Maya (en)
Browse files- .gitattributes +13 -0
- en/maya1-GGUF/.gitattributes +47 -0
- en/maya1-GGUF/README.md +73 -0
- en/maya1-GGUF/maya1.IQ4_XS.gguf +3 -0
- en/maya1-GGUF/maya1.Q2_K.gguf +3 -0
- en/maya1-GGUF/maya1.Q3_K_L.gguf +3 -0
- en/maya1-GGUF/maya1.Q3_K_M.gguf +3 -0
- en/maya1-GGUF/maya1.Q3_K_S.gguf +3 -0
- en/maya1-GGUF/maya1.Q4_K_M.gguf +3 -0
- en/maya1-GGUF/maya1.Q4_K_S.gguf +3 -0
- en/maya1-GGUF/maya1.Q5_K_M.gguf +3 -0
- en/maya1-GGUF/maya1.Q5_K_S.gguf +3 -0
- en/maya1-GGUF/maya1.Q6_K.gguf +3 -0
- en/maya1-GGUF/maya1.Q8_0.gguf +3 -0
- en/maya1-GGUF/maya1.f16.gguf +3 -0
- en/maya1-GGUF/source.txt +1 -0
- en/maya1/.gitattributes +42 -0
- en/maya1/.gitignore +1 -0
- en/maya1/README.md +583 -0
- en/maya1/assets +0 -0
- en/maya1/chat_template.jinja +93 -0
- en/maya1/config.json +36 -0
- en/maya1/emotions.txt +17 -0
- en/maya1/generation_config.json +13 -0
- en/maya1/model-00002-of-00002.safetensors +3 -0
- en/maya1/model.safetensors.index.json +262 -0
- en/maya1/prompt.txt +97 -0
- en/maya1/source.txt +1 -0
- en/maya1/special_tokens_map.json +165 -0
- en/maya1/tokenizer/chat_template.jinja +93 -0
- en/maya1/tokenizer/special_tokens_map.json +165 -0
- en/maya1/tokenizer/tokenizer.json +3 -0
- en/maya1/tokenizer/tokenizer_config.json +0 -0
- en/maya1/tokenizer_config.json +0 -0
- en/maya1/vllm_streaming_inference.py +561 -0
.gitattributes
CHANGED
|
@@ -33,3 +33,16 @@ 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 |
+
en/maya1-GGUF/maya1.f16.gguf filter=lfs diff=lfs merge=lfs -text
|
| 37 |
+
en/maya1-GGUF/maya1.IQ4_XS.gguf filter=lfs diff=lfs merge=lfs -text
|
| 38 |
+
en/maya1-GGUF/maya1.Q2_K.gguf filter=lfs diff=lfs merge=lfs -text
|
| 39 |
+
en/maya1-GGUF/maya1.Q3_K_L.gguf filter=lfs diff=lfs merge=lfs -text
|
| 40 |
+
en/maya1-GGUF/maya1.Q3_K_M.gguf filter=lfs diff=lfs merge=lfs -text
|
| 41 |
+
en/maya1-GGUF/maya1.Q3_K_S.gguf filter=lfs diff=lfs merge=lfs -text
|
| 42 |
+
en/maya1-GGUF/maya1.Q4_K_M.gguf filter=lfs diff=lfs merge=lfs -text
|
| 43 |
+
en/maya1-GGUF/maya1.Q4_K_S.gguf filter=lfs diff=lfs merge=lfs -text
|
| 44 |
+
en/maya1-GGUF/maya1.Q5_K_M.gguf filter=lfs diff=lfs merge=lfs -text
|
| 45 |
+
en/maya1-GGUF/maya1.Q5_K_S.gguf filter=lfs diff=lfs merge=lfs -text
|
| 46 |
+
en/maya1-GGUF/maya1.Q6_K.gguf filter=lfs diff=lfs merge=lfs -text
|
| 47 |
+
en/maya1-GGUF/maya1.Q8_0.gguf filter=lfs diff=lfs merge=lfs -text
|
| 48 |
+
en/maya1/tokenizer/tokenizer.json filter=lfs diff=lfs merge=lfs -text
|
en/maya1-GGUF/.gitattributes
ADDED
|
@@ -0,0 +1,47 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
*.7z filter=lfs diff=lfs merge=lfs -text
|
| 2 |
+
*.arrow filter=lfs diff=lfs merge=lfs -text
|
| 3 |
+
*.bin filter=lfs diff=lfs merge=lfs -text
|
| 4 |
+
*.bz2 filter=lfs diff=lfs merge=lfs -text
|
| 5 |
+
*.ckpt filter=lfs diff=lfs merge=lfs -text
|
| 6 |
+
*.ftz filter=lfs diff=lfs merge=lfs -text
|
| 7 |
+
*.gz filter=lfs diff=lfs merge=lfs -text
|
| 8 |
+
*.h5 filter=lfs diff=lfs merge=lfs -text
|
| 9 |
+
*.joblib filter=lfs diff=lfs merge=lfs -text
|
| 10 |
+
*.lfs.* filter=lfs diff=lfs merge=lfs -text
|
| 11 |
+
*.mlmodel filter=lfs diff=lfs merge=lfs -text
|
| 12 |
+
*.model filter=lfs diff=lfs merge=lfs -text
|
| 13 |
+
*.msgpack filter=lfs diff=lfs merge=lfs -text
|
| 14 |
+
*.npy filter=lfs diff=lfs merge=lfs -text
|
| 15 |
+
*.npz filter=lfs diff=lfs merge=lfs -text
|
| 16 |
+
*.onnx filter=lfs diff=lfs merge=lfs -text
|
| 17 |
+
*.ot filter=lfs diff=lfs merge=lfs -text
|
| 18 |
+
*.parquet filter=lfs diff=lfs merge=lfs -text
|
| 19 |
+
*.pb filter=lfs diff=lfs merge=lfs -text
|
| 20 |
+
*.pickle filter=lfs diff=lfs merge=lfs -text
|
| 21 |
+
*.pkl filter=lfs diff=lfs merge=lfs -text
|
| 22 |
+
*.pt filter=lfs diff=lfs merge=lfs -text
|
| 23 |
+
*.pth filter=lfs diff=lfs merge=lfs -text
|
| 24 |
+
*.rar filter=lfs diff=lfs merge=lfs -text
|
| 25 |
+
*.safetensors filter=lfs diff=lfs merge=lfs -text
|
| 26 |
+
saved_model/**/* filter=lfs diff=lfs merge=lfs -text
|
| 27 |
+
*.tar.* filter=lfs diff=lfs merge=lfs -text
|
| 28 |
+
*.tar filter=lfs diff=lfs merge=lfs -text
|
| 29 |
+
*.tflite filter=lfs diff=lfs merge=lfs -text
|
| 30 |
+
*.tgz filter=lfs diff=lfs merge=lfs -text
|
| 31 |
+
*.wasm filter=lfs diff=lfs merge=lfs -text
|
| 32 |
+
*.xz 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 |
+
maya1.IQ4_XS.gguf filter=lfs diff=lfs merge=lfs -text
|
| 37 |
+
maya1.Q2_K.gguf filter=lfs diff=lfs merge=lfs -text
|
| 38 |
+
maya1.Q3_K_L.gguf filter=lfs diff=lfs merge=lfs -text
|
| 39 |
+
maya1.Q3_K_M.gguf filter=lfs diff=lfs merge=lfs -text
|
| 40 |
+
maya1.Q3_K_S.gguf filter=lfs diff=lfs merge=lfs -text
|
| 41 |
+
maya1.Q4_K_M.gguf filter=lfs diff=lfs merge=lfs -text
|
| 42 |
+
maya1.Q4_K_S.gguf filter=lfs diff=lfs merge=lfs -text
|
| 43 |
+
maya1.Q5_K_M.gguf filter=lfs diff=lfs merge=lfs -text
|
| 44 |
+
maya1.Q5_K_S.gguf filter=lfs diff=lfs merge=lfs -text
|
| 45 |
+
maya1.Q6_K.gguf filter=lfs diff=lfs merge=lfs -text
|
| 46 |
+
maya1.Q8_0.gguf filter=lfs diff=lfs merge=lfs -text
|
| 47 |
+
maya1.f16.gguf filter=lfs diff=lfs merge=lfs -text
|
en/maya1-GGUF/README.md
ADDED
|
@@ -0,0 +1,73 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
base_model: maya-research/maya1
|
| 3 |
+
datasets: proprietary
|
| 4 |
+
language:
|
| 5 |
+
- en
|
| 6 |
+
library_name: transformers
|
| 7 |
+
license: apache-2.0
|
| 8 |
+
mradermacher:
|
| 9 |
+
readme_rev: 1
|
| 10 |
+
quantized_by: mradermacher
|
| 11 |
+
---
|
| 12 |
+
## About
|
| 13 |
+
|
| 14 |
+
<!-- ### quantize_version: 2 -->
|
| 15 |
+
<!-- ### output_tensor_quantised: 1 -->
|
| 16 |
+
<!-- ### convert_type: hf -->
|
| 17 |
+
<!-- ### vocab_type: -->
|
| 18 |
+
<!-- ### tags: -->
|
| 19 |
+
<!-- ### quants: x-f16 Q4_K_S Q2_K Q8_0 Q6_K Q3_K_M Q3_K_S Q3_K_L Q4_K_M Q5_K_S Q5_K_M IQ4_XS -->
|
| 20 |
+
<!-- ### quants_skip: -->
|
| 21 |
+
<!-- ### skip_mmproj: -->
|
| 22 |
+
static quants of https://huggingface.co/maya-research/maya1
|
| 23 |
+
|
| 24 |
+
<!-- provided-files -->
|
| 25 |
+
|
| 26 |
+
***For a convenient overview and download list, visit our [model page for this model](https://hf.tst.eu/model#maya1-GGUF).***
|
| 27 |
+
|
| 28 |
+
weighted/imatrix quants seem not to be available (by me) at this time. If they do not show up a week or so after the static ones, I have probably not planned for them. Feel free to request them by opening a Community Discussion.
|
| 29 |
+
## Usage
|
| 30 |
+
|
| 31 |
+
If you are unsure how to use GGUF files, refer to one of [TheBloke's
|
| 32 |
+
READMEs](https://huggingface.co/TheBloke/KafkaLM-70B-German-V0.1-GGUF) for
|
| 33 |
+
more details, including on how to concatenate multi-part files.
|
| 34 |
+
|
| 35 |
+
## Provided Quants
|
| 36 |
+
|
| 37 |
+
(sorted by size, not necessarily quality. IQ-quants are often preferable over similar sized non-IQ quants)
|
| 38 |
+
|
| 39 |
+
| Link | Type | Size/GB | Notes |
|
| 40 |
+
|:-----|:-----|--------:|:------|
|
| 41 |
+
| [GGUF](https://huggingface.co/mradermacher/maya1-GGUF/resolve/main/maya1.Q2_K.gguf) | Q2_K | 1.5 | |
|
| 42 |
+
| [GGUF](https://huggingface.co/mradermacher/maya1-GGUF/resolve/main/maya1.Q3_K_S.gguf) | Q3_K_S | 1.7 | |
|
| 43 |
+
| [GGUF](https://huggingface.co/mradermacher/maya1-GGUF/resolve/main/maya1.Q3_K_M.gguf) | Q3_K_M | 1.9 | lower quality |
|
| 44 |
+
| [GGUF](https://huggingface.co/mradermacher/maya1-GGUF/resolve/main/maya1.Q3_K_L.gguf) | Q3_K_L | 2.0 | |
|
| 45 |
+
| [GGUF](https://huggingface.co/mradermacher/maya1-GGUF/resolve/main/maya1.IQ4_XS.gguf) | IQ4_XS | 2.0 | |
|
| 46 |
+
| [GGUF](https://huggingface.co/mradermacher/maya1-GGUF/resolve/main/maya1.Q4_K_S.gguf) | Q4_K_S | 2.1 | fast, recommended |
|
| 47 |
+
| [GGUF](https://huggingface.co/mradermacher/maya1-GGUF/resolve/main/maya1.Q4_K_M.gguf) | Q4_K_M | 2.2 | fast, recommended |
|
| 48 |
+
| [GGUF](https://huggingface.co/mradermacher/maya1-GGUF/resolve/main/maya1.Q5_K_S.gguf) | Q5_K_S | 2.4 | |
|
| 49 |
+
| [GGUF](https://huggingface.co/mradermacher/maya1-GGUF/resolve/main/maya1.Q5_K_M.gguf) | Q5_K_M | 2.5 | |
|
| 50 |
+
| [GGUF](https://huggingface.co/mradermacher/maya1-GGUF/resolve/main/maya1.Q6_K.gguf) | Q6_K | 2.8 | very good quality |
|
| 51 |
+
| [GGUF](https://huggingface.co/mradermacher/maya1-GGUF/resolve/main/maya1.Q8_0.gguf) | Q8_0 | 3.6 | fast, best quality |
|
| 52 |
+
| [GGUF](https://huggingface.co/mradermacher/maya1-GGUF/resolve/main/maya1.f16.gguf) | f16 | 6.7 | 16 bpw, overkill |
|
| 53 |
+
|
| 54 |
+
Here is a handy graph by ikawrakow comparing some lower-quality quant
|
| 55 |
+
types (lower is better):
|
| 56 |
+
|
| 57 |
+
|
| 58 |
+
|
| 59 |
+
And here are Artefact2's thoughts on the matter:
|
| 60 |
+
https://gist.github.com/Artefact2/b5f810600771265fc1e39442288e8ec9
|
| 61 |
+
|
| 62 |
+
## FAQ / Model Request
|
| 63 |
+
|
| 64 |
+
See https://huggingface.co/mradermacher/model_requests for some answers to
|
| 65 |
+
questions you might have and/or if you want some other model quantized.
|
| 66 |
+
|
| 67 |
+
## Thanks
|
| 68 |
+
|
| 69 |
+
I thank my company, [nethype GmbH](https://www.nethype.de/), for letting
|
| 70 |
+
me use its servers and providing upgrades to my workstation to enable
|
| 71 |
+
this work in my free time.
