TommyAI1986 commited on
Commit
cb05fc2
·
verified ·
1 Parent(s): f87bfd3

Add Qwen3 Embedding 0.6B (Core AI)

Browse files
.gitattributes CHANGED
@@ -33,3 +33,5 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
33
  *.zip filter=lfs diff=lfs merge=lfs -text
34
  *.zst filter=lfs diff=lfs merge=lfs -text
35
  *tfevents* filter=lfs diff=lfs merge=lfs -text
 
 
 
33
  *.zip filter=lfs diff=lfs merge=lfs -text
34
  *.zst filter=lfs diff=lfs merge=lfs -text
35
  *tfevents* filter=lfs diff=lfs merge=lfs -text
36
+ qwen3-embedding-0.6b/qwen3-embedding-0.6b_float16_s512_static.aimodel/main.mlirb filter=lfs diff=lfs merge=lfs -text
37
+ qwen3-embedding-0.6b/tokenizer/tokenizer.json filter=lfs diff=lfs merge=lfs -text
qwen3-embedding-0.6b/SHA256SUMS ADDED
@@ -0,0 +1,13 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ 3ec8c2dad0732aba83ac9359d4a4fba7ddeb63c8ba422959051150c0a187fee6 SOURCE-README.md
2
+ 51a2756b507ebaf2425f4ba4c06084866c8198e627c3ff6f1125158d2637e551 artifact-manifest.json
3
+ 72b89b1e17540462f4d8b80e4225501dc69f415dd395581c5473bb1a49585e59 export-metadata.json
4
+ 0c30b2fa619b45e3344eff10add843759f280e8b9b53c0b647336fa2a930f710 installation-receipt.json
5
+ ab2f5f89a0e78060c01869f846b00e945bf9a7714ef53df530596d0b26833a84 parity-report.json
6
+ ec1bf7feb6e7f8bdc1e24ee0ecf51bffb769f9de8dcb69f2dbfcbcdd2e4e8a84 qwen3-embedding-0.6b_float16_s512_static.aimodel/main.hash
7
+ 7731c0dd39fae320addd0604a26cd035c6316531eaa77c03470e99db1382d46c qwen3-embedding-0.6b_float16_s512_static.aimodel/main.mlirb
8
+ 7f00643721d3283ffe9a14b82582ca2db7c9b66cace7565cfed3b21e6ccd4e69 qwen3-embedding-0.6b_float16_s512_static.aimodel/metadata.json
9
+ ca352bb63c7fa4815ae2ac15e1bf5caee99fcf8d8414512199e6a6056590b771 reference.json
10
+ 87a2728cb8dc9fe424d624542f6060ec05a1d285ebbec578bb078900e33396b5 tokenizer/chat_template.jinja
11
+ a1b5ea7c0de53bc91ffec88f5ad6575d2fd847cf25a2c7bfaadd13d8e2dcefb0 tokenizer/tokenizer.json
12
+ fde77b035fb8a7fbaddebd91d7183076385e21fa67cb3fb8b78e9718fd9ba974 tokenizer/tokenizer_config.json
13
+ 1f8b0e716d0839e36ebabce172f26a96f8e5a059d891f1c69e9470ca7054fb64 validation-report.json
qwen3-embedding-0.6b/SOURCE-README.md ADDED
@@ -0,0 +1,268 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ base_model:
4
+ - Qwen/Qwen3-0.6B-Base
5
+ tags:
6
+ - transformers
7
+ - sentence-transformers
8
+ - sentence-similarity
9
+ - feature-extraction
10
+ ---
11
+ # Qwen3-Embedding-0.6B
12
+
13
+ <p align="center">
14
+ <img src="https://qianwen-res.oss-accelerate-overseas.aliyuncs.com/logo_qwen3.png" width="400"/>
15
+ <p>
16
+
17
+ ## Highlights
18
+
19
+ The Qwen3 Embedding model series is the latest proprietary model of the Qwen family, specifically designed for text embedding and ranking tasks. Building upon the dense foundational models of the Qwen3 series, it provides a comprehensive range of text embeddings and reranking models in various sizes (0.6B, 4B, and 8B). This series inherits the exceptional multilingual capabilities, long-text understanding, and reasoning skills of its foundational model. The Qwen3 Embedding series represents significant advancements in multiple text embedding and ranking tasks, including text retrieval, code retrieval, text classification, text clustering, and bitext mining.
20
+
21
+ **Exceptional Versatility**: The embedding model has achieved state-of-the-art performance across a wide range of downstream application evaluations. The 8B size embedding model ranks **No.1** in the MTEB multilingual leaderboard (as of June 5, 2025, score **70.58**), while the reranking model excels in various text retrieval scenarios.
22
+
23
+ **Comprehensive Flexibility**: The Qwen3 Embedding series offers a full spectrum of sizes (from 0.6B to 8B) for both embedding and reranking models, catering to diverse use cases that prioritize efficiency and effectiveness. Developers can seamlessly combine these two modules. Additionally, the embedding model allows for flexible vector definitions across all dimensions, and both embedding and reranking models support user-defined instructions to enhance performance for specific tasks, languages, or scenarios.
