Skip to content

Commit 330c46b

Browse files
ssss141414ssss141414
committed
recipe(opus-mt-en-ru): add translation composite recipe pair (Goal-L2-encoder PASS on CPU)
1 parent b448652 commit 330c46b

3 files changed

Lines changed: 343 additions & 0 deletions

File tree

‎examples/recipes/Helsinki-NLP_opus-mt-en-ru/translation_fp16_decoder_config.json‎

Lines changed: 285 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,285 @@
1+
{
2+
"export": {
3+
"opset_version": 17,
4+
"batch_size": 1,
5+
"export_params": true,
6+
"do_constant_folding": true,
7+
"verbose": false,
8+
"dynamo": false,
9+
"enable_hierarchy_tags": true,
10+
"clean_onnx": false,
11+
"hierarchy_tag_format": "full",
12+
"input_tensors": [
13+
{
14+
"name": "decoder_input_ids",
15+
"dtype": "int32",
16+
"shape": [
17+
1,
18+
1
19+
],
20+
"value_range": [
21+
0,
22+
62518
23+
]
24+
},
25+
{
26+
"name": "encoder_hidden_states",
27+
"dtype": "float32",
28+
"shape": [
29+
1,
30+
512,
31+
512
32+
],
33+
"value_range": [
34+
0,
35+
1
36+
]
37+
},
38+
{
39+
"name": "attention_mask",
40+
"dtype": "int64",
41+
"shape": [
42+
1,
43+
512
44+
]
45+
},
46+
{
47+
"name": "decoder_attention_mask",
48+
"dtype": "int64",
49+
"shape": [
50+
1,
51+
512
52+
]
53+
},
54+
{
55+
"name": "cache_position",
56+
"dtype": "int64",
57+
"shape": [
58+
1
59+
]
60+
},
61+
{
62+
"name": "past_0_key",
63+
"dtype": "float32",
64+
"shape": [
65+
1,
66+
8,
67+
512,
68+
64
69+
],
70+
"value_range": [
71+
0,
72+
1
73+
]
74+
},
75+
{
76+
"name": "past_0_value",
77+
"dtype": "float32",
78+
"shape": [
79+
1,
80+
8,
81+
512,
82+
64
83+
],
84+
"value_range": [
85+
0,
86+
1
87+
]
88+
},
89+
{
90+
"name": "past_1_key",
91+
"dtype": "float32",
92+
"shape": [
93+
1,
94+
8,
95+
512,
96+
64
97+
],
98+
"value_range": [
99+
0,
100+
1
101+
]
102+
},
103+
{
104+
"name": "past_1_value",
105+
"dtype": "float32",
106+
"shape": [
107+
1,
108+
8,
109+
512,
110+
64
111+
],
112+
"value_range": [
113+
0,
114+
1
115+
]
116+
},
117+
{
118+
"name": "past_2_key",
119+
"dtype": "float32",
120+
"shape": [
121+
1,
122+
8,
123+
512,
124+
64
125+
],
126+
"value_range": [
127+
0,
128+
1
129+
]
130+
},
131+
{
132+
"name": "past_2_value",
133+
"dtype": "float32",
134+
"shape": [
135+
1,
136+
8,
137+
512,
138+
64
139+
],
140+
"value_range": [
141+
0,
142+
1
143+
]
144+
},
145+
{
146+
"name": "past_3_key",
147+
"dtype": "float32",
148+
"shape": [
149+
1,
150+
8,
151+
512,
152+
64
153+
],
154+
"value_range": [
155+
0,
156+
1
157+
]
158+
},
159+
{
160+
"name": "past_3_value",
161+
"dtype": "float32",
162+
"shape": [
163+
1,
164+
8,
165+
512,
166+
64
167+
],
168+
"value_range": [
169+
0,
170+
1
171+
]
172+
},
173+
{
174+
"name": "past_4_key",
175+
"dtype": "float32",
176+
"shape": [
177+
1,
178+
8,
179+
512,
180+
64
181+
],
182+
"value_range": [
183+
0,
184+
1
185+
]
186+
},
187+
{
188+
"name": "past_4_value",
189+
"dtype": "float32",
190+
"shape": [
191+
1,
192+
8,
193+
512,
194+
64
195+
],
196+
"value_range": [
197+
0,
198+
1
199+
]
200+
},
201+
{
202+
"name": "past_5_key",
203+
"dtype": "float32",
204+
"shape": [
205+
1,
206+
8,
207+
512,
208+
64
209+
],
210+
"value_range": [
211+
0,
212+
1
213+
]
214+
},
215+
{
216+
"name": "past_5_value",
217+
"dtype": "float32",
218+
"shape": [
219+
1,
220+
8,
221+
512,
222+
64
223+
],
224+
"value_range": [
225+
0,
226+
1
227+
]
228+
}
229+
],
230+
"output_tensors": [
231+
{
232+
"name": "logits"
233+
},
234+
{
235+
"name": "present_0_key"
236+
},
237+
{
238+
"name": "present_0_value"
239+
},
240+
{
241+
"name": "present_1_key"
242+
},
243+
{
244+
"name": "present_1_value"
245+
},
246+
{
247+
"name": "present_2_key"
248+
},
249+
{
250+
"name": "present_2_value"
251+
},
252+
{
253+
"name": "present_3_key"
254+
},
255+
{
256+
"name": "present_3_value"
257+
},
258+
{
259+
"name": "present_4_key"
260+
},
261+
{
262+
"name": "present_4_value"
263+
},
264+
{
265+
"name": "present_5_key"
266+
},
267+
{
268+
"name": "present_5_value"
269+
}
270+
]
271+
},
272+
"optim": {
273+
"clamp_constant_values": true,
274+
"gelu_fusion": true,
275+
"matmul_add_fusion": true,
276+
"remove_isnan_in_attention_mask": true
277+
},
278+
"quant": null,
279+
"compile": null,
280+
"loader": {
281+
"task": "text2text-generation",
282+
"model_class": "MarianDecoderWrapper",
283+
"model_type": "marian"
284+
}
285+
}