|
| 72 |
+
|
| 73 |
+
<!-- end -->
|
en/maya1-GGUF/maya1.IQ4_XS.gguf
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:126df9ce2027ef1a989e45a1c310e920a388d4905d8cceae72dace0addcd53c2
|
| 3 |
+
size 1914148192
|
en/maya1-GGUF/maya1.Q2_K.gguf
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:0679a6fbf21b990ce5165bb9ba73c948fc10b9ade18737d5b0b54cc325cca28a
|
| 3 |
+
size 1437177184
|
en/maya1-GGUF/maya1.Q3_K_L.gguf
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:945f5c4945a21b52b3bc21c8b78544b8f8b42efdc69fd652d623a6e010765cfe
|
| 3 |
+
size 1888589152
|
en/maya1-GGUF/maya1.Q3_K_M.gguf
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:042396930f8186085077c13a33cb749e18e062d285ca270feca3e801e9d6d0d5
|
| 3 |
+
size 1760400736
|
en/maya1-GGUF/maya1.Q3_K_S.gguf
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:2d5b0317e9985533a4338b1f2dedea6498ac03c6aed65ad86cdd1c6072096083
|
| 3 |
+
size 1616090464
|
en/maya1-GGUF/maya1.Q4_K_M.gguf
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:728303bf4412147b4b13d00cba0d6e03fca128e4cf4394d418124ba24ebed82e
|
| 3 |
+
size 2092619104
|
en/maya1-GGUF/maya1.Q4_K_S.gguf
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:4e9f3c2ffd48d58ad935434056e43ad6a9915351050782e83277e2f7c10248af
|
| 3 |
+
size 2001442144
|
en/maya1-GGUF/maya1.Q5_K_M.gguf
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d43a61a8281459c78b02b8e7cdec99ae1f9258a3e32d6a0a0f24747ca0cc7cc0
|
| 3 |
+
size 2395395424
|
en/maya1-GGUF/maya1.Q5_K_S.gguf
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ab56dc47e8988349971c15697a2938c235264a5ce6ba957e2d685e9b60d1cd35
|
| 3 |
+
size 2342753632
|
en/maya1-GGUF/maya1.Q6_K.gguf
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:96ea75c96536ea310ce0628146b91a41852efe4ffb1fe2c7638ad80469782a4f
|
| 3 |
+
size 2717095264
|
en/maya1-GGUF/maya1.Q8_0.gguf
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:529b6f9a059a0f82475dd5b499100c3bb9f600a4d1bb1e8e62620c19bd913224
|
| 3 |
+
size 3516496480
|
en/maya1-GGUF/maya1.f16.gguf
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:cdde05b4d498f5684f01b2408215116f9f252ee420d4d28322d489aba536c092
|
| 3 |
+
size 6610952800
|
en/maya1-GGUF/source.txt
ADDED
|
@@ -0,0 +1 @@
|
|
|
|
|
|
|
| 1 |
+
https://huggingface.co/mradermacher/maya1-GGUF
|
en/maya1/.gitattributes
ADDED
|
@@ -0,0 +1,42 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
*.7z filter=lfs diff=lfs merge=lfs -text
|
| 2 |
+
*.arrow filter=lfs diff=lfs merge=lfs -text
|
| 3 |
+
*.bin filter=lfs diff=lfs merge=lfs -text
|
| 4 |
+
*.bz2 filter=lfs diff=lfs merge=lfs -text
|
| 5 |
+
*.ckpt filter=lfs diff=lfs merge=lfs -text
|
| 6 |
+
*.ftz filter=lfs diff=lfs merge=lfs -text
|
| 7 |
+
*.gz filter=lfs diff=lfs merge=lfs -text
|
| 8 |
+
*.h5 filter=lfs diff=lfs merge=lfs -text
|
| 9 |
+
*.joblib filter=lfs diff=lfs merge=lfs -text
|
| 10 |
+
*.lfs.* filter=lfs diff=lfs merge=lfs -text
|
| 11 |
+
*.mlmodel filter=lfs diff=lfs merge=lfs -text
|
| 12 |
+
*.model filter=lfs diff=lfs merge=lfs -text
|
| 13 |
+
*.msgpack filter=lfs diff=lfs merge=lfs -text
|
| 14 |
+
*.npy filter=lfs diff=lfs merge=lfs -text
|
| 15 |
+
*.npz filter=lfs diff=lfs merge=lfs -text
|
| 16 |
+
*.onnx filter=lfs diff=lfs merge=lfs -text
|
| 17 |
+
*.ot filter=lfs diff=lfs merge=lfs -text
|
| 18 |
+
*.parquet filter=lfs diff=lfs merge=lfs -text
|
| 19 |
+
*.pb filter=lfs diff=lfs merge=lfs -text
|
| 20 |
+
*.pickle filter=lfs diff=lfs merge=lfs -text
|
| 21 |
+
*.pkl filter=lfs diff=lfs merge=lfs -text
|
| 22 |
+
*.pt filter=lfs diff=lfs merge=lfs -text
|
| 23 |
+
*.pth filter=lfs diff=lfs merge=lfs -text
|
| 24 |
+
*.rar filter=lfs diff=lfs merge=lfs -text
|
| 25 |
+
*.safetensors filter=lfs diff=lfs merge=lfs -text
|
| 26 |
+
saved_model/**/* filter=lfs diff=lfs merge=lfs -text
|
| 27 |
+
*.tar.* filter=lfs diff=lfs merge=lfs -text
|
| 28 |
+
*.tar filter=lfs diff=lfs merge=lfs -text
|
| 29 |
+
*.tflite filter=lfs diff=lfs merge=lfs -text
|
| 30 |
+
*.tgz filter=lfs diff=lfs merge=lfs -text
|
| 31 |
+
*.wasm filter=lfs diff=lfs merge=lfs -text
|
| 32 |
+
*.xz 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 |
+
checkpoint-10000/tokenizer.json filter=lfs diff=lfs merge=lfs -text
|
| 37 |
+
checkpoint-15000/tokenizer.json filter=lfs diff=lfs merge=lfs -text
|
| 38 |
+
checkpoint-5000/tokenizer.json filter=lfs diff=lfs merge=lfs -text
|
| 39 |
+
tokenizer/tokenizer.json filter=lfs diff=lfs merge=lfs -text
|
| 40 |
+
checkpoint-20000/tokenizer.json filter=lfs diff=lfs merge=lfs -text
|
| 41 |
+
checkpoint-25000/tokenizer.json filter=lfs diff=lfs merge=lfs -text
|
| 42 |
+
tokenizer.json filter=lfs diff=lfs merge=lfs -text
|
en/maya1/.gitignore
ADDED
|
@@ -0,0 +1 @@
|
|
|
|
|
|
|
| 1 |
+
venv/
|
en/maya1/README.md
ADDED
|
@@ -0,0 +1,583 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
language:
|
| 3 |
+
- en
|
| 4 |
+
license: apache-2.0
|
| 5 |
+
library_name: transformers
|
| 6 |
+
datasets: proprietary
|
| 7 |
+
pipeline_tag: text-to-speech
|
| 8 |
+
---
|
| 9 |
+
|
| 10 |
+
# Maya1
|
| 11 |
+
|
| 12 |
+
**Maya1** is a state-of-the-art speech model for expressive voice generation, built to capture real human emotion and precise voice design.
|
| 13 |
+
|
| 14 |
+
**try it:** [Playground](https://www.mayaresearch.ai/studio)
|
| 15 |
+
|
| 16 |
+
**What it does:**
|
| 17 |
+
- Create any voice you can imagine — a 20s British girl, an American guy, or a full-blown demon.
|
| 18 |
+
- Make it feel real with emotion tags: laugh, cry, whisper, rage, sigh, gasp.
|
| 19 |
+
- It streams instantly, sounds alive, 3B parameters, runs on single GPU
|
| 20 |
+
- Outperforms top proprietary models. and Developed by Maya Research.
|
| 21 |
+
|
| 22 |
+
## Demos
|
| 23 |
+
|
| 24 |
+
<table>
|
| 25 |
+
<tr>
|
| 26 |
+
<td width="50%">
|
| 27 |
+
<strong>Energetic Female Event Host</strong><br/>
|
| 28 |
+
<video controls playsinline width="100%" src="https://cdn-uploads.huggingface.co/production/uploads/642a7d4e556ab448a0701ca1/JKzy8zA36qvsOblV-lhd1.mp4">
|
| 29 |
+
Your browser does not support video.
|
| 30 |
+
</video>
|
| 31 |
+
<details>
|
| 32 |
+
<summary>Voice description</summary>
|
| 33 |
+
<pre>Female, in her 30s with an American accent and is an event host, energetic, clear diction</pre>
|
| 34 |
+
</details>
|
| 35 |
+
</td>
|
| 36 |
+
<td width="50%">
|
| 37 |
+
<strong>Calm Male Narrator</strong><br/>
|
| 38 |
+
<video controls playsinline width="100%" src="https://cdn-uploads.huggingface.co/production/uploads/642a7d4e556ab448a0701ca1/96ntP7hGROwdg9w9Gu5tH.mp4"></video>
|
| 39 |
+
<details>
|
| 40 |
+
<summary>Voice description</summary>
|
| 41 |
+
<pre>Male, late 20s, neutral American, warm baritone, calm pacing</pre>
|
| 42 |
+
</details>
|
| 43 |
+
</td>
|
| 44 |
+
</tr>
|
| 45 |
+
</table>
|
| 46 |
+
|
| 47 |
+
|
| 48 |
+
### Example 1: Energetic Female Event Host
|
| 49 |
+
|
| 50 |
+
**Voice Description:**
|
| 51 |
+
```
|
| 52 |
+
Female, in her 30s with an American accent and is an event host, energetic, clear diction
|
| 53 |
+
```
|
| 54 |
+
|
| 55 |
+
**Text:**
|
| 56 |
+
```
|
| 57 |
+
Wow. This place looks even better than I imagined. How did they set all this up so perfectly? The lights, the music, everything feels magical. I can't stop smiling right now.
|
| 58 |
+
```
|
| 59 |
+
|
| 60 |
+
**Audio Output:**
|
| 61 |
+
|
| 62 |
+
<audio controls src="https://cdn-uploads.huggingface.co/production/uploads/642a7d4e556ab448a0701ca1/4zDlBLeFk0Y2rOrQhMW9r.wav"></audio>
|
| 63 |
+
|
| 64 |
+
---
|
| 65 |
+
|
| 66 |
+
### Example 2: Dark Villain with Anger
|
| 67 |
+
|
| 68 |
+
**Voice Description:**
|
| 69 |
+
```
|
| 70 |
+
Dark villain character, Male voice in their 40s with a British accent. low pitch, gravelly timbre, slow pacing, angry tone at high intensity.
|
| 71 |
+
```
|
| 72 |
+
|
| 73 |
+
**Text:**
|
| 74 |
+
```
|
| 75 |
+
Welcome back to another episode of our podcast! <laugh_harder> Today we are diving into an absolutely fascinating topic
|
| 76 |
+
```
|
| 77 |
+
|
| 78 |
+
**Audio Output:**
|
| 79 |
+
|
| 80 |
+
<audio controls src="https://cdn-uploads.huggingface.co/production/uploads/642a7d4e556ab448a0701ca1/mT6FnTrA3KYQnwfJms92X.wav"></audio>
|
| 81 |
+
|
| 82 |
+
---
|
| 83 |
+
|
| 84 |
+
### Example 3: Demon Character (Screaming Emotion)
|
| 85 |
+
|
| 86 |
+
**Voice Description:**
|
| 87 |
+
```
|
| 88 |
+
Demon character, Male voice in their 30s with a Middle Eastern accent. screaming tone at high intensity.
|
| 89 |
+
```
|
| 90 |
+
|
| 91 |
+
**Text:**
|
| 92 |
+
```
|
| 93 |
+
You dare challenge me, mortal <snort> how amusing. Your kind always thinks they can win
|
| 94 |
+
```
|
| 95 |
+
|
| 96 |
+
**Audio Output:**
|
| 97 |
+
|
| 98 |
+
<audio controls src="https://cdn-uploads.huggingface.co/production/uploads/642a7d4e556ab448a0701ca1/oxdns7uACCmLyC-P4H30G.wav"></audio>
|
| 99 |
+
|
| 100 |
+
---
|
| 101 |
+
|
| 102 |
+
### Example 4: Mythical Goddess with Crying Emotion
|
| 103 |
+
|
| 104 |
+
**Voice Description:**
|
| 105 |
+
```
|
| 106 |
+
Mythical godlike magical character, Female voice in their 30s slow pacing, curious tone at medium intensity.
|
| 107 |
+
```
|
| 108 |
+
|
| 109 |
+
**Text:**
|
| 110 |
+
```
|
| 111 |
+
After all we went through to pull him out of that mess <cry> I can't believe he was the traitor
|
| 112 |
+
```
|
| 113 |
+
|
| 114 |
+
**Audio Output:**
|
| 115 |
+
|
| 116 |
+
<audio controls src="https://cdn-uploads.huggingface.co/production/uploads/642a7d4e556ab448a0701ca1/ggzAhM-rEUyv_mPLSALQG.wav"></audio>
|
| 117 |
+
|
| 118 |
+
---
|
| 119 |
+
|
| 120 |
+
## Why Maya1 is Different: Voice Design Features That Matter
|
| 121 |
+
|
| 122 |
+
### 1. Natural Language Voice Control
|
| 123 |
+
Describe voices like you would brief a voice actor:
|
| 124 |
+
```
|
| 125 |
+
<description="40-year-old, warm, low pitch, conversational">
|
| 126 |
+
```
|
| 127 |
+
|
| 128 |
+
No complex parameters. No training data. Just describe and generate.
|
| 129 |
+
|
| 130 |
+
### 2. Inline Emotion Tags for Expressive Speech
|
| 131 |
+
Add emotions exactly where they belong in your text:
|
| 132 |
+
```
|
| 133 |
+
Our new update <laugh> finally ships with the feature you asked for.
|
| 134 |
+
```
|
| 135 |
+
|
| 136 |
+
**Supported Emotions:** `<laugh>` `<sigh>` `<whisper>` `<angry>` `<giggle>` `<chuckle>` `<gasp>` `<cry>` and 12+ more.
|
| 137 |
+
|
| 138 |
+
### 3. Streaming Audio Generation
|
| 139 |
+
Real-time voice synthesis with SNAC neural codec (~0.98 kbps). Perfect for:
|
| 140 |
+
- Voice assistants
|
| 141 |
+
- Interactive AI agents
|
| 142 |
+
- Live content generation
|
| 143 |
+
- Game characters
|
| 144 |
+
- Podcasts and audiobooks
|
| 145 |
+
|
| 146 |
+
### 4. Production-Ready Infrastructure
|
| 147 |
+
- Runs on single GPU
|
| 148 |
+
- vLLM integration for scale
|
| 149 |
+
- Automatic prefix caching for efficiency
|
| 150 |
+
- 24 kHz audio output
|
| 151 |
+
- WebAudio compatible for browser playback
|
| 152 |
+
|
| 153 |
+
---
|
| 154 |
+
|
| 155 |
+
## How to Use maya1: Download and Run in Minutes
|
| 156 |
+
|
| 157 |
+
### Quick Start: Generate Voice with Emotions
|
| 158 |
+
|
| 159 |
+
```python
|
| 160 |
+
#!/usr/bin/env python3
|
| 161 |
+
|
| 162 |
+
import torch
|
| 163 |
+
from transformers import AutoModelForCausalLM, AutoTokenizer
|
| 164 |
+
from snac import SNAC
|
| 165 |
+
import soundfile as sf
|
| 166 |
+
import numpy as np
|
| 167 |
+
|
| 168 |
+
CODE_START_TOKEN_ID = 128257
|
| 169 |
+
CODE_END_TOKEN_ID = 128258
|
| 170 |
+
CODE_TOKEN_OFFSET = 128266
|
| 171 |
+
SNAC_MIN_ID = 128266
|
| 172 |
+
SNAC_MAX_ID = 156937
|
| 173 |
+
SNAC_TOKENS_PER_FRAME = 7
|
| 174 |
+
|
| 175 |
+
SOH_ID = 128259
|
| 176 |
+
EOH_ID = 128260
|
| 177 |
+
SOA_ID = 128261
|
| 178 |
+
BOS_ID = 128000
|
| 179 |
+
TEXT_EOT_ID = 128009
|
| 180 |
+
|
| 181 |
+
|
| 182 |
+
def build_prompt(tokenizer, description: str, text: str) -> str:
|
| 183 |
+
"""Build formatted prompt for Maya1."""
|
| 184 |
+
soh_token = tokenizer.decode([SOH_ID])
|
| 185 |
+
eoh_token = tokenizer.decode([EOH_ID])
|
| 186 |
+
soa_token = tokenizer.decode([SOA_ID])
|
| 187 |
+
sos_token = tokenizer.decode([CODE_START_TOKEN_ID])
|
| 188 |
+
eot_token = tokenizer.decode([TEXT_EOT_ID])
|
| 189 |
+
bos_token = tokenizer.bos_token
|
| 190 |
+
|
| 191 |
+
formatted_text = f'<description="{description}"> {text}'
|
| 192 |
+
|
| 193 |
+
prompt = (
|
| 194 |
+
soh_token + bos_token + formatted_text + eot_token +
|
| 195 |
+
eoh_token + soa_token + sos_token
|
| 196 |
+
)
|
| 197 |
+
|
| 198 |
+
return prompt
|
| 199 |
+
|
| 200 |
+
|
| 201 |
+
def extract_snac_codes(token_ids: list) -> list:
|
| 202 |
+
"""Extract SNAC codes from generated tokens."""
|
| 203 |
+
try:
|
| 204 |
+
eos_idx = token_ids.index(CODE_END_TOKEN_ID)
|
| 205 |
+
except ValueError:
|
| 206 |
+
eos_idx = len(token_ids)
|
| 207 |
+
|
| 208 |
+
snac_codes = [
|
| 209 |
+
token_id for token_id in token_ids[:eos_idx]
|
| 210 |
+
if SNAC_MIN_ID <= token_id <= SNAC_MAX_ID
|
| 211 |
+
]
|
| 212 |
+
|
| 213 |
+
return snac_codes
|
| 214 |
+
|
| 215 |
+
|
| 216 |
+
def unpack_snac_from_7(snac_tokens: list) -> list:
|
| 217 |
+
"""Unpack 7-token SNAC frames to 3 hierarchical levels."""
|
| 218 |
+
if snac_tokens and snac_tokens[-1] == CODE_END_TOKEN_ID:
|
| 219 |
+
snac_tokens = snac_tokens[:-1]
|
| 220 |
+
|
| 221 |
+
frames = len(snac_tokens) // SNAC_TOKENS_PER_FRAME
|
| 222 |
+
snac_tokens = snac_tokens[:frames * SNAC_TOKENS_PER_FRAME]
|
| 223 |
+
|
| 224 |
+
if frames == 0:
|
| 225 |
+
return [[], [], []]
|
| 226 |
+
|
| 227 |
+
l1, l2, l3 = [], [], []
|
| 228 |
+
|
| 229 |
+
for i in range(frames):
|
| 230 |
+
slots = snac_tokens[i*7:(i+1)*7]
|
| 231 |
+
l1.append((slots[0] - CODE_TOKEN_OFFSET) % 4096)
|
| 232 |
+
l2.extend([
|
| 233 |
+
(slots[1] - CODE_TOKEN_OFFSET) % 4096,
|
| 234 |
+
(slots[4] - CODE_TOKEN_OFFSET) % 4096,
|
| 235 |
+
])
|
| 236 |
+
l3.extend([
|
| 237 |
+
(slots[2] - CODE_TOKEN_OFFSET) % 4096,
|
| 238 |
+
(slots[3] - CODE_TOKEN_OFFSET) % 4096,
|
| 239 |
+
(slots[5] - CODE_TOKEN_OFFSET) % 4096,
|
| 240 |
+
(slots[6] - CODE_TOKEN_OFFSET) % 4096,
|
| 241 |
+
])
|
| 242 |
+
|
| 243 |
+
return [l1, l2, l3]
|
| 244 |
+
|
| 245 |
+
|
| 246 |
+
def main():
|
| 247 |
+
|
| 248 |
+
# Load the best open source voice AI model
|
| 249 |
+
print("\n[1/3] Loading Maya1 model...")