24
+
25
+ **Multilingual Capability**: The Qwen3 Embedding series offer support for over 100 languages, thanks to the multilingual capabilites of Qwen3 models. This includes various programming languages, and provides robust multilingual, cross-lingual, and code retrieval capabilities.
26
+ ## Model Overview
27
+
28
+ **Qwen3-Embedding-0.6B** has the following features:
29
+
30
+ - Model Type: Text Embedding
31
+ - Supported Languages: 100+ Languages
32
+ - Number of Paramaters: 0.6B
33
+ - Context Length: 32k
34
+ - Embedding Dimension: Up to 1024, supports user-defined output dimensions ranging from 32 to 1024
35
+
36
+ For more details, including benchmark evaluation, hardware requirements, and inference performance, please refer to our [blog](https://qwenlm.github.io/blog/qwen3-embedding/), [GitHub](https://github.com/QwenLM/Qwen3-Embedding).
37
+
38
+ ## Qwen3 Embedding Series Model list
39
+
40
+ | Model Type | Models | Size | Layers | Sequence Length | Embedding Dimension | MRL Support | Instruction Aware |
41
+ |------------------|----------------------|------|--------|-----------------|---------------------|-------------|----------------|
42
+ | Text Embedding | [Qwen3-Embedding-0.6B](https://huggingface.co/Qwen/Qwen3-Embedding-0.6B) | 0.6B | 28 | 32K | 1024 | Yes | Yes |
43
+ | Text Embedding | [Qwen3-Embedding-4B](https://huggingface.co/Qwen/Qwen3-Embedding-4B) | 4B | 36 | 32K | 2560 | Yes | Yes |
44
+ | Text Embedding | [Qwen3-Embedding-8B](https://huggingface.co/Qwen/Qwen3-Embedding-8B) | 8B | 36 | 32K | 4096 | Yes | Yes |
45
+ | Text Reranking | [Qwen3-Reranker-0.6B](https://huggingface.co/Qwen/Qwen3-Reranker-0.6B) | 0.6B | 28 | 32K | - | - | Yes |
46
+ | Text Reranking | [Qwen3-Reranker-4B](https://huggingface.co/Qwen/Qwen3-Reranker-4B) | 4B | 36 | 32K | - | - | Yes |
47
+ | Text Reranking | [Qwen3-Reranker-8B](https://huggingface.co/Qwen/Qwen3-Reranker-8B) | 8B | 36 | 32K | - | - | Yes |
48
+
49
+ > **Note**:
50
+ > - `MRL Support` indicates whether the embedding model supports custom dimensions for the final embedding.
51
+ > - `Instruction Aware` notes whether the embedding or reranking model supports customizing the input instruction according to different tasks.
52
+ > - Our evaluation indicates that, for most downstream tasks, using instructions (instruct) typically yields an improvement of 1% to 5% compared to not using them. Therefore, we recommend that developers create tailored instructions specific to their tasks and scenarios. In multilingual contexts, we also advise users to write their instructions in English, as most instructions utilized during the model training process were originally written in English.
53
+
54
+ ## Usage
55
+
56
+ With Transformers versions earlier than 4.51.0, you may encounter the following error:
57
+ ```
58
+ KeyError: 'qwen3'
59
+ ```
60
+
61
+ ### Sentence Transformers Usage
62
+
63
+ ```python
64
+ # Requires transformers>=4.51.0
65
+ # Requires sentence-transformers>=2.7.0
66
+
67
+ from sentence_transformers import SentenceTransformer
68
+
69
+ # Load the model
70
+ model = SentenceTransformer("Qwen/Qwen3-Embedding-0.6B")
71
+
72
+ # We recommend enabling flash_attention_2 for better acceleration and memory saving,
73
+ # together with setting `padding_side` to "left":
74
+ # model = SentenceTransformer(
75
+ # "Qwen/Qwen3-Embedding-0.6B",
76
+ # model_kwargs={"attn_implementation": "flash_attention_2", "device_map": "auto"},
77
+ # tokenizer_kwargs={"padding_side": "left"},
78
+ # )
79
+
80
+ # The queries and documents to embed
81
+ queries = [
82
+ "What is the capital of China?",
83
+ "Explain gravity",
84
+ ]
85
+ documents = [
86
+ "The capital of China is Beijing.",
87
+ "Gravity is a force that attracts two bodies towards each other. It gives weight to physical objects and is responsible for the movement of planets around the sun.",
88
+ ]
89
+
90
+ # Encode the queries and documents. Note that queries benefit from using a prompt
91
+ # Here we use the prompt called "query" stored under `model.prompts`, but you can
92
+ # also pass your own prompt via the `prompt` argument
93
+ query_embeddings = model.encode(queries, prompt_name="query")
94
+ document_embeddings = model.encode(documents)
95
+
96
+ # Compute the (cosine) similarity between the query and document embeddings
97
+ similarity = model.similarity(query_embeddings, document_embeddings)
