artificial-styletts2 / audiocraft

Commit History

determinis

bc7f42e

Dionyssos commited on Apr 17

4.49

4eabff6

Dionyssos commited on Apr 13

AudioGen dtypes

0230db1

Dionyssos commited on Apr 13

fx pos sinus

4898f81

Dionyssos commited on Apr 10

debug static on sound prompts

8e60374

Dionyssos commited on Apr 9

fx overwriting pattern antidiagonals

bc08da5

Dionyssos commited on Apr 9

pattern no overwrite 2048

999bd69

Dionyssos commited on Apr 9

preserve only few last kv

17a68db

Dionyssos commited on Apr 8

try L2 kv

cf02fb0

Dionyssos commited on Apr 8

warn non-atomic kv cache fill

4795b97

Dionyssos commited on Feb 25

offset

8a9a2fe

Dionyssos commited on Feb 24

antidiagonal fast delay pattern

376e6a0

Dionyssos commited on Feb 24

numerals

b399825

Dionyssos commited on Feb 23

fx sounds batch inference

86b9ce4

Dionyssos commited on Feb 20

del adhoc soundscape fading

df63ff0

Dionyssos commited on Feb 3

fx live dem

bb2cd38

Dionyssos commited on Jan 21

soudscape discard last 1s from AudioGen - avoids splash sound

a84b206

Dionyssos commited on Jan 11

debug long sounds

3ac9f34

Dionyssos commited on Nov 5, 2024

apply null cond lm.forw()

6cb7713

Dionyssos commited on Nov 3, 2024

revert pattern preserves 4

dcfe0d4

Dionyssos commited on Nov 2, 2024

AudioGen class

54adc39

Dionyssos commited on Nov 2, 2024

distortion on barging DEBUG

e70ad00

Dionyssos commited on Nov 2, 2024

del diffusion [unused]

3dfbf53

Dionyssos commited on Nov 1, 2024

instantiate audiogen in demo

d912185

Dionyssos commited on Nov 1, 2024

fx k,v history after project

60fbcf9

Dionyssos commited on Oct 29, 2024

kv history @ each transf layer

731cb10

Dionyssos commited on Oct 28, 2024

DEBUG steaming past key changes 47x during next token calc

e83a997

Dionyssos commited on Oct 25, 2024

DEBUG: cross_attention_src = query or key?

a0ce150

Dionyssos commited on Oct 23, 2024

revert transformer.py

2e6c69d

Dionyssos commited on Oct 21, 2024

LM has virtual 2nd batch

3eec6d2

Dionyssos commited on Oct 18, 2024

duplicate xN_DRAW - for long gen

d9889a1

Dionyssos commited on Oct 17, 2024

index out of bounds 2048 @ dequantize()

27d24be

Dionyssos commited on Oct 16, 2024

prompt_token defaults to torch.zeros()

8639464

Dionyssos commited on Oct 16, 2024

fx

d8e2a3d

Dionyssos commited on Oct 16, 2024

cleanup Vq

fe62fb4

Dionyssos commited on Oct 16, 2024

clean unused functions

e366cd5

Dionyssos commited on Oct 15, 2024

debug special token

0a8807e

Dionyssos commited on Oct 15, 2024

cleanup 3

5067878

Dionyssos commited on Oct 11, 2024

cleanup2

c5e1f80

Dionyssos commited on Oct 11, 2024

del xtr funs

531e776

Dionyssos commited on Oct 10, 2024

clean unused funs

06aa0fc

Dionyssos commited on Oct 10, 2024

links

4377106

Dionyssos commited on Oct 2, 2024

freq = 16000

2d0e2b6

Dionyssos commited on Sep 30, 2024

audiocraft

d72b2c3

Dionyssos commited on Sep 24, 2024

Commit History

determinis bc7f42e

4.49 4eabff6

AudioGen dtypes 0230db1

fx pos sinus 4898f81

debug static on sound prompts 8e60374

fx overwriting pattern antidiagonals bc08da5

pattern no overwrite 2048 999bd69

preserve only few last kv 17a68db

try L2 kv cf02fb0

warn non-atomic kv cache fill 4795b97

offset 8a9a2fe

antidiagonal fast delay pattern 376e6a0

numerals b399825

fx sounds batch inference 86b9ce4

del adhoc soundscape fading df63ff0

fx live dem bb2cd38

soudscape discard last 1s from AudioGen - avoids splash sound a84b206

debug long sounds 3ac9f34

apply null cond lm.forw() 6cb7713

revert pattern preserves 4 dcfe0d4

AudioGen class 54adc39

distortion on barging DEBUG e70ad00

del diffusion [unused] 3dfbf53

instantiate audiogen in demo d912185

fx k,v history after project 60fbcf9

kv history @ each transf layer 731cb10

DEBUG steaming past key changes 47x during next token calc e83a997

DEBUG: cross_attention_src = query or key? a0ce150

revert transformer.py 2e6c69d

LM has virtual 2nd batch 3eec6d2

duplicate xN_DRAW - for long gen d9889a1

index out of bounds 2048 @ dequantize() 27d24be

prompt_token defaults to torch.zeros() 8639464

fx d8e2a3d

cleanup Vq fe62fb4

clean unused functions e366cd5

debug special token 0a8807e

cleanup 3 5067878

cleanup2 c5e1f80

del xtr funs 531e776

clean unused funs 06aa0fc

links 4377106

freq = 16000 2d0e2b6

audiocraft d72b2c3

determinis

bc7f42e

4.49

4eabff6

AudioGen dtypes

0230db1

fx pos sinus

4898f81

debug static on sound prompts

8e60374

fx overwriting pattern antidiagonals

bc08da5

pattern no overwrite 2048

999bd69

preserve only few last kv

17a68db

try L2 kv

cf02fb0

warn non-atomic kv cache fill

4795b97

offset

8a9a2fe

antidiagonal fast delay pattern

376e6a0

numerals

b399825

fx sounds batch inference

86b9ce4

del adhoc soundscape fading

df63ff0

fx live dem

bb2cd38

soudscape discard last 1s from AudioGen - avoids splash sound

a84b206

debug long sounds

3ac9f34

apply null cond lm.forw()

6cb7713

revert pattern preserves 4

dcfe0d4

AudioGen class

54adc39

distortion on barging DEBUG

e70ad00

del diffusion [unused]

3dfbf53

instantiate audiogen in demo

d912185

fx k,v history after project

60fbcf9

kv history @ each transf layer

731cb10

DEBUG steaming past key changes 47x during next token calc

e83a997

DEBUG: cross_attention_src = query or key?

a0ce150

revert transformer.py

2e6c69d

LM has virtual 2nd batch

3eec6d2

duplicate xN_DRAW - for long gen

d9889a1

index out of bounds 2048 @ dequantize()

27d24be

prompt_token defaults to torch.zeros()

8639464

fx

d8e2a3d

cleanup Vq

fe62fb4

clean unused functions

e366cd5

debug special token

0a8807e

cleanup 3

5067878

cleanup2

c5e1f80

del xtr funs

531e776

clean unused funs

06aa0fc

links

4377106

freq = 16000

2d0e2b6

audiocraft

d72b2c3