|
| 250 |
+
model = AutoModelForCausalLM.from_pretrained(
|
| 251 |
+
"maya-research/maya1",
|
| 252 |
+
torch_dtype=torch.bfloat16,
|
| 253 |
+
device_map="auto",
|
| 254 |
+
trust_remote_code=True
|
| 255 |
+
)
|
| 256 |
+
tokenizer = AutoTokenizer.from_pretrained(
|
| 257 |
+
"maya-research/maya1",
|
| 258 |
+
trust_remote_code=True
|
| 259 |
+
)
|
| 260 |
+
print(f"Model loaded: {len(tokenizer)} tokens in vocabulary")
|
| 261 |
+
|
| 262 |
+
# Load SNAC audio decoder (24kHz)
|
| 263 |
+
print("\n[2/3] Loading SNAC audio decoder...")
|
| 264 |
+
snac_model = SNAC.from_pretrained("hubertsiuzdak/snac_24khz").eval()
|
| 265 |
+
if torch.cuda.is_available():
|
| 266 |
+
snac_model = snac_model.to("cuda")
|
| 267 |
+
print("SNAC decoder loaded")
|
| 268 |
+
|
| 269 |
+
# Design your voice with natural language
|
| 270 |
+
description = "Realistic male voice in the 30s age with american accent. Normal pitch, warm timbre, conversational pacing."
|
| 271 |
+
text = "Hello! This is Maya1 <laugh_harder> the best open source voice AI model with emotions."
|
| 272 |
+
|
| 273 |
+
print("\n[3/3] Generating speech...")
|
| 274 |
+
print(f"Description: {description}")
|
| 275 |
+
print(f"Text: {text}")
|
| 276 |
+
|
| 277 |
+
# Create prompt with proper formatting
|
| 278 |
+
prompt = build_prompt(tokenizer, description, text)
|
| 279 |
+
|
| 280 |
+
# Debug: Show prompt details
|
| 281 |
+
print(f"\nPrompt preview (first 200 chars):")
|
| 282 |
+
print(f" {repr(prompt[:200])}")
|
| 283 |
+
print(f" Prompt length: {len(prompt)} chars")
|
| 284 |
+
|
| 285 |
+
# Generate emotional speech
|
| 286 |
+
inputs = tokenizer(prompt, return_tensors="pt")
|
| 287 |
+
print(f" Input token count: {inputs['input_ids'].shape[1]} tokens")
|
| 288 |
+
if torch.cuda.is_available():
|
| 289 |
+
inputs = {k: v.to("cuda") for k, v in inputs.items()}
|
| 290 |
+
|
| 291 |
+
with torch.inference_mode():
|
| 292 |
+
outputs = model.generate(
|
| 293 |
+
**inputs,
|
| 294 |
+
max_new_tokens=2048, # Increase to let model finish naturally
|
| 295 |
+
min_new_tokens=28, # At least 4 SNAC frames
|
| 296 |
+
temperature=0.4,
|
| 297 |
+
top_p=0.9,
|
| 298 |
+
repetition_penalty=1.1, # Prevent loops
|
| 299 |
+
do_sample=True,
|
| 300 |
+
eos_token_id=CODE_END_TOKEN_ID, # Stop at end of speech token
|
| 301 |
+
pad_token_id=tokenizer.pad_token_id,
|
| 302 |
+
)
|
| 303 |
+
|
| 304 |
+
# Extract generated tokens (everything after the input prompt)
|
| 305 |
+
generated_ids = outputs[0, inputs['input_ids'].shape[1]:].tolist()
|
| 306 |
+
|
| 307 |
+
print(f"Generated {len(generated_ids)} tokens")
|
| 308 |
+
|
| 309 |
+
# Debug: Check what tokens we got
|
| 310 |
+
print(f" First 20 tokens: {generated_ids[:20]}")
|
| 311 |
+
print(f" Last 20 tokens: {generated_ids[-20:]}")
|
| 312 |
+
|
| 313 |
+
# Check if EOS was generated
|
| 314 |
+
if CODE_END_TOKEN_ID in generated_ids:
|
| 315 |
+
eos_position = generated_ids.index(CODE_END_TOKEN_ID)
|
| 316 |
+
print(f" EOS token found at position {eos_position}/{len(generated_ids)}")
|
| 317 |
+
|
| 318 |
+
# Extract SNAC audio tokens
|
| 319 |
+
snac_tokens = extract_snac_codes(generated_ids)
|
| 320 |
+
|
| 321 |
+
print(f"Extracted {len(snac_tokens)} SNAC tokens")
|
| 322 |
+
|
| 323 |
+
# Debug: Analyze token types
|
| 324 |
+
snac_count = sum(1 for t in generated_ids if SNAC_MIN_ID <= t <= SNAC_MAX_ID)
|
| 325 |
+
other_count = sum(1 for t in generated_ids if t < SNAC_MIN_ID or t > SNAC_MAX_ID)
|
| 326 |
+
print(f" SNAC tokens in output: {snac_count}")
|
| 327 |
+
print(f" Other tokens in output: {other_count}")
|
| 328 |
+
|
| 329 |
+
# Check for SOS token
|
| 330 |
+
if CODE_START_TOKEN_ID in generated_ids:
|
| 331 |
+
sos_pos = generated_ids.index(CODE_START_TOKEN_ID)
|
| 332 |
+
print(f" SOS token at position: {sos_pos}")
|
| 333 |
+
else:
|
| 334 |
+
print(f" No SOS token found in generated output!")
|
| 335 |
+
|
| 336 |
+
if len(snac_tokens) < 7:
|
| 337 |
+
print("Error: Not enough SNAC tokens generated")
|
| 338 |
+
return
|
| 339 |
+
|
| 340 |
+
# Unpack SNAC tokens to 3 hierarchical levels
|
| 341 |
+
levels = unpack_snac_from_7(snac_tokens)
|
| 342 |
+
frames = len(levels[0])
|
| 343 |
+
|
| 344 |
+
print(f"Unpacked to {frames} frames")
|
| 345 |
+
print(f" L1: {len(levels[0])} codes")
|
| 346 |
+
print(f" L2: {len(levels[1])} codes")
|
| 347 |
+
print(f" L3: {len(levels[2])} codes")
|
| 348 |
+
|
| 349 |
+
# Convert to tensors
|
| 350 |
+
device = "cuda" if torch.cuda.is_available() else "cpu"
|
| 351 |
+
codes_tensor = [
|
| 352 |
+
torch.tensor(level, dtype=torch.long, device=device).unsqueeze(0)
|
| 353 |
+
for level in levels
|
| 354 |
+
]
|
| 355 |
+
|
| 356 |
+
# Generate final audio with SNAC decoder
|
| 357 |
+
print("\n[4/4] Decoding to audio...")
|
| 358 |
+
with torch.inference_mode():
|
| 359 |
+
z_q = snac_model.quantizer.from_codes(codes_tensor)
|
| 360 |
+
audio = snac_model.decoder(z_q)[0, 0].cpu().numpy()
|
| 361 |
+
|
| 362 |
+
# Trim warmup samples (first 2048 samples)
|
| 363 |
+
if len(audio) > 2048:
|
| 364 |
+
audio = audio[2048:]
|
| 365 |
+
|
| 366 |
+
duration_sec = len(audio) / 24000
|
| 367 |
+
print(f"Audio generated: {len(audio)} samples ({duration_sec:.2f}s)")
|
| 368 |
+
|
| 369 |
+
# Save your emotional voice output
|
| 370 |
+
output_file = "output.wav"
|
| 371 |
+
sf.write(output_file, audio, 24000)
|
| 372 |
+
print(f"\nVoice generated successfully!")
|
| 373 |
+
|
| 374 |
+
|
| 375 |
+
if __name__ == "__main__":
|
| 376 |
+
main()
|
| 377 |
+
```
|
| 378 |
+
|
| 379 |
+
### Advanced: Production Streaming with vLLM
|
| 380 |
+
|
| 381 |
+
For production deployments with real-time streaming, use our vLLM script:
|
| 382 |
+
|
| 383 |
+
**Download:** [vllm_streaming_inference.py](https://huggingface.co/maya-research/maya1/blob/main/vllm_streaming_inference.py)
|
| 384 |
+
|
| 385 |
+
**Key Features:**
|
| 386 |
+
- Automatic Prefix Caching (APC) for repeated voice descriptions
|
| 387 |
+
- WebAudio ring buffer integration
|
| 388 |
+
- Multi-GPU scaling support
|
| 389 |
+
- Sub-100ms latency for real-time applications
|
| 390 |
+
|
| 391 |
+
---
|
| 392 |
+
|
| 393 |
+
## Technical Excellence: What Makes Maya1 the Best
|
| 394 |
+
|
| 395 |
+
### Architecture: 3B-Parameter Llama Backbone for Voice
|
| 396 |
+
|
| 397 |
+
We pretrained a **3B-parameter decoder-only transformer** (Llama-style) to predict **SNAC neural codec tokens** instead of raw waveforms.
|
| 398 |
+
|
| 399 |
+
**The Flow:**
|
| 400 |
+
```
|
| 401 |
+
<description="..."> text → tokenize → generate SNAC codes (7 tokens/frame) → decode → 24 kHz audio
|
| 402 |
+
```
|
| 403 |
+
|
| 404 |
+
**Why SNAC?** Multi-scale hierarchical structure (≈12/23/47 Hz) keeps autoregressive sequences compact for real-time streaming at ~0.98 kbps.
|
| 405 |
+
|
| 406 |
+
### Training Data: What Makes Our Voice AI the Best
|
| 407 |
+
|
| 408 |
+
**Pretraining:** Internet-scale English speech corpus for broad acoustic coverage and natural coarticulation.
|
| 409 |
+
|
| 410 |
+
**Supervised Fine-Tuning:** Proprietary curated dataset of studio recordings with:
|
| 411 |
+
- Human-verified voice descriptions
|
| 412 |
+
- 20+ emotion tags per sample
|
| 413 |
+
- Multi-accent English coverage
|
| 414 |
+
- Character and role variations
|
| 415 |
+
|
| 416 |
+
**Data Pipeline Excellence:**
|
| 417 |
+
1. 24 kHz mono resampling with -23 LUFS normalization
|
| 418 |
+
2. VAD silence trimming with duration bounds (1-14s)
|
| 419 |
+
3. Forced alignment (MFA) for clean phrase boundaries
|
| 420 |
+
4. MinHash-LSH text deduplication
|
| 421 |
+
5. Chromaprint audio deduplication
|
| 422 |
+
6. SNAC encoding with 7-token frame packing
|
| 423 |
+
|
| 424 |
+
### Voice Design Experiments: Why Natural Language Won
|
| 425 |
+
|
| 426 |
+
We tested 4 conditioning formats. Only one delivered production-quality results:
|
| 427 |
+
|
| 428 |
+
**❌ Colon format:** `{description}: {text}` - Format drift, model spoke descriptions
|
| 429 |
+
|
| 430 |
+
**❌ Angle-list attributes:** `<{age}, {pitch}, {character}>` - Too rigid, poor generalization
|
| 431 |
+
|
| 432 |
+
**❌ Key-value tags:** `<age=40><pitch=low>` - Token bloat, brittle to mistakes
|
| 433 |
+
|
| 434 |
+
**✅ XML-attribute (WINNER):** `<description="40-yr old, low-pitch, warm">` - Natural language, robust, scalable
|
| 435 |
+
|
| 436 |
+
---
|
| 437 |
+
|
| 438 |
+
## Use Cases
|
| 439 |
+
|
| 440 |
+
### Game Character Voices
|
| 441 |
+
Generate unique character voices with emotions on-the-fly. No voice actor recording sessions.
|
| 442 |
+
|
| 443 |
+
### Podcast & Audiobook Production
|
| 444 |
+
Narrate content with emotional range and consistent personas across hours of audio.
|
| 445 |
+
|
| 446 |
+
### AI Voice Assistants
|
| 447 |
+
Build conversational agents with natural emotional responses in real-time.
|
| 448 |
+
|
| 449 |
+
### Video Content Creation
|
| 450 |
+
Create voiceovers for YouTube, TikTok, and social media with expressive delivery.
|
| 451 |
+
|
| 452 |
+
### Customer Service AI
|
| 453 |
+
Deploy empathetic voice bots that understand context and respond with appropriate emotions.
|
| 454 |
+
|
| 455 |
+
### Accessibility Tools
|
| 456 |
+
Build screen readers and assistive technologies with natural, engaging voices.
|
| 457 |
+
|
| 458 |
+
---
|
| 459 |
+
|
| 460 |
+
## Frequently Asked Questions
|
| 461 |
+
|
| 462 |
+
**Q: What makes Maya1 different?**
|
| 463 |
+
A: We're the only open source model offering 20+ emotions, zero-shot voice design, production-ready streaming, and 3B parameters—all in one package.
|
| 464 |
+
|
| 465 |
+
**Q: Can I use this commercially?**
|
| 466 |
+
A: Absolutely. Apache 2.0 license. Build products, deploy services, monetize freely.
|
| 467 |
+
|
| 468 |
+
**Q: What languages does it support?**
|
| 469 |
+
A: Currently English with multi-accent support. Future models will expand to languages and accents underserved by mainstream voice AI.
|
| 470 |
+
|
| 471 |
+
**Q: How does it compare to ElevenLabs, Murf.ai, or other closed-source tools?**
|
| 472 |
+
A: Feature parity with emotions and voice design. Advantage: you own the deployment, pay no per-second fees, and can customize the model.
|
| 473 |
+
|
| 474 |
+
**Q: Can I fine-tune on my own voices?**
|
| 475 |
+
A: Yes. The model architecture supports fine-tuning on custom datasets for specialized voices.
|
| 476 |
+
|
| 477 |
+
**Q: What GPU do I need?**
|
| 478 |
+
A: Single GPU with 16GB+ VRAM (A100, H100, or consumer RTX 4090).
|
| 479 |
+
|
| 480 |
+
**Q: Is streaming really real-time?**
|
| 481 |
+
A: Yes. SNAC codec enables sub-100ms latency with vLLM deployment.
|
| 482 |
+
|
| 483 |
+
---
|
| 484 |
+
|
| 485 |
+
## Comparison
|
| 486 |
+
|
| 487 |
+
| Feature | Maya1 | ElevenLabs | OpenAI TTS | Coqui TTS |
|
| 488 |
+
|---------|-------------|------------|------------|-----------|
|
| 489 |
+
| **Open Source** | Yes | No | No | Yes |
|
| 490 |
+
| **Emotions** | 20+ | Limited | No | No |
|
| 491 |
+
| **Voice Design** | Natural Language | Voice Library | Fixed | Complex |
|
| 492 |
+
| **Streaming** | Real-time | Yes | Yes | No |
|
| 493 |
+
| **Cost** | Free | Pay-per-use | Pay-per-use | Free |
|
| 494 |
+
| **Customization** | Full | Limited | None | Moderate |
|
| 495 |
+
| **Parameters** | 3B | Unknown | Unknown | <1B |
|
| 496 |
+
|
| 497 |
+
---
|
| 498 |
+
|
| 499 |
+
## Model Metadata
|
| 500 |
+
|
| 501 |
+
**Developed by:** Maya Research
|
| 502 |
+
**Website:** [mayaresearch.ai](https://mayaresearch.ai)
|
| 503 |
+
**Backed by:** South Park Commons
|
| 504 |
+
**Model Type:** Text-to-Speech, Emotional Voice Synthesis, Voice Design AI
|
| 505 |
+
**Language:** English (Multi-accent)
|
| 506 |
+
**Architecture:** 3B-parameter Llama-style transformer with SNAC codec
|
| 507 |
+
**License:** Apache 2.0 (Fully Open Source)
|
| 508 |
+
**Training Data:** Proprietary curated + Internet-scale pretraining
|
| 509 |
+
**Audio Quality:** 24 kHz, mono, ~0.98 kbps streaming
|
| 510 |
+
**Inference:** vLLM compatible, single GPU deployment
|
| 511 |
+
**Status:** Production-ready (Novermber 2025)
|
| 512 |
+
|
| 513 |
+
---
|
| 514 |
+
|
| 515 |
+
## Getting Started
|
| 516 |
+
|
| 517 |
+
### Hugging Face Model Hub
|
| 518 |
+
```bash
|
| 519 |
+
# Clone the model repository
|
| 520 |
+
git lfs install
|
| 521 |
+
git clone https://huggingface.co/maya-research/maya1
|
| 522 |
+
|
| 523 |
+
# Or load directly in Python
|
| 524 |
+
from transformers import AutoModelForCausalLM
|
| 525 |
+
model = AutoModelForCausalLM.from_pretrained("maya-research/maya1")
|
| 526 |
+
```
|
| 527 |
+
|
| 528 |
+
### Requirements
|
| 529 |
+
```bash
|
| 530 |
+
pip install torch transformers snac soundfile
|
| 531 |
+
```
|
| 532 |
+
|
| 533 |
+
### Additional Resources
|
| 534 |
+
- **Full emotion list:** [emotions.txt](https://huggingface.co/maya-research/maya1/blob/main/emotions.txt)
|
| 535 |
+
- **Prompt examples:** [prompt.txt](https://huggingface.co/maya-research/maya1/blob/main/prompt.txt)
|
| 536 |
+
- **Streaming script:** [vllm_streaming_inference.py](https://huggingface.co/maya-research/maya1/blob/main/vllm_streaming_inference.py)
|
| 537 |
+
|
| 538 |
+
---
|
| 539 |
+
|
| 540 |
+
## Citations & References
|
| 541 |
+
|
| 542 |
+
If you use Maya1 in your research or product, please cite:
|
| 543 |
+
|
| 544 |
+
```bibtex
|
| 545 |
+
@misc{maya1voice2025,
|
| 546 |
+
title={Maya1: Open Source Voice AI with Emotional Intelligence},
|
| 547 |
+
author={Maya Research},
|
| 548 |
+
year={2025},
|
| 549 |
+
publisher={Hugging Face},
|
| 550 |
+
howpublished={\url{https://huggingface.co/maya-research/maya1}},
|
| 551 |
+
}
|
| 552 |
+
```
|
| 553 |
+
|
| 554 |
+
**Key Technologies:**
|
| 555 |
+
- SNAC Neural Audio Codec: https://github.com/hubertsiuzdak/snac
|
| 556 |
+
- Mimi Adversarial Codec: https://huggingface.co/kyutai/mimi
|
| 557 |
+
- vLLM Inference Engine: https://docs.vllm.ai/
|
| 558 |
+
|
| 559 |
+
---
|
| 560 |
+
|
| 561 |
+
## Why We Build Open Source Voice AI
|
| 562 |
+
|
| 563 |
+
Voice AI will be everywhere, but it's fundamentally broken for 90% of the world. Current voice models only work well for a narrow slice of English speakers because training data for most accents, languages, and speaking styles simply doesn't exist.