98
+ print(similarity)
99
+ # tensor([[0.7646, 0.1414],
100
+ # [0.1355, 0.6000]])
101
+ ```
102
+
103
+ ### Transformers Usage
104
+
105
+ ```python
106
+ # Requires transformers>=4.51.0
107
+
108
+ import torch
109
+ import torch.nn.functional as F
110
+
111
+ from torch import Tensor
112
+ from transformers import AutoTokenizer, AutoModel
113
+
114
+
115
+ def last_token_pool(last_hidden_states: Tensor,
116
+ attention_mask: Tensor) -> Tensor:
117
+ left_padding = (attention_mask[:, -1].sum() == attention_mask.shape[0])
118
+ if left_padding:
119
+ return last_hidden_states[:, -1]
120
+ else:
121
+ sequence_lengths = attention_mask.sum(dim=1) - 1
122
+ batch_size = last_hidden_states.shape[0]
123
+ return last_hidden_states[torch.arange(batch_size, device=last_hidden_states.device), sequence_lengths]
124
+
125
+
126
+ def get_detailed_instruct(task_description: str, query: str) -> str:
127
+ return f'Instruct: {task_description}\nQuery:{query}'
128
+
129
+ # Each query must come with a one-sentence instruction that describes the task
130
+ task = 'Given a web search query, retrieve relevant passages that answer the query'
131
+
132
+ queries = [
133
+ get_detailed_instruct(task, 'What is the capital of China?'),
134
+ get_detailed_instruct(task, 'Explain gravity')
135
+ ]
136
+ # No need to add instruction for retrieval documents
137
+ documents = [
138
+ "The capital of China is Beijing.",
139
+ "Gravity is a force that attracts two bodies towards each other. It gives weight to physical objects and is responsible for the movement of planets around the sun."
140
+ ]
141
+ input_texts = queries + documents
142
+
143
+ tokenizer = AutoTokenizer.from_pretrained('Qwen/Qwen3-Embedding-0.6B', padding_side='left')
144
+ model = AutoModel.from_pretrained('Qwen/Qwen3-Embedding-0.6B')
145
+
146
+ # We recommend enabling flash_attention_2 for better acceleration and memory saving.
147
+ # model = AutoModel.from_pretrained('Qwen/Qwen3-Embedding-0.6B', attn_implementation="flash_attention_2", torch_dtype=torch.float16).cuda()
148
+
149
+ max_length = 8192
150
+
151
+ # Tokenize the input texts
152
+ batch_dict = tokenizer(
153
+ input_texts,
154
+ padding=True,
155
+ truncation=True,
156
+ max_length=max_length,
157
+ return_tensors="pt",
158
+ )
159
+ batch_dict.to(model.device)
160
+ with torch.no_grad():
161
+ outputs = model(**batch_dict)
162
+ embeddings = last_token_pool(outputs.last_hidden_state, batch_dict['attention_mask'])
163
+
164
+ # normalize embeddings
165
+ embeddings = F.normalize(embeddings, p=2, dim=1)
166
+ scores = (embeddings[:2] @ embeddings[2:].T)
167
+ print(scores.tolist())
168
+ # [[0.7645568251609802, 0.14142508804798126], [0.13549736142158508, 0.5999549627304077]]
169
+ ```
170
+
171
+ ### vLLM Usage
172
+
173
+ ```python
174
+ # Requires vllm>=0.8.5
175
+ import torch
176
+ import vllm
177
+ from vllm import LLM
178
+
179
+ def get_detailed_instruct(task_description: str, query: str) -> str:
180
+ return f'Instruct: {task_description}\nQuery:{query}'
181
+
182
+ # Each query must come with a one-sentence instruction that describes the task
183
+ task = 'Given a web search query, retrieve relevant passages that answer the query'
184
+
185
+ queries = [
186
+ get_detailed_instruct(task, 'What is the capital of China?'),
187
+ get_detailed_instruct(task, 'Explain gravity')
188
+ ]
189
+ # No need to add instruction for retrieval documents
190
+ documents = [
191
+ "The capital of China is Beijing.",
192
+ "Gravity is a force that attracts two bodies towards each other. It gives weight to physical objects and is responsible for the movement of planets around the sun."
193
+ ]
194
+ input_texts = queries + documents
195
+
196
+ model = LLM(model="Qwen/Qwen3-Embedding-0.6B", task="embed")
197
+
198
+ outputs = model.embed(input_texts)
199
+ embeddings = torch.tensor([o.outputs.embedding for o in outputs])
200
+ scores = (embeddings[:2] @ embeddings[2:].T)
201
+ print(scores.tolist())
202
+ # [[0.7620252966880798, 0.14078938961029053], [0.1358368694782257, 0.6013815999031067]]
203
+ ```
204
+
205
+ 📌 **Tip**: We recommend that developers customize the `instruct` according to their specific scenarios, tasks, and languages. Our tests have shown that in most retrieval scenarios, not using an `instruct` on the query side can lead to a drop in retrieval performance by approximately 1% to 5%.
