU4fRqUwWjC0O3iO
|
|
87
87
|
ipex_llm/serving/fastchat/tgi_api_server.py,sha256=agNTAEiZPSuj3dEdIdYKwkoY0cXOUDX06DiM9VP2knQ,24418
|
88
88
|
ipex_llm/serving/fastchat/vllm_worker.py,sha256=ZLz2Q9GxJO6r_LOiP6epgCRjBGk-K4EB1SNEWSJp5DA,11091
|
89
89
|
ipex_llm/transformers/__init__.py,sha256=BreA3EY6hkNq0rVixb_sUuTLzMrcWXTt3yRsshCPHHQ,1214
|
90
|
-
ipex_llm/transformers/convert.py,sha256=
|
90
|
+
ipex_llm/transformers/convert.py,sha256=niCLZiEQbiMz8UkfYDNwYl2hE1caj93FUR1qjgVh9ZM,107538
|
91
91
|
ipex_llm/transformers/convert_ipex.py,sha256=_nSnUTQy-yfkKaqGdqnBdWztZf3NGmnbZ0TKaDrF4X4,14617
|
92
92
|
ipex_llm/transformers/embedding.py,sha256=bdgk59DvD4ZZyxRzewXOR7g56nThgO6uhIwk8QL7f-s,9299
|
93
|
-
ipex_llm/transformers/kv.py,sha256=
|
93
|
+
ipex_llm/transformers/kv.py,sha256=src_HcVDKFwQ1V8hdTrFQw5RIwUewM9VOR47GVTPJG4,21187
|
94
94
|
ipex_llm/transformers/lisa.py,sha256=F5WxbtXQ7RdKulj83h_2DnEIgKiKGZf7zvOmg6QBl2s,3289
|
95
95
|
ipex_llm/transformers/loader.py,sha256=c9qfJSC6-in-mkd-iKb1igk3nHWUYS3QtyH2cOazmKc,6825
|
96
96
|
ipex_llm/transformers/lookup.py,sha256=b6OlZ9OV10R9qeWw8mVryVpDxszkjwLkldvi7GPMJY8,19614
|
@@ -146,7 +146,7 @@ ipex_llm/transformers/models/chatglm.py,sha256=UHai1t2AUtGmF765_eHF8LUMVQzp_oCBx
|
|
146
146
|
ipex_llm/transformers/models/chatglm2.py,sha256=KyAIX7zGVQDQuwwM3QMBNWZbTeMHEzKUIgAryT0voHc,14933
|
147
147
|
ipex_llm/transformers/models/chatglm4.py,sha256=QvUehdaCePB3MNHyWg3dneDxmjtBdxYeKUyQUVcsgfM,16886
|
148
148
|
ipex_llm/transformers/models/chatglm4v.py,sha256=Ba9Xtzwtzk_rzg5khGqDrlHfJsDwc5YcM5_yPoord7o,13324
|
149
|
-
ipex_llm/transformers/models/common.py,sha256=
|
149
|
+
ipex_llm/transformers/models/common.py,sha256=LVA9nL_qJ61NEkEn9T985PjrrWPGpDTCALknH4Qv5aw,13040
|
150
150
|
ipex_llm/transformers/models/decilm.py,sha256=P-PBuDPf07GvKggLwJx_wPwIn6esN3rX8ai2JxRuZmE,5246
|
151
151
|
ipex_llm/transformers/models/deepseek.py,sha256=w6tGeyJ9joD7lQBiZ6A01Z00g8hAXC1N2yGtJh8kyuk,13096
|
152
152
|
ipex_llm/transformers/models/deepseek_v3.py,sha256=CTgwIKQlUPlUCbOxc9Id5GapWkXOP6pMtkguYrWpCio,10003
|
@@ -180,7 +180,7 @@ ipex_llm/transformers/models/rwkv5.py,sha256=OkRNj1pCAZg1z2Fw-I0DEnxLEdZyPeRSQ6m
|
|
180
180
|
ipex_llm/transformers/models/sd.py,sha256=VvHV5u-0k2MgHu3NL9113hPj7DgfxqctuKzEEeNfRDU,5981
|
181
181
|