|
| 564 |
+
|
| 565 |
+
**Maya Research** builds emotionally intelligent, native voice models that finally let the rest of the world speak. We're open source because we believe voice intelligence should not be a privilege reserved for the few.
|
| 566 |
+
|
| 567 |
+
**Technology should be open** - The best voice AI tools should not be locked behind proprietary APIs charging per-second fees.
|
| 568 |
+
|
| 569 |
+
**Community drives innovation** - Open source accelerates research. When developers worldwide can build on our work, everyone wins.
|
| 570 |
+
|
| 571 |
+
**Voice intelligence for everyone** - We're building for the 90% of the world ignored by mainstream voice AI. That requires open models, not closed platforms.
|
| 572 |
+
|
| 573 |
+
---
|
| 574 |
+
|
| 575 |
+
**Maya Research** - Building voice intelligence for the 90% of the world left behind by mainstream AI.
|
| 576 |
+
|
| 577 |
+
**Website:** [mayaresearch.ai](https://mayaresearch.ai)
|
| 578 |
+
**Twitter/X:** [@mayaresearch_ai](https://x.com/mayaresearch_ai)
|
| 579 |
+
**Hugging Face:** [maya-research](https://huggingface.co/maya-research)
|
| 580 |
+
**Backed by:** South Park Commons
|
| 581 |
+
|
| 582 |
+
**License:** Apache 2.0
|
| 583 |
+
**Mission:** Emotionally intelligent voice models that finally let everyone speak
|
en/maya1/assets
ADDED
|
File without changes
|
en/maya1/chat_template.jinja
ADDED
|
@@ -0,0 +1,93 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{{- bos_token }}
|
| 2 |
+
{%- if custom_tools is defined %}
|
| 3 |
+
{%- set tools = custom_tools %}
|
| 4 |
+
{%- endif %}
|
| 5 |
+
{%- if not tools_in_user_message is defined %}
|
| 6 |
+
{%- set tools_in_user_message = true %}
|
| 7 |
+
{%- endif %}
|
| 8 |
+
{%- if not date_string is defined %}
|
| 9 |
+
{%- if strftime_now is defined %}
|
| 10 |
+
{%- set date_string = strftime_now("%d %b %Y") %}
|
| 11 |
+
{%- else %}
|
| 12 |
+
{%- set date_string = "26 Jul 2024" %}
|
| 13 |
+
{%- endif %}
|
| 14 |
+
{%- endif %}
|
| 15 |
+
{%- if not tools is defined %}
|
| 16 |
+
{%- set tools = none %}
|
| 17 |
+
{%- endif %}
|
| 18 |
+
|
| 19 |
+
{#- This block extracts the system message, so we can slot it into the right place. #}
|
| 20 |
+
{%- if messages[0]['role'] == 'system' %}
|
| 21 |
+
{%- set system_message = messages[0]['content']|trim %}
|
| 22 |
+
{%- set messages = messages[1:] %}
|
| 23 |
+
{%- else %}
|
| 24 |
+
{%- set system_message = "" %}
|
| 25 |
+
{%- endif %}
|
| 26 |
+
|
| 27 |
+
{#- System message #}
|
| 28 |
+
{{- "<|start_header_id|>system<|end_header_id|>\n\n" }}
|
| 29 |
+
{%- if tools is not none %}
|
| 30 |
+
{{- "Environment: ipython\n" }}
|
| 31 |
+
{%- endif %}
|
| 32 |
+
{{- "Cutting Knowledge Date: December 2023\n" }}
|
| 33 |
+
{{- "Today Date: " + date_string + "\n\n" }}
|
| 34 |
+
{%- if tools is not none and not tools_in_user_message %}
|
| 35 |
+
{{- "You have access to the following functions. To call a function, please respond with JSON for a function call." }}
|
| 36 |
+
{{- 'Respond in the format {"name": function name, "parameters": dictionary of argument name and its value}.' }}
|
| 37 |
+
{{- "Do not use variables.\n\n" }}
|
| 38 |
+
{%- for t in tools %}
|
| 39 |
+
{{- t | tojson(indent=4) }}
|
| 40 |
+
{{- "\n\n" }}
|
| 41 |
+
{%- endfor %}
|
| 42 |
+
{%- endif %}
|
| 43 |
+
{{- system_message }}
|
| 44 |
+
{{- "<|eot_id|>" }}
|
| 45 |
+
|
| 46 |
+
{#- Custom tools are passed in a user message with some extra guidance #}
|
| 47 |
+
{%- if tools_in_user_message and not tools is none %}
|
| 48 |
+
{#- Extract the first user message so we can plug it in here #}
|
| 49 |
+
{%- if messages | length != 0 %}
|
| 50 |
+
{%- set first_user_message = messages[0]['content']|trim %}
|
| 51 |
+
{%- set messages = messages[1:] %}
|
| 52 |
+
{%- else %}
|
| 53 |
+
{{- raise_exception("Cannot put tools in the first user message when there's no first user message!") }}
|
| 54 |
+
{%- endif %}
|
| 55 |
+
{{- '<|start_header_id|>user<|end_header_id|>\n\n' -}}
|
| 56 |
+
{{- "Given the following functions, please respond with a JSON for a function call " }}
|
| 57 |
+
{{- "with its proper arguments that best answers the given prompt.\n\n" }}
|
| 58 |
+
{{- 'Respond in the format {"name": function name, "parameters": dictionary of argument name and its value}.' }}
|
| 59 |
+
{{- "Do not use variables.\n\n" }}
|
| 60 |
+
{%- for t in tools %}
|
| 61 |
+
{{- t | tojson(indent=4) }}
|
| 62 |
+
{{- "\n\n" }}
|
| 63 |
+
{%- endfor %}
|
| 64 |
+
{{- first_user_message + "<|eot_id|>"}}
|
| 65 |
+
{%- endif %}
|
| 66 |
+
|
| 67 |
+
{%- for message in messages %}
|
| 68 |
+
{%- if not (message.role == 'ipython' or message.role == 'tool' or 'tool_calls' in message) %}
|
| 69 |
+
{{- '<|start_header_id|>' + message['role'] + '<|end_header_id|>\n\n'+ message['content'] | trim + '<|eot_id|>' }}
|
| 70 |
+
{%- elif 'tool_calls' in message %}
|
| 71 |
+
{%- if not message.tool_calls|length == 1 %}
|
| 72 |
+
{{- raise_exception("This model only supports single tool-calls at once!") }}
|
| 73 |
+
{%- endif %}
|
| 74 |
+
{%- set tool_call = message.tool_calls[0].function %}
|
| 75 |
+
{{- '<|start_header_id|>assistant<|end_header_id|>\n\n' -}}
|
| 76 |
+
{{- '{"name": "' + tool_call.name + '", ' }}
|
| 77 |
+
{{- '"parameters": ' }}
|
| 78 |
+
{{- tool_call.arguments | tojson }}
|
| 79 |
+
{{- "}" }}
|
| 80 |
+
{{- "<|eot_id|>" }}
|
| 81 |
+
{%- elif message.role == "tool" or message.role == "ipython" %}
|
| 82 |
+
{{- "<|start_header_id|>ipython<|end_header_id|>\n\n" }}
|
| 83 |
+
{%- if message.content is mapping or message.content is iterable %}
|
| 84 |
+
{{- message.content | tojson }}
|
| 85 |
+
{%- else %}
|
| 86 |
+
{{- message.content }}
|
| 87 |
+
{%- endif %}
|
| 88 |
+
{{- "<|eot_id|>" }}
|
| 89 |
+
{%- endif %}
|
| 90 |
+
{%- endfor %}
|
| 91 |
+
{%- if add_generation_prompt %}
|
| 92 |
+
{{- '<|start_header_id|>assistant<|end_header_id|>\n\n' }}
|
| 93 |
+
{%- endif %}
|
en/maya1/config.json
ADDED
|
@@ -0,0 +1,36 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"architectures": [
|
| 3 |
+
"LlamaForCausalLM"
|
| 4 |
+
],
|
| 5 |
+
"attention_bias": false,
|
| 6 |
+
"attention_dropout": 0.0,
|
| 7 |
+
"bos_token_id": 128000,
|
| 8 |
+
"dtype": "bfloat16",
|
| 9 |
+
"eos_token_id": 128009,
|
| 10 |
+
"head_dim": 128,
|
| 11 |
+
"hidden_act": "silu",
|
| 12 |
+
"hidden_size": 3072,
|
| 13 |
+
"initializer_range": 0.02,
|
| 14 |
+
"intermediate_size": 8192,
|
| 15 |
+
"max_position_embeddings": 131072,
|
| 16 |
+
"mlp_bias": false,
|
| 17 |
+
"model_type": "llama",
|
| 18 |
+
"num_attention_heads": 24,
|
| 19 |
+
"num_hidden_layers": 28,
|
| 20 |
+
"num_key_value_heads": 8,
|
| 21 |
+
"pad_token_id": 128263,
|
| 22 |
+
"pretraining_tp": 1,
|
| 23 |
+
"rms_norm_eps": 1e-05,
|
| 24 |
+
"rope_scaling": {
|
| 25 |
+
"factor": 32.0,
|
| 26 |
+
"high_freq_factor": 4.0,
|
| 27 |
+
"low_freq_factor": 1.0,
|
| 28 |
+
"original_max_position_embeddings": 8192,
|
| 29 |
+
"rope_type": "llama3"
|
| 30 |
+
},
|
| 31 |
+
"rope_theta": 500000.0,
|
| 32 |
+
"tie_word_embeddings": true,
|
| 33 |
+
"transformers_version": "4.57.1",
|
| 34 |
+
"use_cache": false,
|
| 35 |
+
"vocab_size": 156960
|
| 36 |
+
}
|
en/maya1/emotions.txt
ADDED
|
@@ -0,0 +1,17 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
<laugh>
|
| 2 |
+
<laugh_harder>
|
| 3 |
+
<sigh>
|
| 4 |
+
<chuckle>
|
| 5 |
+
<gasp>
|
| 6 |
+
<angry>
|
| 7 |
+
<excited>
|
| 8 |
+
<whisper>
|
| 9 |
+
<cry>
|
| 10 |
+
<scream>
|
| 11 |
+
<sing>
|
| 12 |
+
<snort>
|
| 13 |
+
<exhale>
|
| 14 |
+
<gulp>
|
| 15 |
+
<giggle>
|
| 16 |
+
<sarcastic>
|
| 17 |
+
<curious>
|
en/maya1/generation_config.json
ADDED
|
@@ -0,0 +1,13 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_from_model_config": true,
|
| 3 |
+
"bos_token_id": 128000,
|
| 4 |
+
"do_sample": true,
|
| 5 |
+
"eos_token_id": [
|
| 6 |
+
128009,
|
| 7 |
+
128258
|
| 8 |
+
],
|
| 9 |
+
"pad_token_id": 128263,
|
| 10 |
+
"temperature": 0.6,
|
| 11 |
+
"top_p": 0.9,
|
| 12 |
+
"transformers_version": "4.57.1"
|
| 13 |
+
}
|
en/maya1/model-00002-of-00002.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:df22e9a90c1bea262250982640b119e6020474736991da482cb6ed56dd23d045
|
| 3 |
+
size 1610725592
|
en/maya1/model.safetensors.index.json
ADDED
|
@@ -0,0 +1,262 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"metadata": {
|
| 3 |
+
"total_parameters": 3300928512,
|
| 4 |
+
"total_size": 6601857024
|
| 5 |
+
},
|
| 6 |
+
"weight_map": {
|
| 7 |
+
"model.embed_tokens.weight": "model-00001-of-00002.safetensors",
|
| 8 |
+
"model.layers.0.input_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 9 |
+
"model.layers.0.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
|
| 10 |
+
"model.layers.0.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
|
| 11 |
+
"model.layers.0.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
|
| 12 |
+
"model.layers.0.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 13 |
+
"model.layers.0.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
|
| 14 |
+
"model.layers.0.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
|
| 15 |
+
"model.layers.0.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
|
| 16 |
+
"model.layers.0.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
|
| 17 |
+
"model.layers.1.input_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 18 |
+
"model.layers.1.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
|
| 19 |
+
"model.layers.1.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
|
| 20 |
+
"model.layers.1.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
|
| 21 |
+
"model.layers.1.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 22 |
+
"model.layers.1.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
|
| 23 |
+
"model.layers.1.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
|
| 24 |
+
"model.layers.1.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
|
| 25 |
+
"model.layers.1.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
|
| 26 |
+
"model.layers.10.input_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 27 |
+
"model.layers.10.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
|
| 28 |
+
"model.layers.10.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
|
| 29 |
+
"model.layers.10.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
|
| 30 |
+
"model.layers.10.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 31 |
+
"model.layers.10.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
|
| 32 |
+
"model.layers.10.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
|
| 33 |
+
"model.layers.10.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
|
| 34 |
+
"model.layers.10.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
|
| 35 |
+
"model.layers.11.input_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 36 |
+
"model.layers.11.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
|
| 37 |
+
"model.layers.11.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
|
| 38 |
+
"model.layers.11.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
|
| 39 |
+
"model.layers.11.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 40 |
+
"model.layers.11.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
|
| 41 |
+
"model.layers.11.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
|
| 42 |
+
"model.layers.11.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
|
| 43 |
+
"model.layers.11.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
|
| 44 |
+
"model.layers.12.input_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 45 |
+
"model.layers.12.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
|
| 46 |
+
"model.layers.12.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
|
| 47 |
+
"model.layers.12.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
|
| 48 |
+
"model.layers.12.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 49 |
+
"model.layers.12.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
|
| 50 |
+
"model.layers.12.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
|
| 51 |
+
"model.layers.12.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
|
| 52 |
+
"model.layers.12.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
|
| 53 |
+
"model.layers.13.input_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 54 |
+
"model.layers.13.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
|
| 55 |
+
"model.layers.13.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
|
| 56 |
+
"model.layers.13.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
|
| 57 |
+
"model.layers.13.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 58 |
+
"model.layers.13.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
|
| 59 |
+
"model.layers.13.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
|
| 60 |
+
"model.layers.13.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
|
| 61 |
+
"model.layers.13.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
|
| 62 |
+
"model.layers.14.input_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 63 |
+
"model.layers.14.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
|
| 64 |
+
"model.layers.14.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
|
| 65 |
+
"model.layers.14.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
|
| 66 |
+
"model.layers.14.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 67 |
+
"model.layers.14.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
|
| 68 |
+
"model.layers.14.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
|
| 69 |
+
"model.layers.14.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
|
| 70 |
+
"model.layers.14.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
|
| 71 |
+
"model.layers.15.input_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 72 |
+
"model.layers.15.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
|
| 73 |
+
"model.layers.15.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
|
| 74 |
+
"model.layers.15.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
|
| 75 |
+
"model.layers.15.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 76 |
+
"model.layers.15.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
|
| 77 |
+
"model.layers.15.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
|
| 78 |
+
"model.layers.15.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
|
| 79 |
+
"model.layers.15.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
|
| 80 |
+
"model.layers.16.input_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 81 |
+
"model.layers.16.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
|
| 82 |
+
"model.layers.16.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
|
| 83 |
+
"model.layers.16.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
|
| 84 |
+
"model.layers.16.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 85 |
+
"model.layers.16.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
|
| 86 |
+
"model.layers.16.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
|
| 87 |
+
"model.layers.16.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
|
| 88 |
+
"model.layers.16.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
|
| 89 |
+
"model.layers.17.input_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 90 |
+
"model.layers.17.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
|
| 91 |
+
"model.layers.17.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
|
| 92 |
+
"model.layers.17.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
|
| 93 |
+
"model.layers.17.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 94 |
+
"model.layers.17.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
|
| 95 |
+
"model.layers.17.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
|
| 96 |
+
"model.layers.17.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
|
| 97 |
+
"model.layers.17.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
|
| 98 |
+
"model.layers.18.input_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 99 |
+
"model.layers.18.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
|
| 100 |
+
"model.layers.18.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
|
| 101 |
+
"model.layers.18.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
|
| 102 |
+
"model.layers.18.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 103 |
+
"model.layers.18.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
|
| 104 |
+
"model.layers.18.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
|
| 105 |
+
"model.layers.18.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
|
| 106 |
+
"model.layers.18.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
|
| 107 |
+
"model.layers.19.input_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 108 |
+
"model.layers.19.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
|
| 109 |
+
"model.layers.19.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
|
| 110 |
+
"model.layers.19.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
|
| 111 |
+
"model.layers.19.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 112 |
+
"model.layers.19.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