206
+
207
+ ## Evaluation
208
+
209
+ ### MTEB (Multilingual)
210
+
211
+ | Model | Size | Mean (Task) | Mean (Type) | Bitxt Mining | Class. | Clust. | Inst. Retri. | Multi. Class. | Pair. Class. | Rerank | Retri. | STS |
212
+ |----------------------------------|:-------:|:-------------:|:-------------:|:--------------:|:--------:|:--------:|:--------------:|:---------------:|:--------------:|:--------:|:--------:|:------:|
213
+ | NV-Embed-v2 | 7B | 56.29 | 49.58 | 57.84 | 57.29 | 40.80 | 1.04 | 18.63 | 78.94 | 63.82 | 56.72 | 71.10|
214
+ | GritLM-7B | 7B | 60.92 | 53.74 | 70.53 | 61.83 | 49.75 | 3.45 | 22.77 | 79.94 | 63.78 | 58.31 | 73.33|
215
+ | BGE-M3 | 0.6B | 59.56 | 52.18 | 79.11 | 60.35 | 40.88 | -3.11 | 20.1 | 80.76 | 62.79 | 54.60 | 74.12|
216
+ | multilingual-e5-large-instruct | 0.6B | 63.22 | 55.08 | 80.13 | 64.94 | 50.75 | -0.40 | 22.91 | 80.86 | 62.61 | 57.12 | 76.81|
217
+ | gte-Qwen2-1.5B-instruct | 1.5B | 59.45 | 52.69 | 62.51 | 58.32 | 52.05 | 0.74 | 24.02 | 81.58 | 62.58 | 60.78 | 71.61|
218
+ | gte-Qwen2-7b-Instruct | 7B | 62.51 | 55.93 | 73.92 | 61.55 | 52.77 | 4.94 | 25.48 | 85.13 | 65.55 | 60.08 | 73.98|
219
+ | text-embedding-3-large | - | 58.93 | 51.41 | 62.17 | 60.27 | 46.89 | -2.68 | 22.03 | 79.17 | 63.89 | 59.27 | 71.68|
220
+ | Cohere-embed-multilingual-v3.0 | - | 61.12 | 53.23 | 70.50 | 62.95 | 46.89 | -1.89 | 22.74 | 79.88 | 64.07 | 59.16 | 74.80|
221
+ | Gemini Embedding | - | 68.37 | 59.59 | 79.28 | 71.82 | 54.59 | 5.18 | **29.16** | 83.63 | 65.58 | 67.71 | 79.40|
222
+ | **Qwen3-Embedding-0.6B** | 0.6B | 64.33 | 56.00 | 72.22 | 66.83 | 52.33 | 5.09 | 24.59 | 80.83 | 61.41 | 64.64 | 76.17|
223
+ | **Qwen3-Embedding-4B** | 4B | 69.45 | 60.86 | 79.36 | 72.33 | 57.15 | **11.56** | 26.77 | 85.05 | 65.08 | 69.60 | 80.86|
224
+ | **Qwen3-Embedding-8B** | 8B | **70.58** | **61.69** | **80.89** | **74.00** | **57.65** | 10.06 | 28.66 | **86.40** | **65.63** | **70.88** | **81.08** |
225
+
226
+ > **Note**: For compared models, the scores are retrieved from MTEB online [leaderboard](https://huggingface.co/spaces/mteb/leaderboard) on May 24th, 2025.
227
+
228
+ ### MTEB (Eng v2)
229
+
230
+ | MTEB English / Models | Param. | Mean(Task) | Mean(Type) | Class. | Clust. | Pair Class. | Rerank. | Retri. | STS | Summ. |
231
+ |--------------------------------|:--------:|:------------:|:------------:|:--------:|:--------:|:-------------:|:---------:|:--------:|:-------:|:-------:|
232
+ | multilingual-e5-large-instruct | 0.6B | 65.53 | 61.21 | 75.54 | 49.89 | 86.24 | 48.74 | 53.47 | 84.72 | 29.89 |
233
+ | NV-Embed-v2 | 7.8B | 69.81 | 65.00 | 87.19 | 47.66 | 88.69 | 49.61 | 62.84 | 83.82 | 35.21 |
234
+ | GritLM-7B | 7.2B | 67.07 | 63.22 | 81.25 | 50.82 | 87.29 | 49.59 | 54.95 | 83.03 | 35.65 |
235
+ | gte-Qwen2-1.5B-instruct | 1.5B | 67.20 | 63.26 | 85.84 | 53.54 | 87.52 | 49.25 | 50.25 | 82.51 | 33.94 |
236
+ | stella_en_1.5B_v5 | 1.5B | 69.43 | 65.32 | 89.38 | 57.06 | 88.02 | 50.19 | 52.42 | 83.27 | 36.91 |
237
+ | gte-Qwen2-7B-instruct | 7.6B | 70.72 | 65.77 | 88.52 | 58.97 | 85.9 | 50.47 | 58.09 | 82.69 | 35.74 |
238
+ | gemini-embedding-exp-03-07 | - | 73.3 | 67.67 | 90.05 | 59.39 | 87.7 | 48.59 | 64.35 | 85.29 | 38.28 |