‎examples/recipes/Helsinki-NLP_opus-mt-en-ru/translation_fp16_encoder_config.json‎

Lines changed: 57 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,57 @@
1+
{
2+
"export": {
3+
"opset_version": 17,
4+
"batch_size": 1,
5+
"export_params": true,
6+
"do_constant_folding": true,
7+
"verbose": false,
8+
"dynamo": false,
9+
"enable_hierarchy_tags": true,
10+
"clean_onnx": false,
11+
"hierarchy_tag_format": "full",
12+
"input_tensors": [
13+
{
14+
"name": "input_ids",
15+
"dtype": "int32",
16+
"shape": [
17+
1,
18+
512
19+
],
20+
"value_range": [
21+
0,
22+
62518
23+
]
24+
},
25+
{
26+
"name": "attention_mask",
27+
"dtype": "int32",
28+
"shape": [
29+
1,
30+
512
31+
],
32+
"value_range": [
33+
0,
34+
2
35+
]
36+
}
37+
],
38+
"output_tensors": [
39+
{
40+
"name": "encoder_hidden_states"
41+
}
42+
]
43+
},
44+
"optim": {
45+
"clamp_constant_values": true,
46+
"gelu_fusion": true,
47+
"matmul_add_fusion": true,
48+
"remove_isnan_in_attention_mask": true
49+
},
50+
"quant": null,
51+
"compile": null,
52+
"loader": {
53+
"task": "feature-extraction",
54+
"model_class": "MarianEncoderWrapper",
55+
"model_type": "marian"
56+
}
57+
}

‎examples/recipes/README.md‎

Lines changed: 1 addition & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -17,6 +17,7 @@ Each *(model, task)* includes:
1717
| Model | Task |
1818
|---|---|
1919
| BAAI/bge-large-en-v1.5 | sentence-similarity |
20+
| Helsinki-NLP/opus-mt-en-ru | translation |
2021
| cardiffnlp/twitter-roberta-base-sentiment-latest | text-classification |
2122
| deepset/roberta-base-squad2 | question-answering |
2223
| deepset/tinyroberta-squad2 | question-answering |

0 commit comments

Comments
 (0)