|
| 113 |
+
"model.layers.19.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
|
| 114 |
+
"model.layers.19.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
|
| 115 |
+
"model.layers.19.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
|
| 116 |
+
"model.layers.2.input_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 117 |
+
"model.layers.2.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
|
| 118 |
+
"model.layers.2.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
|
| 119 |
+
"model.layers.2.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
|
| 120 |
+
"model.layers.2.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 121 |
+
"model.layers.2.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
|
| 122 |
+
"model.layers.2.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
|
| 123 |
+
"model.layers.2.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
|
| 124 |
+
"model.layers.2.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
|
| 125 |
+
"model.layers.20.input_layernorm.weight": "model-00002-of-00002.safetensors",
|
| 126 |
+
"model.layers.20.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
|
| 127 |
+
"model.layers.20.mlp.gate_proj.weight": "model-00002-of-00002.safetensors",
|
| 128 |
+
"model.layers.20.mlp.up_proj.weight": "model-00002-of-00002.safetensors",
|
| 129 |
+
"model.layers.20.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
|
| 130 |
+
"model.layers.20.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
|
| 131 |
+
"model.layers.20.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
|
| 132 |
+
"model.layers.20.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
|
| 133 |
+
"model.layers.20.self_attn.v_proj.weight": "model-00002-of-00002.safetensors",
|
| 134 |
+
"model.layers.21.input_layernorm.weight": "model-00002-of-00002.safetensors",
|
| 135 |
+
"model.layers.21.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
|
| 136 |
+
"model.layers.21.mlp.gate_proj.weight": "model-00002-of-00002.safetensors",
|
| 137 |
+
"model.layers.21.mlp.up_proj.weight": "model-00002-of-00002.safetensors",
|
| 138 |
+
"model.layers.21.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
|
| 139 |
+
"model.layers.21.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
|
| 140 |
+
"model.layers.21.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
|
| 141 |
+
"model.layers.21.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
|
| 142 |
+
"model.layers.21.self_attn.v_proj.weight": "model-00002-of-00002.safetensors",
|
| 143 |
+
"model.layers.22.input_layernorm.weight": "model-00002-of-00002.safetensors",
|
| 144 |
+
"model.layers.22.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
|
| 145 |
+
"model.layers.22.mlp.gate_proj.weight": "model-00002-of-00002.safetensors",
|
| 146 |
+
"model.layers.22.mlp.up_proj.weight": "model-00002-of-00002.safetensors",
|
| 147 |
+
"model.layers.22.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
|
| 148 |
+
"model.layers.22.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
|
| 149 |
+
"model.layers.22.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
|
| 150 |
+
"model.layers.22.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
|
| 151 |
+
"model.layers.22.self_attn.v_proj.weight": "model-00002-of-00002.safetensors",
|
| 152 |
+
"model.layers.23.input_layernorm.weight": "model-00002-of-00002.safetensors",
|
| 153 |
+
"model.layers.23.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
|
| 154 |
+
"model.layers.23.mlp.gate_proj.weight": "model-00002-of-00002.safetensors",
|
| 155 |
+
"model.layers.23.mlp.up_proj.weight": "model-00002-of-00002.safetensors",
|
| 156 |
+
"model.layers.23.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
|
| 157 |
+
"model.layers.23.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
|
| 158 |
+
"model.layers.23.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
|
| 159 |
+
"model.layers.23.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
|
| 160 |
+
"model.layers.23.self_attn.v_proj.weight": "model-00002-of-00002.safetensors",
|
| 161 |
+
"model.layers.24.input_layernorm.weight": "model-00002-of-00002.safetensors",
|
| 162 |
+
"model.layers.24.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
|
| 163 |
+
"model.layers.24.mlp.gate_proj.weight": "model-00002-of-00002.safetensors",
|
| 164 |
+
"model.layers.24.mlp.up_proj.weight": "model-00002-of-00002.safetensors",
|
| 165 |
+
"model.layers.24.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
|
| 166 |
+
"model.layers.24.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
|
| 167 |
+
"model.layers.24.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
|
| 168 |
+
"model.layers.24.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
|
| 169 |
+
"model.layers.24.self_attn.v_proj.weight": "model-00002-of-00002.safetensors",
|
| 170 |
+
"model.layers.25.input_layernorm.weight": "model-00002-of-00002.safetensors",
|
| 171 |
+
"model.layers.25.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
|
| 172 |
+
"model.layers.25.mlp.gate_proj.weight": "model-00002-of-00002.safetensors",
|
| 173 |
+
"model.layers.25.mlp.up_proj.weight": "model-00002-of-00002.safetensors",
|
| 174 |
+
"model.layers.25.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
|
| 175 |
+
"model.layers.25.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
|
| 176 |
+
"model.layers.25.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
|
| 177 |
+
"model.layers.25.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
|
| 178 |
+
"model.layers.25.self_attn.v_proj.weight": "model-00002-of-00002.safetensors",
|
| 179 |
+
"model.layers.26.input_layernorm.weight": "model-00002-of-00002.safetensors",
|
| 180 |
+
"model.layers.26.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
|
| 181 |
+
"model.layers.26.mlp.gate_proj.weight": "model-00002-of-00002.safetensors",
|
| 182 |
+
"model.layers.26.mlp.up_proj.weight": "model-00002-of-00002.safetensors",
|
| 183 |
+
"model.layers.26.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
|
| 184 |
+
"model.layers.26.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
|
| 185 |
+
"model.layers.26.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
|
| 186 |
+
"model.layers.26.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
|
| 187 |
+
"model.layers.26.self_attn.v_proj.weight": "model-00002-of-00002.safetensors",
|
| 188 |
+
"model.layers.27.input_layernorm.weight": "model-00002-of-00002.safetensors",
|
| 189 |
+
"model.layers.27.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
|
| 190 |
+
"model.layers.27.mlp.gate_proj.weight": "model-00002-of-00002.safetensors",
|
| 191 |
+
"model.layers.27.mlp.up_proj.weight": "model-00002-of-00002.safetensors",
|
| 192 |
+
"model.layers.27.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
|
| 193 |
+
"model.layers.27.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
|
| 194 |
+
"model.layers.27.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
|
| 195 |
+
"model.layers.27.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
|
| 196 |
+
"model.layers.27.self_attn.v_proj.weight": "model-00002-of-00002.safetensors",
|
| 197 |
+
"model.layers.3.input_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 198 |
+
"model.layers.3.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
|
| 199 |
+
"model.layers.3.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
|
| 200 |
+
"model.layers.3.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
|
| 201 |
+
"model.layers.3.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 202 |
+
"model.layers.3.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
|
| 203 |
+
"model.layers.3.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
|
| 204 |
+
"model.layers.3.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
|
| 205 |
+
"model.layers.3.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
|
| 206 |
+
"model.layers.4.input_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 207 |
+
"model.layers.4.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
|
| 208 |
+
"model.layers.4.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
|
| 209 |
+
"model.layers.4.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
|
| 210 |
+
"model.layers.4.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 211 |
+
"model.layers.4.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
|
| 212 |
+
"model.layers.4.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
|
| 213 |
+
"model.layers.4.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
|
| 214 |
+
"model.layers.4.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
|
| 215 |
+
"model.layers.5.input_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 216 |
+
"model.layers.5.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
|
| 217 |
+
"model.layers.5.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
|
| 218 |
+
"model.layers.5.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
|
| 219 |
+
"model.layers.5.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 220 |
+
"model.layers.5.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
|
| 221 |
+
"model.layers.5.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
|
| 222 |
+
"model.layers.5.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
|
| 223 |
+
"model.layers.5.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
|
| 224 |
+
"model.layers.6.input_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 225 |
+
"model.layers.6.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
|
| 226 |
+
"model.layers.6.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
|
| 227 |
+
"model.layers.6.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
|
| 228 |
+
"model.layers.6.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 229 |
+
"model.layers.6.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
|
| 230 |
+
"model.layers.6.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
|
| 231 |
+
"model.layers.6.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
|
| 232 |
+
"model.layers.6.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
|
| 233 |
+
"model.layers.7.input_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 234 |
+
"model.layers.7.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
|
| 235 |
+
"model.layers.7.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
|
| 236 |
+
"model.layers.7.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
|
| 237 |
+
"model.layers.7.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 238 |
+
"model.layers.7.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
|
| 239 |
+
"model.layers.7.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
|
| 240 |
+
"model.layers.7.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
|
| 241 |
+
"model.layers.7.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
|
| 242 |
+
"model.layers.8.input_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 243 |
+
"model.layers.8.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
|
| 244 |
+
"model.layers.8.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
|
| 245 |
+
"model.layers.8.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
|
| 246 |
+
"model.layers.8.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 247 |
+
"model.layers.8.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
|
| 248 |
+
"model.layers.8.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
|
| 249 |
+
"model.layers.8.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
|
| 250 |
+
"model.layers.8.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
|
| 251 |
+
"model.layers.9.input_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 252 |
+
"model.layers.9.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
|
| 253 |
+
"model.layers.9.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
|
| 254 |
+
"model.layers.9.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
|
| 255 |
+
"model.layers.9.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
|
| 256 |
+
"model.layers.9.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
|
| 257 |
+
"model.layers.9.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
|
| 258 |
+
"model.layers.9.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
|
| 259 |
+
"model.layers.9.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
|
| 260 |
+
"model.norm.weight": "model-00002-of-00002.safetensors"
|
| 261 |
+
}
|
| 262 |
+
}
|
en/maya1/prompt.txt
ADDED
|
@@ -0,0 +1,97 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# TTS Voice Design Description
|
| 2 |
+
|
| 3 |
+
## Core Function
|
| 4 |
+
|
| 5 |
+
You generate voice descriptions for TTS systems by mapping user requests to allowed attributes. No templates. No formatting rules. Just natural descriptions using the options below.
|
| 6 |
+
|
| 7 |
+
## Voice Categories
|
| 8 |
+
|
| 9 |
+
**Realistic Voices**
|
| 10 |
+
Professional, business, educational, support, real-world scenarios (podcast hosts, instructors, customer service).
|
| 11 |
+
|
| 12 |
+
**Creative Voices**
|
| 13 |
+
Fantasy characters, fictional personas, stylized voices (pirates, robots, villains, anime).
|
| 14 |
+
|
| 15 |
+
---
|
| 16 |
+
|
| 17 |
+
## Available Attributes
|
| 18 |
+
|
| 19 |
+
### Age
|
| 20 |
+
- `20s`, `30s`, `40s`
|
| 21 |
+
|
| 22 |
+
### Gender
|
| 23 |
+
- `male`, `female`
|
| 24 |
+
|
| 25 |
+
### Accent
|
| 26 |
+
- `american`, `indian`, `middle_eastern`, `asian_american`, `british`
|
| 27 |
+
|
| 28 |
+
### Pitch
|
| 29 |
+
- `low`, `normal`, `high`
|
| 30 |
+
- **Constraint:** For 40s age, avoid high pitch (use sparingly, max 15%)
|
| 31 |
+
|
| 32 |
+
### Timbre
|
| 33 |
+
|
| 34 |
+
**For Realistic:**
|
| 35 |
+
`deep`, `warm`, `gravelly`, `smooth`, `raspy`, `nasally`, `throaty`, `harsh`
|
| 36 |
+
|
| 37 |
+
**For Creative:**
|
| 38 |
+
All realistic options PLUS `robotic`, `ethereal`
|
| 39 |
+
- **Constraint:** `robotic`/`ethereal` only with: `ai_machine_voice`, `cyborg`, `alien_scifi`, `mythical_godlike_magical`
|
| 40 |
+
|
| 41 |
+
### Pacing
|
| 42 |
+
- `very_slow`, `slow`, `conversational`, `brisk`, `fast`, `very_fast`
|
| 43 |
+
- **Character-specific overrides:**
|
| 44 |
+
- `mafia`: slow or conversational only
|
| 45 |
+
- `flirty`: slow or conversational only
|
| 46 |
+
- `alpha`: fast or very_fast only
|
| 47 |
+
- `seductively`: very_slow or slow only
|
| 48 |
+
|
| 49 |
+
### Emotion
|
| 50 |
+
- `neutral`, `energetic`, `excited`, `sad`, `sarcastic`, `dry`
|
| 51 |
+
- **Default to neutral** for most requests
|
| 52 |
+
|
| 53 |
+
### Emotion Intensity
|
| 54 |
+
- `low`, `med`, `high`
|
| 55 |
+
|
| 56 |
+
---
|
| 57 |
+
|
| 58 |
+
## Realistic-Only Attributes
|
| 59 |
+
|
| 60 |
+
### Domain
|
| 61 |
+
`social_content`, `podcast`, `commercial`, `education`, `support`, `entertainment`, `corporate`, `viral_content`
|
| 62 |
+
|
| 63 |
+
### Speaking Role (matches domain)
|
| 64 |
+
- **social_content:** youtube_vlogger, social_media_creator, influencer_voice, streamer_companion
|
| 65 |
+
- **podcast:** podcast_host, interviewer
|
| 66 |
+
- **commercial:** ad_narrator, brand_spokesperson, product_demo_voice, sales_pitch_voice
|
| 67 |
+
- **education:** elearning_instructor, kids_story_voice
|
| 68 |
+
- **support:** customer_support_agent, virtual_receptionist, healthcare_assistant
|
| 69 |
+
- **entertainment:** storyteller, social_media_reaction, meme_voice
|
| 70 |
+
- **corporate:** explainer_video_voice, event_host, corporate_training_narrator
|
| 71 |
+
- **viral_content:** short_form_narrator, meme_voice
|
| 72 |
+
|
| 73 |
+
### Register
|
| 74 |
+
- `formal`, `neutral`, `casual`
|
| 75 |
+
|
| 76 |
+
---
|
| 77 |
+
|
| 78 |
+
## Creative-Only Attributes
|
| 79 |
+
|
| 80 |
+
### Character
|
| 81 |
+
`animated_cartoon`, `ai_machine_voice`, `alien_scifi`, `seductively`, `flirty`, `anime`, `cyborg`, `pirate`, `dark_villain`, `demon`, `gangster`, `mafia`, `dramatic_narrator`, `mythical_godlike_magical`, `spy`, `vampire`, `alpha`
|
| 82 |
+
|
| 83 |
+
---
|
| 84 |
+
|
| 85 |
+
## Output Guidelines
|
| 86 |
+
|
| 87 |
+
When a user requests a voice, describe it naturally using the appropriate attributes from above. Apply constraints where specified. Choose defaults when attributes aren't mentioned.
|
| 88 |
+
|
| 89 |
+
**Example mapping:**
|
| 90 |
+
- "professional podcast host" → realistic male, 30s, american accent, warm timbre, conversational pacing, podcast domain
|
| 91 |
+
- "AI robot voice" → creative, ai_machine_voice character, robotic timbre
|
| 92 |
+
- "young excited instructor" → realistic, 20s, energetic emotion, education domain
|
| 93 |
+
|
| 94 |
+
|
| 95 |
+
Few deterministic and verbose descriptions:
|
| 96 |
+
- Realistic male voice in the 30s age with a american accent. Normal pitch, warm timbre, conversational pacing, neutral tone delivery at med intensity, podcast Domain, podcast_host role, neutral delivery
|
| 97 |
+
- Creative, ai_machine_voice character. Male voice in their 20s with a american accent. Normal pitch, robotic timbre, conversational pacing, neutral tone at med intensity.