239
+ | **Qwen3-Embedding-0.6B** | 0.6B | 70.70 | 64.88 | 85.76 | 54.05 | 84.37 | 48.18 | 61.83 | 86.57 | 33.43 |
240
+ | **Qwen3-Embedding-4B** | 4B | 74.60 | 68.10 | 89.84 | 57.51 | 87.01 | 50.76 | 68.46 | 88.72 | 34.39 |
241
+ | **Qwen3-Embedding-8B** | 8B | 75.22 | 68.71 | 90.43 | 58.57 | 87.52 | 51.56 | 69.44 | 88.58 | 34.83 |
242
+
243
+ ### C-MTEB (MTEB Chinese)
244
+
245
+ | C-MTEB | Param. | Mean(Task) | Mean(Type) | Class. | Clust. | Pair Class. | Rerank. | Retr. | STS |
246
+ |------------------|--------|------------|------------|--------|--------|-------------|---------|-------|-------|
247
+ | multilingual-e5-large-instruct | 0.6B | 58.08 | 58.24 | 69.80 | 48.23 | 64.52 | 57.45 | 63.65 | 45.81 |
248
+ | bge-multilingual-gemma2 | 9B | 67.64 | 75.31 | 59.30 | 86.67 | 68.28 | 73.73 | 55.19 | - |
249
+ | gte-Qwen2-1.5B-instruct | 1.5B | 67.12 | 67.79 | 72.53 | 54.61 | 79.5 | 68.21 | 71.86 | 60.05 |
250
+ | gte-Qwen2-7B-instruct | 7.6B | 71.62 | 72.19 | 75.77 | 66.06 | 81.16 | 69.24 | 75.70 | 65.20 |
251
+ | ritrieve_zh_v1 | 0.3B | 72.71 | 73.85 | 76.88 | 66.5 | 85.98 | 72.86 | 76.97 | 63.92 |
252
+ | **Qwen3-Embedding-0.6B** | 0.6B | 66.33 | 67.45 | 71.40 | 68.74 | 76.42 | 62.58 | 71.03 | 54.52 |
253
+ | **Qwen3-Embedding-4B** | 4B | 72.27 | 73.51 | 75.46 | 77.89 | 83.34 | 66.05 | 77.03 | 61.26 |
254
+ | **Qwen3-Embedding-8B** | 8B | 73.84 | 75.00 | 76.97 | 80.08 | 84.23 | 66.99 | 78.21 | 63.53 |
255
+
256
+
257
+ ## Citation
258
+
259
+ If you find our work helpful, feel free to give us a cite.
260
+
261
+ ```
262
+ @article{qwen3embedding,
263
+ title={Qwen3 Embedding: Advancing Text Embedding and Reranking Through Foundation Models},
264
+ author={Zhang, Yanzhao and Li, Mingxin and Long, Dingkun and Zhang, Xin and Lin, Huan and Yang, Baosong and Xie, Pengjun and Yang, An and Liu, Dayiheng and Lin, Junyang and Huang, Fei and Zhou, Jingren},
265
+ journal={arXiv preprint arXiv:2506.05176},
266
+ year={2025}
267
+ }
268
+ ```
qwen3-embedding-0.6b/artifact-manifest.json ADDED
@@ -0,0 +1,76 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "artifactID": "7731c0dd39fae320addd0604a26cd035c6316531eaa77c03470e99db1382d46c",
3
+ "byteCount": 1192697348,
4
+ "contractVersion": 1,
5
+ "coreAICoreVersion": "1.0.0b2",
6
+ "coreAITorchVersion": "0.4.1",
7
+ "files": [
8
+ {
9
+ "byteCount": 16342,
10
+ "path": "SOURCE-README.md",
11
+ "sha256": "3ec8c2dad0732aba83ac9359d4a4fba7ddeb63c8ba422959051150c0a187fee6"
12
+ },
13
+ {
14
+ "byteCount": 661,
15
+ "path": "export-metadata.json",
16
+ "sha256": "72b89b1e17540462f4d8b80e4225501dc69f415dd395581c5473bb1a49585e59"
17
+ },
18
+ {
19
+ "byteCount": 2224,
20
+ "path": "parity-report.json",
21
+ "sha256": "ab2f5f89a0e78060c01869f846b00e945bf9a7714ef53df530596d0b26833a84"
22
+ },
23
+ {
24
+ "byteCount": 32,
25
+ "path": "qwen3-embedding-0.6b_float16_s512_static.aimodel/main.hash",
26
+ "sha256": "ec1bf7feb6e7f8bdc1e24ee0ecf51bffb769f9de8dcb69f2dbfcbcdd2e4e8a84"
27
+ },
28
+ {
29
+ "byteCount": 1192697348,
30
+ "path": "qwen3-embedding-0.6b_float16_s512_static.aimodel/main.mlirb",
31
+ "sha256": "7731c0dd39fae320addd0604a26cd035c6316531eaa77c03470e99db1382d46c"
32
+ },
33
+ {
34
