Hugging Face's logo

yueyulin
/
rwkv_asr

Automatic Speech Recognition
Safetensors
Chinese
English
Model card Files
xet
Community
yueyulin commited on
Commit
9286472
·
verified ·
1 Parent(s): 9de26d2

Upload folder using huggingface_hub

Browse files
Files changed (15) hide show
  1. asr_gradio/rwkv7_0.1b_audio_lm_latents_1.5b_242k/config.json +49 -0
  2. asr_gradio/rwkv7_0.1b_audio_lm_latents_1.5b_242k/model.safetensors +3 -0
  3. asr_gradio/rwkv7_0.1b_audio_lm_latents_1.5b_242k/model_converted.pth +3 -0
  4. asr_gradio/rwkv7_0.1b_audio_lm_latents_1.5b_242k/projector1.pt +3 -0
  5. asr_gradio/rwkv7_0.1b_audio_lm_latents_1.5b_242k/projector2.pt +3 -0
  6. asr_gradio/rwkv7_0.1b_audio_lm_latents_1.5b_252k/config.json +49 -0
  7. asr_gradio/rwkv7_0.1b_audio_lm_latents_1.5b_252k/model.safetensors +3 -0
  8. asr_gradio/rwkv7_0.1b_audio_lm_latents_1.5b_252k/model_converted.pth +3 -0
  9. asr_gradio/rwkv7_0.1b_audio_lm_latents_1.5b_252k/projector1.pt +3 -0
  10. asr_gradio/rwkv7_0.1b_audio_lm_latents_1.5b_252k/projector2.pt +3 -0
  11. asr_gradio/rwkv7_0.1b_audio_lm_latents_1.5b_262k/config.json +49 -0
  12. asr_gradio/rwkv7_0.1b_audio_lm_latents_1.5b_262k/model.safetensors +3 -0
  13. asr_gradio/rwkv7_0.1b_audio_lm_latents_1.5b_262k/model_converted.pth +3 -0
  14. asr_gradio/rwkv7_0.1b_audio_lm_latents_1.5b_262k/projector1.pt +3 -0
  15. asr_gradio/rwkv7_0.1b_audio_lm_latents_1.5b_262k/projector2.pt +3 -0
asr_gradio/rwkv7_0.1b_audio_lm_latents_1.5b_242k/config.json ADDED
@@ -0,0 +1,49 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "a_low_rank_dim": 64,
3
+ "architectures": [
4
+ "RWKV7ModelForLatentInputs"
5
+ ],
6
+ "attn": null,
7
+ "attn_mode": "chunk",
8
+ "bos_token_id": 0,
9
+ "decay_low_rank_dim": 64,
10
+ "eos_token_id": 0,
11
+ "fuse_cross_entropy": true,
12
+ "fuse_linear_cross_entropy": false,
13
+ "fuse_norm": false,
14
+ "gate_low_rank_dim": 128,
15
+ "head_dim": 64,
16
+ "hidden_act": "sqrelu",
17
+ "hidden_ratio": 4.0,
18
+ "hidden_size": 768,
19
+ "initializer_range": 0.006,
20
+ "intermediate_size": 3072,
21
+ "max_position_embeddings": 2048,
22
+ "model_type": "rwkv7",
23
+ "norm_bias": true,
24
+ "norm_eps": 1e-05,
25
+ "norm_first": true,
26
+ "num_heads": 32,
27
+ "num_hidden_layers": 12,
28
+ "tie_word_embeddings": false,
29
+ "torch_dtype": "float32",
30
+ "transformers_version": "4.52.4",
31
+ "use_cache": true,
32
+ "use_l2warp": true,
33
+ "v_low_rank_dim": 32,
34
+ "value_dim": [
35
+ 768,
36
+ 768,
37
+ 768,
38
+ 768,
39
+ 768,
40
+ 768,
41
+ 768,
42
+ 768,
43
+ 768,
44
+ 768,
45
+ 768,
46
+ 768
47
+ ],
48
+ "vocab_size": 10
49
+ }
asr_gradio/rwkv7_0.1b_audio_lm_latents_1.5b_242k/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9dd70716cf94d664f523899a563669feb8c74fa8252622ce76a544d92dac0fae
3
+ size 361523904
asr_gradio/rwkv7_0.1b_audio_lm_latents_1.5b_242k/model_converted.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e6e54ead7875b820d4cd3296524c0e29ec768f8ad3e63a642c60fd28d055f994
3
+ size 361668745
asr_gradio/rwkv7_0.1b_audio_lm_latents_1.5b_242k/projector1.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2782b1165cadcd6b5d75280a482b22abf43c8ae71a6f6c07beaf2f0edd63365a
3
+ size 3937149
asr_gradio/rwkv7_0.1b_audio_lm_latents_1.5b_242k/projector2.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2b0ded11c90e5311e067e22d425cde539b483382d975700cbc57a42e658bf4ab
3
+ size 6301565
asr_gradio/rwkv7_0.1b_audio_lm_latents_1.5b_252k/config.json ADDED
@@ -0,0 +1,49 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "a_low_rank_dim": 64,
3
+ "architectures": [
4
+ "RWKV7ModelForLatentInputs"
5
+ ],
6
+ "attn": null,
7
+ "attn_mode": "chunk",
8