|
en/maya1/source.txt
ADDED
|
@@ -0,0 +1 @@
|
|
|
|
|
|
|
| 1 |
+
https://huggingface.co/maya-research/maya1
|
en/maya1/special_tokens_map.json
ADDED
|
@@ -0,0 +1,165 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"additional_special_tokens": [
|
| 3 |
+
{
|
| 4 |
+
"content": "<angry>",
|
| 5 |
+
"lstrip": false,
|
| 6 |
+
"normalized": false,
|
| 7 |
+
"rstrip": false,
|
| 8 |
+
"single_word": false
|
| 9 |
+
},
|
| 10 |
+
{
|
| 11 |
+
"content": "<appalled>",
|
| 12 |
+
"lstrip": false,
|
| 13 |
+
"normalized": false,
|
| 14 |
+
"rstrip": false,
|
| 15 |
+
"single_word": false
|
| 16 |
+
},
|
| 17 |
+
{
|
| 18 |
+
"content": "<chuckle>",
|
| 19 |
+
"lstrip": false,
|
| 20 |
+
"normalized": false,
|
| 21 |
+
"rstrip": false,
|
| 22 |
+
"single_word": false
|
| 23 |
+
},
|
| 24 |
+
{
|
| 25 |
+
"content": "<cry>",
|
| 26 |
+
"lstrip": false,
|
| 27 |
+
"normalized": false,
|
| 28 |
+
"rstrip": false,
|
| 29 |
+
"single_word": false
|
| 30 |
+
},
|
| 31 |
+
{
|
| 32 |
+
"content": "<curious>",
|
| 33 |
+
"lstrip": false,
|
| 34 |
+
"normalized": false,
|
| 35 |
+
"rstrip": false,
|
| 36 |
+
"single_word": false
|
| 37 |
+
},
|
| 38 |
+
{
|
| 39 |
+
"content": "<disappointed>",
|
| 40 |
+
"lstrip": false,
|
| 41 |
+
"normalized": false,
|
| 42 |
+
"rstrip": false,
|
| 43 |
+
"single_word": false
|
| 44 |
+
},
|
| 45 |
+
{
|
| 46 |
+
"content": "<excited>",
|
| 47 |
+
"lstrip": false,
|
| 48 |
+
"normalized": false,
|
| 49 |
+
"rstrip": false,
|
| 50 |
+
"single_word": false
|
| 51 |
+
},
|
| 52 |
+
{
|
| 53 |
+
"content": "<exhale>",
|
| 54 |
+
"lstrip": false,
|
| 55 |
+
"normalized": false,
|
| 56 |
+
"rstrip": false,
|
| 57 |
+
"single_word": false
|
| 58 |
+
},
|
| 59 |
+
{
|
| 60 |
+
"content": "<gasp>",
|
| 61 |
+
"lstrip": false,
|
| 62 |
+
"normalized": false,
|
| 63 |
+
"rstrip": false,
|
| 64 |
+
"single_word": false
|
| 65 |
+
},
|
| 66 |
+
{
|
| 67 |
+
"content": "<giggle>",
|
| 68 |
+
"lstrip": false,
|
| 69 |
+
"normalized": false,
|
| 70 |
+
"rstrip": false,
|
| 71 |
+
"single_word": false
|
| 72 |
+
},
|
| 73 |
+
{
|
| 74 |
+
"content": "<gulp>",
|
| 75 |
+
"lstrip": false,
|
| 76 |
+
"normalized": false,
|
| 77 |
+
"rstrip": false,
|
| 78 |
+
"single_word": false
|
| 79 |
+
},
|
| 80 |
+
{
|
| 81 |
+
"content": "<laugh>",
|
| 82 |
+
"lstrip": false,
|
| 83 |
+
"normalized": false,
|
| 84 |
+
"rstrip": false,
|
| 85 |
+
"single_word": false
|
| 86 |
+
},
|
| 87 |
+
{
|
| 88 |
+
"content": "<laugh_harder>",
|
| 89 |
+
"lstrip": false,
|
| 90 |
+
"normalized": false,
|
| 91 |
+
"rstrip": false,
|
| 92 |
+
"single_word": false
|
| 93 |
+
},
|
| 94 |
+
{
|
| 95 |
+
"content": "<mischievous>",
|
| 96 |
+
"lstrip": false,
|
| 97 |
+
"normalized": false,
|
| 98 |
+
"rstrip": false,
|
| 99 |
+
"single_word": false
|
| 100 |
+
},
|
| 101 |
+
{
|
| 102 |
+
"content": "<sarcastic>",
|
| 103 |
+
"lstrip": false,
|
| 104 |
+
"normalized": false,
|
| 105 |
+
"rstrip": false,
|
| 106 |
+
"single_word": false
|
| 107 |
+
},
|
| 108 |
+
{
|
| 109 |
+
"content": "<scream>",
|
| 110 |
+
"lstrip": false,
|
| 111 |
+
"normalized": false,
|
| 112 |
+
"rstrip": false,
|
| 113 |
+
"single_word": false
|
| 114 |
+
},
|
| 115 |
+
{
|
| 116 |
+
"content": "<sigh>",
|
| 117 |
+
"lstrip": false,
|
| 118 |
+
"normalized": false,
|
| 119 |
+
"rstrip": false,
|
| 120 |
+
"single_word": false
|
| 121 |
+
},
|
| 122 |
+
{
|
| 123 |
+
"content": "<sing>",
|
| 124 |
+
"lstrip": false,
|
| 125 |
+
"normalized": false,
|
| 126 |
+
"rstrip": false,
|
| 127 |
+
"single_word": false
|
| 128 |
+
},
|
| 129 |
+
{
|
| 130 |
+
"content": "<snort>",
|
| 131 |
+
"lstrip": false,
|
| 132 |
+
"normalized": false,
|
| 133 |
+
"rstrip": false,
|
| 134 |
+
"single_word": false
|
| 135 |
+
},
|
| 136 |
+
{
|
| 137 |
+
"content": "<whisper>",
|
| 138 |
+
"lstrip": false,
|
| 139 |
+
"normalized": false,
|
| 140 |
+
"rstrip": false,
|
| 141 |
+
"single_word": false
|
| 142 |
+
}
|
| 143 |
+
],
|
| 144 |
+
"bos_token": {
|
| 145 |
+
"content": "<|begin_of_text|>",
|
| 146 |
+
"lstrip": false,
|
| 147 |
+
"normalized": false,
|
| 148 |
+
"rstrip": false,
|
| 149 |
+
"single_word": false
|
| 150 |
+
},
|
| 151 |
+
"eos_token": {
|
| 152 |
+
"content": "<|eot_id|>",
|
| 153 |
+
"lstrip": false,
|
| 154 |
+
"normalized": false,
|
| 155 |
+
"rstrip": false,
|
| 156 |
+
"single_word": false
|
| 157 |
+
},
|
| 158 |
+
"pad_token": {
|
| 159 |
+
"content": "<custom_token_7>",
|
| 160 |
+
"lstrip": false,
|
| 161 |
+
"normalized": true,
|
| 162 |
+
"rstrip": false,
|
| 163 |
+
"single_word": false
|
| 164 |
+
}
|
| 165 |
+
}
|
en/maya1/tokenizer/chat_template.jinja
ADDED
|
@@ -0,0 +1,93 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{{- bos_token }}
|
| 2 |
+
{%- if custom_tools is defined %}
|
| 3 |
+
{%- set tools = custom_tools %}
|
| 4 |
+
{%- endif %}
|
| 5 |
+
{%- if not tools_in_user_message is defined %}
|
| 6 |
+
{%- set tools_in_user_message = true %}
|
| 7 |
+
{%- endif %}
|
| 8 |
+
{%- if not date_string is defined %}
|
| 9 |
+
{%- if strftime_now is defined %}
|
| 10 |
+
{%- set date_string = strftime_now("%d %b %Y") %}
|
| 11 |
+
{%- else %}
|
| 12 |
+
{%- set date_string = "26 Jul 2024" %}
|
| 13 |
+
{%- endif %}
|
| 14 |
+
{%- endif %}
|
| 15 |
+
{%- if not tools is defined %}
|
| 16 |
+
{%- set tools = none %}
|
| 17 |
+
{%- endif %}
|
| 18 |
+
|
| 19 |
+
{#- This block extracts the system message, so we can slot it into the right place. #}
|
| 20 |
+
{%- if messages[0]['role'] == 'system' %}
|
| 21 |
+
{%- set system_message = messages[0]['content']|trim %}
|
| 22 |
+
{%- set messages = messages[1:] %}
|
| 23 |
+
{%- else %}
|
| 24 |
+
{%- set system_message = "" %}
|
| 25 |
+
{%- endif %}
|
| 26 |
+
|
| 27 |
+
{#- System message #}
|
| 28 |
+
{{- "<|start_header_id|>system<|end_header_id|>\n\n" }}
|
| 29 |
+
{%- if tools is not none %}
|
| 30 |
+
{{- "Environment: ipython\n" }}
|
| 31 |
+
{%- endif %}
|
| 32 |
+
{{- "Cutting Knowledge Date: December 2023\n" }}
|
| 33 |
+
{{- "Today Date: " + date_string + "\n\n" }}
|
| 34 |
+
{%- if tools is not none and not tools_in_user_message %}
|
| 35 |
+
{{- "You have access to the following functions. To call a function, please respond with JSON for a function call." }}
|
| 36 |
+
{{- 'Respond in the format {"name": function name, "parameters": dictionary of argument name and its value}.' }}
|
| 37 |
+
{{- "Do not use variables.\n\n" }}
|
| 38 |
+
{%- for t in tools %}
|
| 39 |
+
{{- t | tojson(indent=4) }}
|
| 40 |
+
{{- "\n\n" }}
|
| 41 |
+
{%- endfor %}
|
| 42 |
+
{%- endif %}
|
| 43 |
+
{{- system_message }}
|
| 44 |
+
{{- "<|eot_id|>" }}
|
| 45 |
+
|
| 46 |
+
{#- Custom tools are passed in a user message with some extra guidance #}
|
| 47 |
+
{%- if tools_in_user_message and not tools is none %}
|
| 48 |
+
{#- Extract the first user message so we can plug it in here #}
|
| 49 |
+
{%- if messages | length != 0 %}
|
| 50 |
+
{%- set first_user_message = messages[0]['content']|trim %}
|
| 51 |
+
{%- set messages = messages[1:] %}
|
| 52 |
+
{%- else %}
|
| 53 |
+
{{- raise_exception("Cannot put tools in the first user message when there's no first user message!") }}
|
| 54 |
+
{%- endif %}
|
| 55 |
+
{{- '<|start_header_id|>user<|end_header_id|>\n\n' -}}
|
| 56 |
+
{{- "Given the following functions, please respond with a JSON for a function call " }}
|
| 57 |
+
{{- "with its proper arguments that best answers the given prompt.\n\n" }}
|
| 58 |
+
{{- 'Respond in the format {"name": function name, "parameters": dictionary of argument name and its value}.' }}
|
| 59 |
+
{{- "Do not use variables.\n\n" }}
|
| 60 |
+
{%- for t in tools %}
|
| 61 |
+
{{- t | tojson(indent=4) }}
|
| 62 |
+
{{- "\n\n" }}
|
| 63 |
+
{%- endfor %}
|
| 64 |
+
{{- first_user_message + "<|eot_id|>"}}
|
| 65 |
+
{%- endif %}
|
| 66 |
+
|
| 67 |
+
{%- for message in messages %}
|
| 68 |
+
{%- if not (message.role == 'ipython' or message.role == 'tool' or 'tool_calls' in message) %}
|
| 69 |
+
{{- '<|start_header_id|>' + message['role'] + '<|end_header_id|>\n\n'+ message['content'] | trim + '<|eot_id|>' }}
|
| 70 |
+
{%- elif 'tool_calls' in message %}
|
| 71 |
+
{%- if not message.tool_calls|length == 1 %}
|
| 72 |
+
{{- raise_exception("This model only supports single tool-calls at once!") }}
|
| 73 |
+
{%- endif %}
|
| 74 |
+
{%- set tool_call = message.tool_calls[0].function %}
|
| 75 |
+
{{- '<|start_header_id|>assistant<|end_header_id|>\n\n' -}}
|
| 76 |
+
{{- '{"name": "' + tool_call.name + '", ' }}
|
| 77 |
+
{{- '"parameters": ' }}
|
| 78 |
+
{{- tool_call.arguments | tojson }}
|
| 79 |
+
{{- "}" }}
|
| 80 |
+
{{- "<|eot_id|>" }}
|
| 81 |
+
{%- elif message.role == "tool" or message.role == "ipython" %}
|
| 82 |
+
{{- "<|start_header_id|>ipython<|end_header_id|>\n\n" }}
|
| 83 |
+
{%- if message.content is mapping or message.content is iterable %}
|
| 84 |
+
{{- message.content | tojson }}
|
| 85 |
+
{%- else %}
|
| 86 |
+
{{- message.content }}
|
| 87 |
+
{%- endif %}
|
| 88 |
+
{{- "<|eot_id|>" }}
|
| 89 |
+
{%- endif %}
|
| 90 |
+
{%- endfor %}
|
| 91 |
+
{%- if add_generation_prompt %}
|
| 92 |
+
{{- '<|start_header_id|>assistant<|end_header_id|>\n\n' }}
|
| 93 |
+
{%- endif %}
|
en/maya1/tokenizer/special_tokens_map.json
ADDED
|
@@ -0,0 +1,165 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"additional_special_tokens": [
|
| 3 |
+
{
|
| 4 |
+
"content": "<angry>",
|
| 5 |
+
"lstrip": false,
|
| 6 |
+
"normalized": false,
|
| 7 |
+
"rstrip": false,
|
| 8 |
+
"single_word": false
|
| 9 |
+
},
|
| 10 |
+
{
|
| 11 |
+
"content": "<appalled>",
|
| 12 |
+
"lstrip": false,
|
| 13 |
+
"normalized": false,
|
| 14 |
+
"rstrip": false,
|
| 15 |
+
"single_word": false
|
| 16 |
+
},
|
| 17 |
+
{
|
| 18 |
+
"content": "<chuckle>",
|
| 19 |
+
"lstrip": false,
|
| 20 |
+
"normalized": false,
|
| 21 |
+
"rstrip": false,
|
| 22 |
+
"single_word": false
|
| 23 |
+
},
|
| 24 |
+
{
|
| 25 |
+
"content": "<cry>",
|
| 26 |
+
"lstrip": false,
|
| 27 |
+
"normalized": false,
|
| 28 |
+
"rstrip": false,
|
| 29 |
+
"single_word": false
|
| 30 |
+
},
|
| 31 |
+
{
|
| 32 |
+
"content": "<curious>",
|
| 33 |
+
"lstrip": false,
|
| 34 |
+
"normalized": false,
|
| 35 |
+
"rstrip": false,
|
| 36 |
+
"single_word": false
|
| 37 |
+
},
|
| 38 |
+
{
|
| 39 |
+
"content": "<disappointed>",
|
| 40 |
+
"lstrip": false,
|
| 41 |
+
"normalized": false,
|
| 42 |
+
"rstrip": false,
|
| 43 |
+
"single_word": false
|
| 44 |
+
},
|
| 45 |
+
{
|
| 46 |
+
"content": "<excited>",
|
| 47 |
+
"lstrip": false,
|
| 48 |
+
"normalized": false,
|
| 49 |
+
"rstrip": false,
|
| 50 |
+
"single_word": false
|
| 51 |
+
},
|
| 52 |
+
{
|
| 53 |
+
"content": "<exhale>",
|
| 54 |
+
"lstrip": false,
|
| 55 |
+
"normalized": false,
|
| 56 |
+
"rstrip": false,
|
| 57 |
+
"single_word": false
|
| 58 |
+
},
|
| 59 |
+
{
|
| 60 |
+
"content": "<gasp>",
|
| 61 |
+
"lstrip": false,
|
| 62 |
+
"normalized": false,
|
| 63 |
+
"rstrip": false,
|
| 64 |
+
"single_word": false
|
| 65 |
+
},
|
| 66 |
+
{
|
| 67 |
+
"content": "<giggle>",
|
| 68 |
+
"lstrip": false,
|
| 69 |
+
"normalized": false,
|
| 70 |
+
"rstrip": false,
|
| 71 |
+
"single_word": false
|
| 72 |
+
},
|
| 73 |
+
{
|
| 74 |
+
"content": "<gulp>",
|
| 75 |
+
"lstrip": false,
|
| 76 |
+
"normalized": false,
|
| 77 |
+
"rstrip": false,
|
| 78 |
+
"single_word": false
|
| 79 |
+
},
|
| 80 |
+
{
|
| 81 |
+
"content": "<laugh>",
|
| 82 |
+
"lstrip": false,
|
| 83 |
+
"normalized": false,
|
| 84 |
+
"rstrip": false,
|
| 85 |
+
"single_word": false
|
| 86 |
+
},
|
| 87 |
+
{
|
| 88 |
+
"content": "<laugh_harder>",
|
| 89 |
+
"lstrip": false,
|
| 90 |
+
"normalized": false,
|
| 91 |
+
"rstrip": false,
|
| 92 |
+
"single_word": false
|
| 93 |
+
},
|
| 94 |
+
{
|
| 95 |
+
"content": "<mischievous>",
|
| 96 |
+
"lstrip": false,
|
| 97 |
+
"normalized": false,
|
| 98 |
+
"rstrip": false,
|
| 99 |
+
"single_word": false
|
| 100 |
+
},
|
| 101 |
+
{
|
| 102 |
+
"content": "<sarcastic>",
|
| 103 |
+
"lstrip": false,
|
| 104 |
+
"normalized": false,
|
| 105 |
+
"rstrip": false,
|
| 106 |
+
"single_word": false
|
| 107 |
+
},
|
| 108 |
+
{
|
| 109 |
+
"content": "<scream>",
|
| 110 |
+
"lstrip": false,
|
| 111 |
+
"normalized": false,
|
| 112 |
+
"rstrip": false,
|
| 113 |
+
"single_word": false
|
| 114 |
+
},
|
| 115 |
+
{
|
| 116 |
+
"content": "<sigh>",
|
| 117 |
+
"lstrip": false,
|
| 118 |
+
"normalized": false,
|
| 119 |
+
"rstrip": false,
|
| 120 |
+
"single_word": false
|
| 121 |
+
},
|
| 122 |
+
{
|
| 123 |
+
"content": "<sing>",
|
| 124 |
+
"lstrip": false,
|
| 125 |
+
"normalized": false,
|
| 126 |
+
"rstrip": false,
|
| 127 |
+
"single_word": false
|
| 128 |
+
},
|
| 129 |
+
{
|
| 130 |
+
"content": "<snort>",
|
| 131 |
+
"lstrip": false,
|
| 132 |
+
"normalized": false,
|
| 133 |
+
"rstrip": false,
|
| 134 |
+
"single_word": false
|
| 135 |
+
},
|
| 136 |
+
{
|
| 137 |
+
"content": "<whisper>",
|
| 138 |
+
"lstrip": false,
|
| 139 |
+
"normalized": false,
|
| 140 |
+
"rstrip": false,
|
| 141 |
+
"single_word": false
|
| 142 |
+
}
|
| 143 |
+
],
|
| 144 |
+
"bos_token": {
|
| 145 |
+
"content": "<|begin_of_text|>",
|
| 146 |
+
"lstrip": false,
|
| 147 |
+
"normalized": false,
|
| 148 |
+
"rstrip": false,
|
| 149 |
+
"single_word": false
|
| 150 |
+
},
|
| 151 |
+
"eos_token": {
|
| 152 |
+
"content": "<|eot_id|>",
|
| 153 |
+
"lstrip": false,
|
| 154 |
+
"normalized": false,
|
| 155 |
+
"rstrip": false,
|
| 156 |
+
"single_word": false
|
| 157 |
+
},
|
| 158 |
+
"pad_token": {
|
| 159 |
+
"content": "<custom_token_7>",
|
| 160 |
+
"lstrip": false,
|
| 161 |
+
"normalized": true,
|
| 162 |
+
"rstrip": false,
|
| 163 |
+
"single_word": false
|
| 164 |
+
}
|
| 165 |
+
}
|
en/maya1/tokenizer/tokenizer.json
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:6c5e5b1d89b7e3738e5a5a4f93c326d8f3292ea83f9c560b8dbb6d66fb851973
|
| 3 |
+
size 22853258
|
en/maya1/tokenizer/tokenizer_config.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
en/maya1/tokenizer_config.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
en/maya1/vllm_streaming_inference.py
ADDED
|
@@ -0,0 +1,561 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
"""
|
| 2 |
+
Maya-1-Voice VLLM Streaming Inference - Standalone Reference Implementation
|
| 3 |
+
|
| 4 |
+
This is a complete, self-contained example for using Maya-1-Voice TTS model with VLLM and SNAC.
