thepushkarp commited on
Commit
e5962ab
·
verified ·
1 Parent(s): e9a998d

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +32 -1
README.md CHANGED
@@ -22,6 +22,37 @@ Maximum relative tensor difference: 0.229572
22
  Average absolute tensor difference: 0.000010
23
  ```
24
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
25
  <center>
26
  <a href="https://github.com/nari-labs/dia">
27
  <img src="https://github.com/nari-labs/dia/raw/main/dia/static/images/banner.png">
@@ -80,7 +111,7 @@ import soundfile as sf
80
  from dia.model import Dia
81
 
82
 
83
- model = Dia.from_pretrained("thepushkarp/Dia-1.6B-safetensors-fp16")
84
 
85
  text = "[S1] Dia is an open weights text to dialogue model. [S2] You get full control over scripts and voices. [S1] Wow. Amazing. (laughs) [S2] Try it now on Git hub or Hugging Face."
86
 
 
22
  Average absolute tensor difference: 0.000010
23
  ```
24
 
25
+ To use the safetensors file, use this custom script which allows loading from safetensors:
26
+
27
+ First install the library:
28
+ ```
29
+ git clone https://github.com/thepushkarp/dia.git
30
+ cd dia
31
+ python -m venv .venv
32
+ source .venv/bin/activate
33
+ ```
34
+
35
+ Then run:
36
+ ```
37
+ import soundfile as sf
38
+
39
+ from dia.model import Dia
40
+
41
+
42
+ model = Dia.from_pretrained(
43
+ "thepushkarp/Dia-1.6B-safetensors-fp16",
44
+ config_path="config.json",
45
+ checkpoint_path="dia-v0_1-fp16.safetensors",
46
+ )
47
+ text = "[S1] Dia is an open weights text to dialogue model. [S2] You get full control over scripts and voices. [S1] Wow. Amazing. (laughs) [S2] Try it now on Git hub or Hugging Face."
48
+
49
+ output = model.generate(text)
50
+
51
+ sf.write("simple.mp3", output, 44100)
52
+ ```
53
+
54
+ ---
55
+
56
  <center>
57
  <a href="https://github.com/nari-labs/dia">
58
  <img src="https://github.com/nari-labs/dia/raw/main/dia/static/images/banner.png">
 
111
  from dia.model import Dia
112
 
113
 
114
+ model = Dia.from_pretrained("nari-labs/Dia-1.6B")
115
 
116
  text = "[S1] Dia is an open weights text to dialogue model. [S2] You get full control over scripts and voices. [S1] Wow. Amazing. (laughs) [S2] Try it now on Git hub or Hugging Face."
117