+ "bos_token_id": 0,
9
+ "decay_low_rank_dim": 64,
10
+ "eos_token_id": 0,
11
+ "fuse_cross_entropy": true,
12
+ "fuse_linear_cross_entropy": false,
13
+ "fuse_norm": false,
14
+ "gate_low_rank_dim": 128,
15
+ "head_dim": 64,
16
+ "hidden_act": "sqrelu",
17
+ "hidden_ratio": 4.0,
18
+ "hidden_size": 768,
19
+ "initializer_range": 0.006,
20
+ "intermediate_size": 3072,
21
+ "max_position_embeddings": 2048,
22
+ "model_type": "rwkv7",
23
+ "norm_bias": true,
24
+ "norm_eps": 1e-05,
25
+ "norm_first": true,
26
+ "num_heads": 32,
27
+ "num_hidden_layers": 12,
28
+ "tie_word_embeddings": false,
29
+ "torch_dtype": "float32",
30
+ "transformers_version": "4.52.4",
31
+ "use_cache": true,
32
+ "use_l2warp": true,
33
+ "v_low_rank_dim": 32,
34
+ "value_dim": [
35
+ 768,
36
+ 768,
37
+ 768,
38
+ 768,
39
+ 768,
40
+ 768,
41
+ 768,
42
+ 768,
43
+ 768,
44
+ 768,
45
+ 768,
46
+ 768
47
+ ],
48
+ "vocab_size": 10
49
+ }
asr_gradio/rwkv7_0.1b_audio_lm_latents_1.5b_252k/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3c696d5461084d4bd54fb04ffa367dd98789104ae04136b6bb92b1e7452db28a
3
+ size 361523904
asr_gradio/rwkv7_0.1b_audio_lm_latents_1.5b_252k/model_converted.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ff953542c5c603a26f21c08b29ba1e3c5a3f637669a9b78fe453a3f7ed438740
3
+ size 361668745
asr_gradio/rwkv7_0.1b_audio_lm_latents_1.5b_252k/projector1.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f560b0d2ace7d7aba504aeb33a7683e58a96298f857c4399116dd7cf7858c100
3
+ size 3937149
asr_gradio/rwkv7_0.1b_audio_lm_latents_1.5b_252k/projector2.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1f2496f217230c298979d2ba5947afd36e45887033efae04021c4566b158ebe5
3
+ size 6301565
asr_gradio/rwkv7_0.1b_audio_lm_latents_1.5b_262k/config.json ADDED
@@ -0,0 +1,49 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "a_low_rank_dim": 64,
3
+ "architectures": [
4
+ "RWKV7ModelForLatentInputs"
5
+ ],
6
+ "attn": null,
7
+ "attn_mode": "chunk",
8
+ "bos_token_id": 0,
9
+ "decay_low_rank_dim": 64,
10
+ "eos_token_id": 0,
11
+ "fuse_cross_entropy": true,
12
+ "fuse_linear_cross_entropy": false,
13
+ "fuse_norm": false,
14
+ "gate_low_rank_dim": 128,
15
+ "head_dim": 64,
16
+ "hidden_act": "sqrelu",
17
+ "hidden_ratio": 4.0,
18
+ "hidden_size": 768,
19
+ "initializer_range": 0.006,
20
+ "intermediate_size": 3072,
21
+ "max_position_embeddings": 2048,
22
+ "model_type": "rwkv7",
23
+ "norm_bias": true,
24
+ "norm_eps": 1e-05,
25
+ "norm_first": true,
26
+ "num_heads": 32,
27
+ "num_hidden_layers": 12,
28
+ "tie_word_embeddings": false,
29
+ "torch_dtype": "float32",
30
+ "transformers_version": "4.52.4",
31
+ "use_cache": true,
32
+ "use_l2warp": true,
33
+ "v_low_rank_dim": 32,
34
+ "value_dim": [
35
+ 768,
36
+ 768,
37
+ 768,
38
+ 768,
39
+ 768,
40
+ 768,
41
+ 768,
42
+ 768,
43
+ 768,
44
+ 768,
45
+ 768,
46
+ 768
47
+ ],
48
+ "vocab_size": 10
49
+ }
asr_gradio/rwkv7_0.1b_audio_lm_latents_1.5b_262k/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:14024323e0b9fbdc3d1738ff05d169883f15b421c1435727a99cc2a61e95b65c
3
+ size 361523904
asr_gradio/rwkv7_0.1b_audio_lm_latents_1.5b_262k/model_converted.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5669a168531edaed3168c67cb66449271c9eec8c9466d67025ced5a3d681dffb
3
+ size 361668745
asr_gradio/rwkv7_0.1b_audio_lm_latents_1.5b_262k/projector1.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e94d905528b9f63f26feb0e2a5129e9f1c435eaa9c78cee700ffd6d7a8423fff
3
+ size 3937149
asr_gradio/rwkv7_0.1b_audio_lm_latents_1.5b_262k/projector2.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5ee9b5df2d6fca873adc65bc3741d57d502e3ef39455cf40546bfd516589235c
3
+ size 6301565