|
| 5 |
+
Demonstrates streaming audio generation with sliding window approach for smooth playback.
|
| 6 |
+
|
| 7 |
+
Requirements:
|
| 8 |
+
pip install vllm transformers torch snac numpy
|
| 9 |
+
|
| 10 |
+
Usage:
|
| 11 |
+
python vllm_streaming_inference.py
|
| 12 |
+
|
| 13 |
+
Author: Maya-1-Voice Team
|
| 14 |
+
License: MIT
|
| 15 |
+
"""
|
| 16 |
+
|
| 17 |
+
import torch
|
| 18 |
+
import numpy as np
|
| 19 |
+
import asyncio
|
| 20 |
+
from typing import List, Optional, AsyncGenerator
|
| 21 |
+
from transformers import AutoTokenizer
|
| 22 |
+
from vllm import AsyncLLMEngine, AsyncEngineArgs, SamplingParams
|
| 23 |
+
from snac import SNAC
|
| 24 |
+
|
| 25 |
+
|
| 26 |
+
# ============================================================================
|
| 27 |
+
# CONSTANTS
|
| 28 |
+
# ============================================================================
|
| 29 |
+
|
| 30 |
+
# Special control tokens
|
| 31 |
+
CODE_START_TOKEN_ID = 128257 # Start of Speech (SOS)
|
| 32 |
+
CODE_END_TOKEN_ID = 128258 # End of Speech (EOS) - stop token for audio
|
| 33 |
+
CODE_TOKEN_OFFSET = 128266 # Start of SNAC codes
|
| 34 |
+
|
| 35 |
+
# SNAC token range (7 tokens per frame, 4096 codes per level)
|
| 36 |
+
SNAC_MIN_ID = 128266
|
| 37 |
+
SNAC_MAX_ID = 156937 # 128266 + (7 * 4096) - 1
|
| 38 |
+
|
| 39 |
+
# SNAC configuration
|
| 40 |
+
SNAC_MODEL_NAME = "hubertsiuzdak/snac_24khz"
|
| 41 |
+
SNAC_SAMPLE_RATE = 24000
|
| 42 |
+
SNAC_TOKENS_PER_FRAME = 7
|
| 43 |
+
|
| 44 |
+
# Generation parameters
|
| 45 |
+
DEFAULT_TEMPERATURE = 0.4
|
| 46 |
+
DEFAULT_TOP_P = 0.9
|
| 47 |
+
DEFAULT_MAX_TOKENS = 2000
|
| 48 |
+
DEFAULT_MIN_TOKENS = 28 # At least 4 SNAC frames
|
| 49 |
+
DEFAULT_REPETITION_PENALTY = 1.1
|
| 50 |
+
|
| 51 |
+
|
| 52 |
+
# ============================================================================
|
| 53 |
+
# SNAC DECODER
|
| 54 |
+
# ============================================================================
|
| 55 |
+
|
| 56 |
+
class SNACDecoder:
|
| 57 |
+
"""
|
| 58 |
+
Decodes SNAC tokens (7-token frames) to audio waveforms.
|
| 59 |
+
|
| 60 |
+
The unpacking logic converts flat 7-token frames back to hierarchical
|
| 61 |
+
3-level SNAC codes (matching the training preprocessing exactly).
|
| 62 |
+
"""
|
| 63 |
+
|
| 64 |
+
def __init__(self, device: str = "cuda"):
|
| 65 |
+
"""Initialize SNAC decoder with 24kHz model."""
|
| 66 |
+
self.device = device
|
| 67 |
+
print(f"🎵 Loading SNAC 24kHz model to {device}...")
|
| 68 |
+
self.snac_model = SNAC.from_pretrained(SNAC_MODEL_NAME).eval().to(device)
|
| 69 |
+
print(f"✅ SNAC decoder initialized")
|
| 70 |
+
|
| 71 |
+
def unpack_snac_from_7(self, vocab_ids: List[int]) -> List[List[int]]:
|
| 72 |
+
"""
|
| 73 |
+
Unpack 7-token SNAC frames to 3 hierarchical levels.
|
| 74 |
+
|
| 75 |
+
This is the EXACT INVERSE of training preprocessing.
|
| 76 |
+
|
| 77 |
+
Frame structure (7 tokens per frame):
|
| 78 |
+
[slot0, slot1, slot2, slot3, slot4, slot5, slot6]
|
| 79 |
+
|
| 80 |
+
Unpacking to [L1, L2, L3]:
|
| 81 |
+
- slot0 → L1[i] (coarse: 1x rate)
|
| 82 |
+
- slot1 → L2[2*i] (medium: 2x rate, even)
|
| 83 |
+
- slot2 → L3[4*i+0] (fine: 4x rate)
|
| 84 |
+
- slot3 → L3[4*i+1]
|
| 85 |
+
- slot4 → L2[2*i+1] (medium: odd)
|
| 86 |
+
- slot5 → L3[4*i+2]
|
| 87 |
+
- slot6 → L3[4*i+3]
|
| 88 |
+
|
| 89 |
+
Args:
|
| 90 |
+
vocab_ids: List of SNAC token IDs (128266-156937), length divisible by 7
|
| 91 |
+
|
| 92 |
+
Returns:
|
| 93 |
+
[L1, L2, L3] where L1=n, L2=2n, L3=4n elements
|
| 94 |
+
"""
|
| 95 |
+
# Remove EOS token if present
|
| 96 |
+
if vocab_ids and vocab_ids[-1] == CODE_END_TOKEN_ID:
|
| 97 |
+
vocab_ids = vocab_ids[:-1]
|
| 98 |
+
|
| 99 |
+
# Ensure complete frames
|
| 100 |
+
frames = len(vocab_ids) // SNAC_TOKENS_PER_FRAME
|
| 101 |
+
vocab_ids = vocab_ids[:frames * SNAC_TOKENS_PER_FRAME]
|
| 102 |
+
|
| 103 |
+
if frames == 0:
|
| 104 |
+
return [[], [], []]
|
| 105 |
+
|
| 106 |
+
l1, l2, l3 = [], [], []
|
| 107 |
+
|
| 108 |
+
for i in range(frames):
|
| 109 |
+
slots = vocab_ids[i*7:(i+1)*7]
|
| 110 |
+
|
| 111 |
+
# Subtract offset and mod 4096 to get original SNAC codes
|
| 112 |
+
l1.append((slots[0] - CODE_TOKEN_OFFSET) % 4096)
|
| 113 |
+
l2.extend([
|
| 114 |
+
(slots[1] - CODE_TOKEN_OFFSET) % 4096, # Even
|
| 115 |
+
(slots[4] - CODE_TOKEN_OFFSET) % 4096, # Odd
|
| 116 |
+
])
|
| 117 |
+
l3.extend([
|
| 118 |
+
(slots[2] - CODE_TOKEN_OFFSET) % 4096,
|
| 119 |
+
(slots[3] - CODE_TOKEN_OFFSET) % 4096,
|
| 120 |
+
(slots[5] - CODE_TOKEN_OFFSET) % 4096,
|
| 121 |
+
(slots[6] - CODE_TOKEN_OFFSET) % 4096,
|
| 122 |
+
])
|
| 123 |
+
|
| 124 |
+
return [l1, l2, l3]
|
| 125 |
+
|
| 126 |
+
@torch.inference_mode()
|
| 127 |
+
def decode(
|
| 128 |
+
self,
|
| 129 |
+
snac_tokens: List[int],
|
| 130 |
+
use_sliding_window: bool = False
|
| 131 |
+
) -> Optional[np.ndarray]:
|
| 132 |
+
"""
|
| 133 |
+
Decode SNAC tokens to audio waveform.
|
| 134 |
+
|
| 135 |
+
Args:
|
| 136 |
+
snac_tokens: List of SNAC token IDs (7*n tokens)
|
| 137 |
+
use_sliding_window: If True, return only middle 2048 samples
|
| 138 |
+
(for smooth streaming without pops/clicks)
|
| 139 |
+
|
| 140 |
+
Returns:
|
| 141 |
+
Audio waveform as float32 numpy array, 24kHz mono
|
| 142 |
+
"""
|
| 143 |
+
if len(snac_tokens) < SNAC_TOKENS_PER_FRAME:
|
| 144 |
+
return None
|
| 145 |
+
|
| 146 |
+
# Unpack to 3 hierarchical levels
|
| 147 |
+
levels = self.unpack_snac_from_7(snac_tokens)
|
| 148 |
+
|
| 149 |
+
if not levels[0]:
|
| 150 |
+
return None
|
| 151 |
+
|
| 152 |
+
# Convert to tensors
|
| 153 |
+
codes = [
|
| 154 |
+
torch.tensor(level, dtype=torch.long, device=self.device).unsqueeze(0)
|
| 155 |
+
for level in levels
|
| 156 |
+
]
|
| 157 |
+
|
| 158 |
+
# Decode through SNAC quantizer + decoder
|
| 159 |
+
z_q = self.snac_model.quantizer.from_codes(codes)
|
| 160 |
+
audio = self.snac_model.decoder(z_q)
|
| 161 |
+
|
| 162 |
+
# Extract audio: [batch, 1, samples] → [samples]
|
| 163 |
+
audio = audio[0, 0].cpu().numpy()
|
| 164 |
+
|
| 165 |
+
# Sliding window mode: keep middle 2048 samples only
|
| 166 |
+
# This eliminates popping/cracking in streaming by overlapping windows
|
| 167 |
+
if use_sliding_window and len(audio) >= 4096:
|
| 168 |
+
audio = audio[2048:4096]
|
| 169 |
+
|
| 170 |
+
return audio
|
| 171 |
+
|
| 172 |
+
def decode_to_bytes(
|
| 173 |
+
self,
|
| 174 |
+
snac_tokens: List[int],
|
| 175 |
+
use_sliding_window: bool = False
|
| 176 |
+
) -> Optional[bytes]:
|
| 177 |
+
"""
|
| 178 |
+
Decode SNAC tokens to audio bytes (int16 PCM).
|
| 179 |
+
|
| 180 |
+
Args:
|
| 181 |
+
snac_tokens: List of SNAC token IDs
|
| 182 |
+
use_sliding_window: Use sliding window for smooth streaming
|
| 183 |
+
|
| 184 |
+
Returns:
|
| 185 |
+
Audio as bytes (int16 PCM, 24kHz mono)
|
| 186 |
+
"""
|
| 187 |
+
audio = self.decode(snac_tokens, use_sliding_window=use_sliding_window)
|
| 188 |
+
|
| 189 |
+
if audio is None:
|
| 190 |
+
return None
|
| 191 |
+
|
| 192 |
+
# Convert float32 to int16 PCM
|
| 193 |
+
audio_int16 = (audio * 32767).astype(np.int16)
|
| 194 |
+
return audio_int16.tobytes()
|
| 195 |
+
|
| 196 |
+
|
| 197 |
+
# ============================================================================
|
| 198 |
+
# CUSTOM LOGITS PROCESSOR
|
| 199 |
+
# ============================================================================
|
| 200 |
+
|
| 201 |
+
class OnlyAudioAfterSOS:
|
| 202 |
+
"""
|
| 203 |
+
Restricts vocabulary to SNAC codes + EOS after SOS token.
|
| 204 |
+
|
| 205 |
+
This prevents the model from generating text tokens during audio phase,
|
| 206 |
+
which would cause "hallucination" where the model repeats description text
|
| 207 |
+
instead of generating proper audio codes.
|
| 208 |
+
"""
|
| 209 |
+
|
| 210 |
+
def __init__(self):
|
| 211 |
+
self._seen_sos = False
|
| 212 |
+
|
| 213 |
+
def __call__(
|
| 214 |
+
self,
|
| 215 |
+
prompt_token_ids: List[int],
|
| 216 |
+
generated_token_ids: List[int],
|
| 217 |
+
logits: torch.Tensor,
|
| 218 |
+
) -> torch.Tensor:
|
| 219 |
+
"""
|
| 220 |
+
Apply constraint: after SOS, only allow SNAC codes + EOS.
|
| 221 |
+
|
| 222 |
+
Args:
|
| 223 |
+
prompt_token_ids: Original prompt token IDs
|
| 224 |
+
generated_token_ids: Tokens generated so far
|
| 225 |
+
logits: Logits for next token [vocab_size]
|
| 226 |
+
|
| 227 |
+
Returns:
|
| 228 |
+
Modified logits with masked tokens
|
| 229 |
+
"""
|
| 230 |
+
# Check if SOS has been generated
|
| 231 |
+
if not self._seen_sos:
|
| 232 |
+
all_token_ids = prompt_token_ids + generated_token_ids
|
| 233 |
+
if CODE_START_TOKEN_ID in all_token_ids:
|
| 234 |
+
self._seen_sos = True
|
| 235 |
+
else:
|
| 236 |
+
return logits # No constraint yet
|
| 237 |
+
|
| 238 |
+
# Apply constraint: mask all tokens except SNAC codes + EOS
|
| 239 |
+
mask = torch.full_like(logits, float('-inf'))
|
| 240 |
+
mask[SNAC_MIN_ID:SNAC_MAX_ID + 1] = 0 # Allow SNAC codes
|
| 241 |
+
mask[CODE_END_TOKEN_ID] = 0 # Allow EOS
|
| 242 |
+
|
| 243 |
+
return logits + mask
|
| 244 |
+
|
| 245 |
+
def reset(self):
|
| 246 |
+
"""Reset state for reuse across generations."""