+ "byteCount": 318,
35
+ "path": "qwen3-embedding-0.6b_float16_s512_static.aimodel/metadata.json",
36
+ "sha256": "7f00643721d3283ffe9a14b82582ca2db7c9b66cace7565cfed3b21e6ccd4e69"
37
+ },
38
+ {
39
+ "byteCount": 343656,
40
+ "path": "reference.json",
41
+ "sha256": "ca352bb63c7fa4815ae2ac15e1bf5caee99fcf8d8414512199e6a6056590b771"
42
+ },
43
+ {
44
+ "byteCount": 4116,
45
+ "path": "tokenizer/chat_template.jinja",
46
+ "sha256": "87a2728cb8dc9fe424d624542f6060ec05a1d285ebbec578bb078900e33396b5"
47
+ },
48
+ {
49
+ "byteCount": 11423799,
50
+ "path": "tokenizer/tokenizer.json",
51
+ "sha256": "a1b5ea7c0de53bc91ffec88f5ad6575d2fd847cf25a2c7bfaadd13d8e2dcefb0"
52
+ },
53
+ {
54
+ "byteCount": 376,
55
+ "path": "tokenizer/tokenizer_config.json",
56
+ "sha256": "fde77b035fb8a7fbaddebd91d7183076385e21fa67cb3fb8b78e9718fd9ba974"
57
+ },
58
+ {
59
+ "byteCount": 579,
60
+ "path": "validation-report.json",
61
+ "sha256": "1f8b0e716d0839e36ebabce172f26a96f8e5a059d891f1c69e9470ca7054fb64"
62
+ }
63
+ ],
64
+ "modelAssetName": "qwen3-embedding-0.6b_float16_s512_static.aimodel",
65
+ "modelSHA256": "7731c0dd39fae320addd0604a26cd035c6316531eaa77c03470e99db1382d46c",
66
+ "precision": "float16",
67
+ "pythonVersion": "3.12.7",
68
+ "schemaVersion": 1,
69
+ "seed": 0,
70
+ "sentenceTransformersVersion": "5.5.1",
71
+ "sequenceLength": 512,
72
+ "sourceModel": "Qwen/Qwen3-Embedding-0.6B",
73
+ "sourceRevision": "fdbdb0320177f59749e02fd66fbdc86082872ef0",
74
+ "torchVersion": "2.11.0",
75
+ "transformersVersion": "5.12.0"
76
+ }
qwen3-embedding-0.6b/export-metadata.json ADDED
@@ -0,0 +1,25 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "coreAICoreVersion": "1.0.0b2",
3
+ "coreAITorchVersion": "0.4.1",
4
+ "minimumWrapperCosine": 1.0,
5
+ "precision": "float16",
6
+ "pythonVersion": "3.12.7",
7
+ "schemaVersion": 1,
8
+ "seed": 0,
9
+ "sentenceTransformersVersion": "5.5.1",
10
+ "sequenceLength": 512,
11
+ "sourceModel": "Qwen/Qwen3-Embedding-0.6B",
12
+ "sourceRevision": "fdbdb0320177f59749e02fd66fbdc86082872ef0",
13
+ "torchVersion": "2.11.0",
14
+ "transformersVersion": "5.12.0",
15
+ "wrapperCosines": {
16
+ "d_beesting": 1.0000001192092896,
17
+ "d_fuji_ja": 1.0,
18
+ "d_python": 1.0,
19
+ "d_tokyo": 1.0000001192092896,
20
+ "q_beesting": 1.0,
21
+ "q_capital": 1.0,
22
+ "q_fuji_ja": 1.0,
23
+ "q_mllang": 1.0
24
+ }
25
+ }
qwen3-embedding-0.6b/installation-receipt.json ADDED
@@ -0,0 +1,15 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "descriptor": {
3
+ "artifactID": "7731c0dd39fae320addd0604a26cd035c6316531eaa77c03470e99db1382d46c",
4
+ "byteCount": 1192697348,
5
+ "contractVersion": 1,
6
+ "manifestSHA256": "51a2756b507ebaf2425f4ba4c06084866c8198e627c3ff6f1125158d2637e551",
7
+ "modelAssetName": "qwen3-embedding-0.6b_float16_s512_static.aimodel",
8
+ "modelSHA256": "7731c0dd39fae320addd0604a26cd035c6316531eaa77c03470e99db1382d46c",
9
+ "schemaVersion": 1,
10
+ "sourceRevision": "fdbdb0320177f59749e02fd66fbdc86082872ef0"
11
+ },
12
+ "installedAt": "2026-07-16T15:20:45.890047+00:00",
13
+ "schemaVersion": 1,
14
+ "validationReportSHA256": "1f8b0e716d0839e36ebabce172f26a96f8e5a059d891f1c69e9470ca7054fb64"
15
+ }