ipex_llm/transformers/models/stablelm.py,sha256=fj-XtOnR6kggnFUQTMPCOOzolkPztN06WAv8QW-XRnI,7054
|
182
182
|
ipex_llm/transformers/models/starcoder2.py,sha256=ONKvD7JCkRM0DI-R56x28QFBJ7CjD5hOZBQ_3WfOcNk,6626
|
183
|
-
ipex_llm/transformers/models/utils.py,sha256=
|
183
|
+
ipex_llm/transformers/models/utils.py,sha256=Rj7QK1s3QxsTT4HzGNR00Q7izG5xYiZEX29YoQIM1tA,16998
|
184
184
|
ipex_llm/transformers/models/whisper.py,sha256=ju3WP8Eq-KvD7kb3Qy51r4FOfSX3NBxfp5RBcq__gzc,4241
|
185
185
|
ipex_llm/transformers/models/yuan.py,sha256=JYAn_ZaSGK0NBJLEIxCACfAq084a66GFJkdd5NbpmMA,7732
|
186
186
|
ipex_llm/transformers/npu_models/__init__.py,sha256=ulEUGLjaP48LCrVeury3UxLjXxKzRi0UpSG4bYu-7f8,585
|
@@ -256,11 +256,11 @@ ipex_llm/vllm/xpu/engine/__init__.py,sha256=pY_CpyuZd72fr6s32ejeKHKFW0K4vUU2rzZj
|
|
256
256
|
ipex_llm/vllm/xpu/engine/engine.py,sha256=NvCMbp0X8NVrOqbwm4FTvXOptTRLzu9jQsy37ZHnTk8,9493
|
257
257
|
ipex_llm/vllm/xpu/entrypoints/openai/api_server.py,sha256=IjiSze9vzBCAkLu_VwIcJwuO1jyFna7DLrj6aSL7RaQ,35220
|
258
258
|
ipex_llm/vllm/xpu/entrypoints/openai/cli_args.py,sha256=hB398yYtKauASRzevctScdbFIjiiSGMAe1bwEuIHrhY,10893
|
259
|
-
ipex_llm-2.3.
|
260
|
-
ipex_llm-2.3.
|
261
|
-
ipex_llm-2.3.
|
262
|
-
ipex_llm-2.3.
|
263
|
-
ipex_llm-2.3.
|
264
|
-
ipex_llm-2.3.
|
265
|
-
ipex_llm-2.3.
|
266
|
-
ipex_llm-2.3.
|
259
|
+
ipex_llm-2.3.0b20250414.data/scripts/ipex-llm-init.bat,sha256=HPtCYuDYwEatq7dAwOvdfVcHYCpAVdbj75K1qh0vQek,2578
|
260
|
+
ipex_llm-2.3.0b20250414.data/scripts/llm-chat.ps1,sha256=6qrs-hGVAV8IKh7Jx8nq_XrnZcjd7qGU5wndArM7Yag,2769
|
261
|
+
ipex_llm-2.3.0b20250414.data/scripts/llm-cli.ps1,sha256=3qBtTLs_EjYDnM8YyCpJhzLnGCKTEGssu9UNqfkjVXs,3009
|
262
|
+
ipex_llm-2.3.0b20250414.dist-info/METADATA,sha256=4kiLch6GAWOIkpZGeb-YiWgSUq8Linmj_qrwVY-hZew,13917
|
263
|
+
ipex_llm-2.3.0b20250414.dist-info/WHEEL,sha256=6iYPr8vTHsyDK75jr9X0V3I9wPSVmtwr_8fdATBciGk,98
|
264
|
+
ipex_llm-2.3.0b20250414.dist-info/entry_points.txt,sha256=TiUyBB2MRmfF3ko-pyAEzqeBCRnyhu27bNOAsWPp3e8,61
|
265
|
+
ipex_llm-2.3.0b20250414.dist-info/top_level.txt,sha256=CGCMHM-SyqUabU4h8RqJ2KTYckQUO3LvIWwmUQ6Qbzw,9
|
266
|
+
ipex_llm-2.3.0b20250414.dist-info/RECORD,,
|
File without changes
|
File without changes
|
File without changes
|
File without changes
|
File without changes
|
File without changes
|