Add paper and code links to model card

#1
by nielsr HF Staff - opened
Files changed (1) hide show
  1. README.md +8 -118
README.md CHANGED
@@ -1,132 +1,21 @@
1
  ---
2
- license: apache-2.0
 
3
  language:
4
  - en
5
  - ko
6
  library_name: transformers
 
 
7
  tags:
8
  - audio
9
  - text-generation
10
- pipeline_tag: audio-text-to-text
11
- base_model:
12
- - Qwen/Qwen3-4B
13
- # model-index:
14
- # - name: FastSLM
15
- # results:
16
- # - task:
17
- # name: Automatic Speech Recognition
18
- # type: automatic-speech-recognition
19
- # dataset:
20
- # name: AMI (Meetings test)
21
- # type: edinburghcstr/ami
22
- # config: ihm
23
- # split: test
24
- # args:
25
- # language: en
26
- # metrics:
27
- # - name: Test WER
28
- # type: wer
29
- # value: 10.8
30
- # - task:
31
- # name: Automatic Speech Recognition
32
- # type: automatic-speech-recognition
33
- # dataset:
34
- # name: Earnings-22
35
- # type: revdotcom/earnings22
36
- # split: test
37
- # args:
38
- # language: en
39
- # metrics:
40
- # - name: Test WER
41
- # type: wer
42
- # value: 10.7
43
- # - task:
44
- # name: Automatic Speech Recognition
45
- # type: automatic-speech-recognition
46
- # dataset:
47
- # name: GigaSpeech
48
- # type: speechcolab/gigaspeech
49
- # split: test
50
- # args:
51
- # language: en
52
- # metrics:
53
- # - name: Test WER
54
- # type: wer
55
- # value: 10.7
56
- # - task:
57
- # name: Automatic Speech Recognition
58
- # type: automatic-speech-recognition
59
- # dataset:
60
- # name: LibriSpeech (clean)
61
- # type: librispeech_asr
62
- # config: other
63
- # split: test
64
- # args:
65
- # language: en
66
- # metrics:
67
- # - name: Test WER
68
- # type: wer
69
- # value: 2.26
70
- # - task:
71
- # name: Automatic Speech Recognition
72
- # type: automatic-speech-recognition
73
- # dataset:
74
- # name: LibriSpeech (other)
75
- # type: librispeech_asr
76
- # config: other
77
- # split: test
78
- # args:
79
- # language: en
80
- # metrics:
81
- # - name: Test WER
82
- # type: wer
83
- # value: 4.87
84
- # - task:
85
- # name: Automatic Speech Recognition
86
- # type: automatic-speech-recognition
87
- # dataset:
88
- # name: Vox Populi
89
- # type: facebook/voxpopuli
90
- # config: en
91
- # split: test
92
- # args:
93
- # language: en
94
- # metrics:
95
- # - name: Test WER
96
- # type: wer
97
- # value: 6.55
98
- # - task:
99
- # type: Automatic Speech Recognition
100
- # name: automatic-speech-recognition
101
- # dataset:
102
- # name: tedlium-v3
103
- # type: LIUM/tedlium
104
- # config: release1
105
- # split: test
106
- # args:
107
- # language: en
108
- # metrics:
109
- # - name: Test WER
110
- # type: wer
111
- # value: 3.97
112
- # - task:
113
- # type: Automatic Speech Recognition
114
- # name: automatic-speech-recognition
115
- # dataset:
116
- # name: SPGI Speech
117
- # type: kensho/spgispeech
118
- # config: test
119
- # split: test
120
- # args:
121
- # language: en
122
- # metrics:
123
- # - name: Test WER
124
- # type: wer
125
- # value: 2.33
126
  ---
127
 
128
  * FastSLM is an designed for efficient and accurate speech-to-text transcription.
129
 
 
 
130
 
131
  <!-- * 🔊 **HFQ-Former**: Hierarchically compresses high-frame-rate audio features while preserving the audio's local and global contextual information. -->
132
  <!-- * 🔊 **Adaptor**: -->
@@ -217,7 +106,8 @@ else:
217
  # A task token is not required, but it is recommended for achieving a more appropriate task.
218
  TASK_TOKEN = "<|ASR|>"
219
  AUDIO_TOKEN = "<|audio_bos|><|AUDIO|><|audio_eos|>"
220
- user_prompt = f"{TASK_TOKEN}{AUDIO_TOKEN}\nTranscribe the audio clip into text."
 
221
 
222
  prompt = [{"role": "user", "content": user_prompt}]
223
  input_ids = tokenizer.apply_chat_template(
 
1
  ---
2
+ base_model:
3
+ - Qwen/Qwen3-4B
4
  language:
5
  - en
6
  - ko
7
  library_name: transformers
8
+ license: apache-2.0
9
+ pipeline_tag: audio-text-to-text
10
  tags:
11
  - audio
12
  - text-generation
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
13
  ---
14
 
15
  * FastSLM is an designed for efficient and accurate speech-to-text transcription.
16
 
17
+ **Paper**: [FastSLM: Hierarchical Temporal Abstraction for Efficient Long-Form Speech Adaptation](https://huggingface.co/papers/2601.06199)
18
+ **Code**: [GitHub - Lee-junseok1025/FastSLM](https://github.com/Lee-junseok1025/FastSLM)
19
 
20
  <!-- * 🔊 **HFQ-Former**: Hierarchically compresses high-frame-rate audio features while preserving the audio's local and global contextual information. -->
21
  <!-- * 🔊 **Adaptor**: -->
 
106
  # A task token is not required, but it is recommended for achieving a more appropriate task.
107
  TASK_TOKEN = "<|ASR|>"
108
  AUDIO_TOKEN = "<|audio_bos|><|AUDIO|><|audio_eos|>"
109
+ user_prompt = f"{TASK_TOKEN}{AUDIO_TOKEN}
110
+ Transcribe the audio clip into text."
111
 
112
  prompt = [{"role": "user", "content": user_prompt}]
113
  input_ids = tokenizer.apply_chat_template(