qwen3-embedding-0.6b/parity-report.json ADDED
@@ -0,0 +1,108 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "artifactID": "7731c0dd39fae320addd0604a26cd035c6316531eaa77c03470e99db1382d46c",
3
+ "coreAICosines": {
4
+ "d_beesting": 0.999998668498719,
5
+ "d_fuji_ja": 0.9999980314382556,
6
+ "d_python": 0.9999989100667815,
7
+ "d_tokyo": 0.9999983726631115,
8
+ "q_beesting": 0.9999985254240714,
9
+ "q_capital": 0.9999986042186417,
10
+ "q_fuji_ja": 0.9999982020130729,
11
+ "q_mllang": 0.9999987209482887
12
+ },
13
+ "coreAIRankings": {
14
+ "q_beesting": [
15
+ "d_beesting",
16
+ "d_tokyo",
17
+ "d_python",
18
+ "d_fuji_ja"
19
+ ],
20
+ "q_capital": [
21
+ "d_tokyo",
22
+ "d_fuji_ja",
23
+ "d_beesting",
24
+ "d_python"
25
+ ],
26
+ "q_fuji_ja": [
27
+ "d_fuji_ja",
28
+ "d_tokyo",
29
+ "d_python",
30
+ "d_beesting"
31
+ ],
32
+ "q_mllang": [
33
+ "d_python",
34
+ "d_fuji_ja",
35
+ "d_tokyo",
36
+ "d_beesting"
37
+ ]
38
+ },
39
+ "correctedReferenceRankings": {
40
+ "q_beesting": [
41
+ "d_beesting",
42
+ "d_tokyo",
43
+ "d_python",
44
+ "d_fuji_ja"
45
+ ],
46
+ "q_capital": [
47
+ "d_tokyo",
48
+ "d_fuji_ja",
49
+ "d_beesting",
50
+ "d_python"
51
+ ],
52
+ "q_fuji_ja": [
53
+ "d_fuji_ja",
54
+ "d_tokyo",
55
+ "d_python",
56
+ "d_beesting"
57
+ ],
58
+ "q_mllang": [
59
+ "d_python",
60
+ "d_fuji_ja",
61
+ "d_tokyo",
62
+ "d_beesting"
63
+ ]
64
+ },
65
+ "expectedRankings": {
66
+ "q_beesting": [
67
+ "d_beesting",
68
+ "d_tokyo",
69
+ "d_python",
70
+ "d_fuji_ja"
71
+ ],
72
+ "q_capital": [
73
+ "d_tokyo",
74
+ "d_fuji_ja",
75
+ "d_beesting",
76
+ "d_python"
77
+ ],
78
+ "q_fuji_ja": [
79
+ "d_fuji_ja",
80
+ "d_tokyo",
81
+ "d_python",
82
+ "d_beesting"
83
+ ],
84
+ "q_mllang": [
85
+ "d_python",
86
+ "d_fuji_ja",
87
+ "d_tokyo",
88
+ "d_beesting"
89
+ ]
90
+ },
91
+ "schemaVersion": 1,
92
+ "semanticBaselineCosines": {
93
+ "d_beesting": 0.9999999797939612,
94
+ "d_fuji_ja": 0.9999999780099337,
95
+ "d_python": 0.9999999794280262,
96
+ "d_tokyo": 0.9999999775174578,
97
+ "q_beesting": 0.9999999801232986,
98
+ "q_capital": 0.9999999787482269,
99
+ "q_fuji_ja": 0.9999999774654617,
100
+ "q_mllang": 0.9999999787463958
101
+ },
102
+ "thresholds": {
103
+ "maximumNormDeviation": 0.02,
104
+ "minimumCoreAICosine": 0.999,
105
+ "minimumSemanticBaselineCosine": 0.9999
106
+ },
107
+ "validatedAt": "2026-07-16T15:20:44.789269+00:00"
108
+ }
qwen3-embedding-0.6b/qwen3-embedding-0.6b_float16_s512_static.aimodel/main.hash ADDED
@@ -0,0 +1 @@
 
 
1
+ w1��9�� ���l�5�1e1�|G����l
qwen3-embedding-0.6b/qwen3-embedding-0.6b_float16_s512_static.aimodel/main.mlirb ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7731c0dd39fae320addd0604a26cd035c6316531eaa77c03470e99db1382d46c
3
+ size 1192697348
qwen3-embedding-0.6b/qwen3-embedding-0.6b_float16_s512_static.aimodel/metadata.json ADDED
@@ -0,0 +1,8 @@
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "license" : "Apache-2.0",
3
+ "assetVersion" : "2.0",
4
+ "author" : "Alibaba Qwen",
5
+ "producer" : "coreai-core 1.0.0b2",
6
+ "creationDate" : "20260716T152031Z",
7
+ "description" : "Qwen3-Embedding-0.6B, pinned at fdbdb0320177f59749e02fd66fbdc86082872ef0; last-token pooling; normalized 1024-dimensional fp16 output."