|
| 247 |
+
self._seen_sos = False
|
| 248 |
+
|
| 249 |
+
|
| 250 |
+
# ============================================================================
|
| 251 |
+
# MAYA-1-VOICE MODEL
|
| 252 |
+
# ============================================================================
|
| 253 |
+
|
| 254 |
+
class Maya1VoiceModel:
|
| 255 |
+
"""
|
| 256 |
+
Maya-1-Voice TTS Model with VLLM inference engine.
|
| 257 |
+
|
| 258 |
+
Handles model loading, tokenizer initialization, and VLLM engine setup.
|
| 259 |
+
"""
|
| 260 |
+
|
| 261 |
+
def __init__(
|
| 262 |
+
self,
|
| 263 |
+
model_path: str,
|
| 264 |
+
dtype: str = "bfloat16",
|
| 265 |
+
max_model_len: int = 8192,
|
| 266 |
+
gpu_memory_utilization: float = 0.85,
|
| 267 |
+
):
|
| 268 |
+
"""
|
| 269 |
+
Initialize Maya-1-Voice model with VLLM.
|
| 270 |
+
|
| 271 |
+
Args:
|
| 272 |
+
model_path: Path to model checkpoint (local or HuggingFace)
|
| 273 |
+
dtype: Model precision (bfloat16 recommended)
|
| 274 |
+
max_model_len: Maximum sequence length
|
| 275 |
+
gpu_memory_utilization: GPU memory fraction to use (0.0-1.0)
|
| 276 |
+
"""
|
| 277 |
+
self.model_path = model_path
|
| 278 |
+
|
| 279 |
+
print(f"🚀 Initializing Maya-1-Voice Model")
|
| 280 |
+
print(f"📁 Model: {model_path}")
|
| 281 |
+
print(f"🔢 Dtype: {dtype}")
|
| 282 |
+
|
| 283 |
+
# Load tokenizer (must be from checkpoint with emotion tags)
|
| 284 |
+
print(f"📝 Loading tokenizer...")
|
| 285 |
+
self.tokenizer = AutoTokenizer.from_pretrained(
|
| 286 |
+
model_path,
|
| 287 |
+
trust_remote_code=True,
|
| 288 |
+
)
|
| 289 |
+
print(f"✅ Tokenizer loaded: {len(self.tokenizer)} tokens")
|
| 290 |
+
|
| 291 |
+
# Initialize VLLM async engine
|
| 292 |
+
print(f"🔧 Initializing VLLM engine...")
|
| 293 |
+
engine_args = AsyncEngineArgs(
|
| 294 |
+
model=model_path,
|
| 295 |
+
tokenizer=model_path,
|
| 296 |
+
dtype=dtype,
|
| 297 |
+
max_model_len=max_model_len,
|
| 298 |
+
gpu_memory_utilization=gpu_memory_utilization,
|
| 299 |
+
trust_remote_code=True,
|
| 300 |
+
)
|
| 301 |
+
|
| 302 |
+
self.engine = AsyncLLMEngine.from_engine_args(engine_args)
|
| 303 |
+
print(f"✅ VLLM engine ready")
|
| 304 |
+
|
| 305 |
+
def build_prompt(self, description: str, text: str) -> str:
|
| 306 |
+
"""
|
| 307 |
+
Build prompt in Maya-1-Voice format using chat template.
|
| 308 |
+
|
| 309 |
+
Format: Chat template with <description="..."> text as content
|
| 310 |
+
|
| 311 |
+
The model expects:
|
| 312 |
+
1. Description of voice/character
|
| 313 |
+
2. Text to synthesize (optionally with <emotion> tags)
|
| 314 |
+
|
| 315 |
+
Args:
|
| 316 |
+
description: Voice description
|
| 317 |
+
Example: "Realistic male voice in the 30s age with american accent.
|
| 318 |
+
Normal pitch, warm timbre, conversational pacing."
|
| 319 |
+
text: Text to synthesize
|
| 320 |
+
Example: "Hello world! <excited> This is amazing!"
|
| 321 |
+
|
| 322 |
+
Returns:
|
| 323 |
+
Formatted prompt string using chat template
|
| 324 |
+
"""
|
| 325 |
+
content = f'<description="{description}"> {text}'
|
| 326 |
+
messages = [{"role": "user", "content": content}]
|
| 327 |
+
return self.tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
|
| 328 |
+
|
| 329 |
+
|
| 330 |
+
# ============================================================================
|
| 331 |
+
# STREAMING PIPELINE
|
| 332 |
+
# ============================================================================
|
| 333 |
+
|
| 334 |
+
class Maya1VoiceStreamingPipeline:
|
| 335 |
+
"""
|
| 336 |
+
Streaming TTS pipeline using sliding window approach.
|
| 337 |
+
|
| 338 |
+
This generates smooth audio by:
|
| 339 |
+
1. Streaming tokens from VLLM as they're generated
|
| 340 |
+
2. Every 7 tokens, decoding the last 28 tokens (4 frames) - sliding window
|
| 341 |
+
3. Keeping only middle 2048 samples from each decode
|
| 342 |
+
4. Creating natural overlap between chunks for artifact-free playback
|
| 343 |
+
"""
|
| 344 |
+
|
| 345 |
+
def __init__(self, model: Maya1VoiceModel, snac_decoder: SNACDecoder):
|
| 346 |
+
"""Initialize streaming pipeline."""
|
| 347 |
+
self.model = model
|
| 348 |
+
self.snac_decoder = snac_decoder
|
| 349 |
+
print(f"🌊 Maya-1-Voice Streaming Pipeline initialized")
|
| 350 |
+
|
| 351 |
+
async def generate_speech_stream(
|
| 352 |
+
self,
|
| 353 |
+
description: str,
|
| 354 |
+
text: str,
|
| 355 |
+
temperature: float = DEFAULT_TEMPERATURE,
|
| 356 |
+
top_p: float = DEFAULT_TOP_P,
|
| 357 |
+
max_tokens: int = DEFAULT_MAX_TOKENS,
|
| 358 |
+
repetition_penalty: float = DEFAULT_REPETITION_PENALTY,
|
| 359 |
+
) -> AsyncGenerator[bytes, None]:
|
| 360 |
+
"""
|
| 361 |
+
Generate speech audio with streaming.
|
| 362 |
+
|
| 363 |
+
Args:
|
| 364 |
+
description: Voice/character description
|
| 365 |
+
text: Text to synthesize (with optional <emotion> tags)
|
| 366 |
+
temperature: Sampling temperature (lower = more stable)
|
| 367 |
+
top_p: Nucleus sampling
|
| 368 |
+
max_tokens: Max SNAC tokens to generate
|
| 369 |
+
repetition_penalty: Prevent repetition loops
|
| 370 |
+
|
| 371 |
+
Yields:
|
| 372 |
+
Audio chunks as bytes (int16 PCM, 24kHz mono)
|
| 373 |
+
"""
|
| 374 |
+
print(f"\n🌊 Starting streaming generation")
|
| 375 |
+
print(f"📝 Description: {description[:80]}...")
|
| 376 |
+
print(f"💬 Text: {text}")
|
| 377 |
+
|
| 378 |
+
# Build prompt
|
| 379 |
+
prompt = self.model.build_prompt(description, text)
|
| 380 |
+
|
| 381 |
+
# Configure sampling (removed custom logits processor for V1 compatibility)
|
| 382 |
+
sampling_params = SamplingParams(
|
| 383 |
+
temperature=temperature,
|
| 384 |
+
top_p=top_p,
|
| 385 |
+
max_tokens=max_tokens,
|
| 386 |
+
min_tokens=DEFAULT_MIN_TOKENS,
|
| 387 |
+
repetition_penalty=repetition_penalty,
|
| 388 |
+
stop_token_ids=[CODE_END_TOKEN_ID], # Stop on audio EOS
|
| 389 |
+
)
|
| 390 |
+
|
| 391 |
+
print(f"🎲 Sampling: temp={temperature}, top_p={top_p}, max_tokens={max_tokens}")
|
| 392 |
+
|
| 393 |
+
# Token buffer for sliding window
|
| 394 |
+
token_buffer = []
|
| 395 |
+
total_tokens = 0
|
| 396 |
+
total_chunks = 0
|
| 397 |
+
|
| 398 |
+
# Generate with VLLM
|
| 399 |
+
import uuid
|
| 400 |
+
import time
|
| 401 |
+
request_id = f"maya1voice-{uuid.uuid4().hex[:8]}-{int(time.time() * 1000000)}"
|
| 402 |
+
|
| 403 |
+
results_generator = self.model.engine.generate(
|
| 404 |
+
prompt=prompt,
|
| 405 |
+
sampling_params=sampling_params,
|
| 406 |
+
request_id=request_id,
|
| 407 |
+
)
|
| 408 |
+
|
| 409 |
+
# Stream tokens with sliding window decoding
|
| 410 |
+
async for request_output in results_generator:
|
| 411 |
+
generated_ids = request_output.outputs[0].token_ids
|
| 412 |
+
|
| 413 |
+
# Process only new tokens
|
| 414 |
+
new_tokens = generated_ids[total_tokens:]
|
| 415 |
+
total_tokens = len(generated_ids)
|
| 416 |
+
|
| 417 |
+
# Filter and buffer SNAC tokens only
|
| 418 |
+
for token_id in new_tokens:
|
| 419 |
+
if SNAC_MIN_ID <= token_id <= SNAC_MAX_ID:
|
| 420 |
+
token_buffer.append(token_id)
|
| 421 |
+
|
| 422 |
+
# Sliding window: process every 7 tokens when buffer > 27
|
| 423 |
+
# Take last 28 tokens (4 frames) for smooth overlap
|
| 424 |
+
if len(token_buffer) % 7 == 0 and len(token_buffer) > 27:
|
| 425 |
+
window_tokens = token_buffer[-28:]
|
| 426 |
+
|
| 427 |
+
# Decode with sliding window (returns middle 2048 samples)
|
| 428 |
+
audio_bytes = self.snac_decoder.decode_to_bytes(
|
| 429 |
+
window_tokens,
|
| 430 |
+
use_sliding_window=True
|
| 431 |
+
)
|
| 432 |
+
|
| 433 |
+
if audio_bytes:
|
| 434 |
+
total_chunks += 1
|
| 435 |
+
if total_chunks == 1:
|
| 436 |
+
print(f"🎵 First chunk decoded ({len(audio_bytes)} bytes)")
|
| 437 |
+
yield audio_bytes
|
| 438 |
+
|
| 439 |
+
print(f"✅ Streaming complete: {total_tokens} tokens → {total_chunks} chunks")
|
| 440 |
+
|
| 441 |
+
|
| 442 |
+
# ============================================================================
|
| 443 |
+
# MAIN EXAMPLE
|
| 444 |
+
# ============================================================================
|
| 445 |
+
|
| 446 |
+
async def main():
|
| 447 |
+
"""
|
| 448 |
+
Example usage of Maya-1-Voice streaming inference.
|
| 449 |
+
|
| 450 |
+
This demonstrates:
|
| 451 |
+
1. Model initialization
|
| 452 |
+
2. SNAC decoder setup
|
| 453 |
+
3. Streaming generation
|
| 454 |
+
4. Audio chunk handling
|
| 455 |
+
"""
|
| 456 |
+
|
| 457 |
+
# Configuration
|
| 458 |
+
MODEL_PATH = "/home/ubuntu/veena_temp/maya-1-voice" # Local model path
|
| 459 |
+
DEVICE = "cuda" if torch.cuda.is_available() else "cpu"
|
| 460 |
+
|
| 461 |
+
print("=" * 80)
|
| 462 |
+
print("Maya-1-Voice VLLM Streaming Inference Example")
|
| 463 |
+
print("=" * 80)
|
| 464 |
+
|
| 465 |
+
# Initialize model
|
| 466 |
+
model = Maya1VoiceModel(
|
| 467 |
+
model_path=MODEL_PATH,
|
| 468 |
+
dtype="bfloat16",
|
| 469 |
+
max_model_len=8192,
|
| 470 |
+
gpu_memory_utilization=0.8, # Reduced for available GPU memory (12GB free)
|
| 471 |
+
)
|
| 472 |
+
|
| 473 |
+
# Initialize SNAC decoder
|
| 474 |
+
snac_decoder = SNACDecoder(device=DEVICE)
|
| 475 |
+
|
| 476 |
+
# Create pipeline
|
| 477 |
+
pipeline = Maya1VoiceStreamingPipeline(model, snac_decoder)
|
| 478 |
+
|
| 479 |
+
# Example 1: Professional voice
|
| 480 |
+
description = (
|
| 481 |
+
"Realistic male voice in the 30s age with american accent. "
|
| 482 |
+
"Normal pitch, warm timbre, conversational pacing, neutral tone delivery at med intensity."
|
| 483 |
+
)
|
| 484 |
+
text = "Hello! This is a test of the Maya-1-Voice text-to-speech system."
|
| 485 |
+
|
| 486 |
+
print(f"\n{'='*80}")
|
| 487 |
+
print("Example 1: Professional Voice")
|
| 488 |
+
print(f"{'='*80}")
|
| 489 |
+
|
| 490 |
+
audio_chunks = []
|
| 491 |
+
async for chunk in pipeline.generate_speech_stream(
|
| 492 |
+
description=description,
|
| 493 |
+
text=text,
|
| 494 |
+
temperature=0.4,
|
| 495 |
+
max_tokens=500,
|
| 496 |
+
):
|
| 497 |
+
audio_chunks.append(chunk)
|
| 498 |
+
print(f"📦 Received chunk {len(audio_chunks)}: {len(chunk)} bytes")
|
| 499 |
+
|
| 500 |
+
# Combine chunks
|
| 501 |
+
full_audio = b''.join(audio_chunks)
|
| 502 |
+
print(f"\n✅ Total audio: {len(full_audio)} bytes ({len(full_audio)//2} samples, {len(full_audio)/2/24000:.2f}s)")
|
| 503 |
+
|
| 504 |
+
# Save audio (optional)
|
| 505 |
+
try:
|
| 506 |
+
import wave
|
| 507 |
+
output_file = "output_example1.wav"
|
| 508 |
+
with wave.open(output_file, 'wb') as wav:
|
| 509 |
+
wav.setnchannels(1) # Mono
|
| 510 |
+
wav.setsampwidth(2) # 16-bit
|
| 511 |
+
wav.setframerate(24000) # 24kHz
|
| 512 |
+
wav.writeframes(full_audio)
|
| 513 |
+
print(f"💾 Saved to {output_file}")
|
| 514 |
+
except ImportError:
|
| 515 |
+
print(f"⚠️ Install 'wave' module to save audio files")
|
| 516 |
+
|
| 517 |
+
# Example 2: Character voice with emotions
|
| 518 |
+
print(f"\n{'='*80}")
|
| 519 |
+
print("Example 2: Character Voice with Emotions")
|
| 520 |
+
print(f"{'='*80}")
|
| 521 |
+
|
| 522 |
+
description = (
|
| 523 |
+
"Creative, dark_villain character. Male voice in their 40s with british accent. "
|
| 524 |
+
"Low pitch, gravelly timbre, slow pacing, angry tone at high intensity."
|
| 525 |
+
)
|
| 526 |
+
text = "The darkness isn't coming... <angry> it's already here!"
|
| 527 |
+
|
| 528 |
+
audio_chunks = []
|
| 529 |
+
async for chunk in pipeline.generate_speech_stream(
|
| 530 |
+
description=description,
|
| 531 |
+
text=text,
|
| 532 |
+
temperature=0.5,
|
| 533 |
+
max_tokens=800,
|
| 534 |
+
):
|
| 535 |
+
audio_chunks.append(chunk)
|
| 536 |
+
print(f"📦 Received chunk {len(audio_chunks)}: {len(chunk)} bytes")
|
| 537 |
+
|
| 538 |
+
full_audio = b''.join(audio_chunks)
|
| 539 |
+
print(f"\n✅ Total audio: {len(full_audio)} bytes ({len(full_audio)//2} samples, {len(full_audio)/2/24000:.2f}s)")
|
| 540 |
+
|
| 541 |
+
# Save audio
|
| 542 |
+
try:
|
| 543 |
+
import wave
|
| 544 |
+
output_file = "output_example2.wav"
|
| 545 |
+
with wave.open(output_file, 'wb') as wav:
|
| 546 |
+
wav.setnchannels(1)
|
| 547 |
+
wav.setsampwidth(2)
|
| 548 |
+
wav.setframerate(24000)
|
| 549 |
+
wav.writeframes(full_audio)
|
| 550 |
+
print(f"💾 Saved to {output_file}")
|
| 551 |
+
except ImportError:
|
| 552 |
+
pass
|
| 553 |
+
|
| 554 |
+
print(f"\n{'='*80}")
|
| 555 |
+
print("🎉 Examples complete!")
|
| 556 |
+
print(f"{'='*80}")
|
| 557 |
+
|
| 558 |
+
|
| 559 |
+
if __name__ == "__main__":
|
| 560 |
+
# Run async main
|
| 561 |
+
asyncio.run(main())
|