8
+ }
qwen3-embedding-0.6b/reference.json ADDED
The diff for this file is too large to render. See raw diff
 
qwen3-embedding-0.6b/tokenizer/chat_template.jinja ADDED
@@ -0,0 +1,85 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {%- if tools %}
2
+ {{- '<|im_start|>system\n' }}
3
+ {%- if messages[0].role == 'system' %}
4
+ {{- messages[0].content + '\n\n' }}
5
+ {%- endif %}
6
+ {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
7
+ {%- for tool in tools %}
8
+ {{- "\n" }}
9
+ {{- tool | tojson }}
10
+ {%- endfor %}
11
+ {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
12
+ {%- else %}
13
+ {%- if messages[0].role == 'system' %}
14
+ {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
15
+ {%- endif %}
16
+ {%- endif %}
17
+ {%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
18
+ {%- for message in messages[::-1] %}
19
+ {%- set index = (messages|length - 1) - loop.index0 %}
20
+ {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
21
+ {%- set ns.multi_step_tool = false %}
22
+ {%- set ns.last_query_index = index %}
23
+ {%- endif %}
24
+ {%- endfor %}
25
+ {%- for message in messages %}
26
+ {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
27
+ {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
28
+ {%- elif message.role == "assistant" %}
29
+ {%- set content = message.content %}
30
+ {%- set reasoning_content = '' %}
31
+ {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
32
+ {%- set reasoning_content = message.reasoning_content %}
33
+ {%- else %}
34
+ {%- if '</think>' in message.content %}
35
+ {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
36
+ {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
37
+ {%- endif %}
38
+ {%- endif %}
39
+ {%- if loop.index0 > ns.last_query_index %}
40
+ {%- if loop.last or (not loop.last and reasoning_content) %}
41
+ {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
42
+ {%- else %}
43
+ {{- '<|im_start|>' + message.role + '\n' + content }}
44
+ {%- endif %}
45
+ {%- else %}
46
+ {{- '<|im_start|>' + message.role + '\n' + content }}
47
+ {%- endif %}
48
+ {%- if message.tool_calls %}
49
+ {%- for tool_call in message.tool_calls %}
50
+ {%- if (loop.first and content) or (not loop.first) %}
51
+ {{- '\n' }}
52
+ {%- endif %}
53
+ {%- if tool_call.function %}
54
+ {%- set tool_call = tool_call.function %}
55
+ {%- endif %}
56
+ {{- '<tool_call>\n{"name": "' }}
57
+ {{- tool_call.name }}
58
+ {{- '", "arguments": ' }}
59
+ {%- if tool_call.arguments is string %}
60
+ {{- tool_call.arguments }}
61
+ {%- else %}
62
+ {{- tool_call.arguments | tojson }}
63
+ {%- endif %}
64
+ {{- '}\n</tool_call>' }}
65
+ {%- endfor %}
66
+ {%- endif %}
67
+ {{- '<|im_end|>\n' }}
68
+ {%- elif message.role == "tool" %}
69
+ {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
70
+ {{- '<|im_start|>user' }}
71
+ {%- endif %}
72
+ {{- '\n<tool_response>\n' }}
73
+ {{- message.content }}
74
+ {{- '\n</tool_response>' }}
75
+ {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
76
+ {{- '<|im_end|>\n' }}
77
+ {%- endif %}
78
+ {%- endif %}
79
+ {%- endfor %}
80
+ {%- if add_generation_prompt %}
81
+ {{- '<|im_start|>assistant\n' }}
82
+ {%- if enable_thinking is defined and enable_thinking is false %}
83
+ {{- '<think>\n\n</think>\n\n' }}
84
+ {%- endif %}
85
+ {%- endif %}
qwen3-embedding-0.6b/tokenizer/tokenizer.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a1b5ea7c0de53bc91ffec88f5ad6575d2fd847cf25a2c7bfaadd13d8e2dcefb0
3
+ size 11423799
qwen3-embedding-0.6b/tokenizer/tokenizer_config.json ADDED
@@ -0,0 +1,15 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "add_prefix_space": false,
3
+ "backend": "tokenizers",
4
+ "bos_token": null,
5
+ "clean_up_tokenization_spaces": false,
6
+ "eos_token": "<|im_end|>",
7
+ "errors": "replace",
8
+ "is_local": true,
9
+ "local_files_only": false,
10
+ "model_max_length": 32768,
11
+ "pad_token": "<|endoftext|>",
12
+ "split_special_tokens": false,
13
+ "tokenizer_class": "Qwen2Tokenizer",
14
+ "unk_token": null
15
+ }
qwen3-embedding-0.6b/validation-report.json ADDED
@@ -0,0 +1,21 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "artifactID": "7731c0dd39fae320addd0604a26cd035c6316531eaa77c03470e99db1382d46c",
3
+ "embeddingDimension": 1024,
4
+ "inputShape": [
5
+ 1,
6
+ 512
7
+ ],
8
+ "maxNormDeviation": 0.000436029707223895,
9
+ "minimumCoreAICosine": 0.9999980314382556,
10
+ "minimumSemanticBaselineCosine": 0.9999999774654617,
11
+ "modelSHA256": "7731c0dd39fae320addd0604a26cd035c6316531eaa77c03470e99db1382d46c",
12
+ "outputScalarType": "float16",
13
+ "outputShape": [
14
+ 1,
15
+ 1024
16
+ ],
17
+ "passed": true,
18
+ "rankingsUnchanged": true,
19
+ "schemaVersion": 1,
20
+ "validatedAt": "2026-07-16T15:20:44.789269+00:00"
21
+ }