Commits · Luigi/SmolVLM2-on-llama.cpp

reduce n_ctx to 512

1aba000

Luigi commited on Jun 17

fix imencode call

cc08312

Luigi commited on Jun 17

default to smallest model with q8 prcision, enable verbose mode, disable reset clip

65efb90

Luigi commited on Jun 17

fix imencode

3be8e88

Luigi commited on Jun 17

reduce jpg quality for smaller image footprint

4e5fc85

Luigi commited on Jun 17

show llama cpp version

9069c3e

Luigi commited on Jun 17

inject verbose message to debug window

69c8775

Luigi commited on Jun 17

add vebose mode switch

2881733

Luigi commited on Jun 17

increase n_ctx to 8192

4decc4b

Luigi commited on Jun 17

add ui component to allow user enabl or disable reset_clip per frame

5462ff3

Luigi commited on Jun 17

add debug to show which weight files we’re using this run

a459bee

Luigi commited on Jun 17

use all cpu cores

be5c239

Luigi commited on Jun 17

show cpu count in debug message

b56b6ec

Luigi commited on Jun 17

add q2_k precsion weights

07f3263

Luigi commited on Jun 16

add rich debug message and dedicated display ui

34cd1e5

Luigi commited on Jun 16

apply in-memory encoding instead of temp files

45c2159

Luigi commited on Jun 16

avoid memory leak

238a95a

Luigi commited on Jun 16

use more thread for inference

bd12f6b

Luigi commited on Jun 16

remove interval ui doublon

bdd1478

Luigi commited on Jun 16

bugifx on ui about model selection

e1ad065

Luigi commited on Jun 16

increase interval default to 3s

c9c43a8

Luigi commited on Jun 16

1. add more models,

5c50991

Luigi commited on Jun 16

reduce ctx and max tokens for performance

76a0b57

Luigi commited on Jun 16

minor update then add todos

65b3c3a

Luigi commited on Jun 16

resize frame to 384 x384 resolution

c1d8038

Luigi commited on Jun 13

add debug messages

36dacc6

Luigi commited on Jun 13

switch to gradio implementation as streamlit + webrtc requires turn server

970f416

Luigi commited on Jun 13

decouple inference from streaming

292fb3c

Luigi commited on Jun 13

set default interval to 3s

2529cb3

Luigi commited on Jun 13

slightly increase repeat_penalty to reduce token repetition

636baf9

Luigi commited on Jun 13

bugfix on 'NoneType' object has no attribute 'caption'

7b7ed26

Luigi commited on Jun 13

update

221e4b6

Luigi commited on Jun 13

add debug

abec2c1

Luigi commited on Jun 12

update

dd0d47d

Luigi commited on Jun 12

initial commit

ca97f63

Luigi commited on Jun 12

Spaces:

Luigi
/

SmolVLM2-on-llama.cpp

Running

Commit History

reduce n_ctx to 512

1aba000

fix imencode call

cc08312

default to smallest model with q8 prcision, enable verbose mode, disable reset clip

65efb90

fix imencode

3be8e88

reduce jpg quality for smaller image footprint

4e5fc85

show llama cpp version

9069c3e

inject verbose message to debug window

69c8775

add vebose mode switch

2881733

increase n_ctx to 8192

4decc4b

add ui component to allow user enabl or disable reset_clip per frame

5462ff3

add debug to show which weight files we’re using this run

a459bee

use all cpu cores

be5c239

show cpu count in debug message

b56b6ec

add q2_k precsion weights

07f3263

add rich debug message and dedicated display ui

34cd1e5

apply in-memory encoding instead of temp files

45c2159

avoid memory leak

238a95a

use more thread for inference

bd12f6b

remove interval ui doublon

bdd1478

bugifx on ui about model selection

e1ad065

increase interval default to 3s

c9c43a8

1. add more models,

5c50991

reduce ctx and max tokens for performance

76a0b57

minor update then add todos

65b3c3a

resize frame to 384 x384 resolution

c1d8038

add debug messages

36dacc6

switch to gradio implementation as streamlit + webrtc requires turn server

970f416

decouple inference from streaming

292fb3c

set default interval to 3s

2529cb3

slightly increase repeat_penalty to reduce token repetition

636baf9

bugfix on 'NoneType' object has no attribute 'caption'

7b7ed26

update

221e4b6

add debug

abec2c1

update

dd0d47d

initial commit

ca97f63

Commit History

reduce n_ctx to 512 1aba000

fix imencode call cc08312

default to smallest model with q8 prcision, enable verbose mode, disable reset clip 65efb90

fix imencode 3be8e88

reduce jpg quality for smaller image footprint 4e5fc85

show llama cpp version 9069c3e

inject verbose message to debug window 69c8775

add vebose mode switch 2881733

increase n_ctx to 8192 4decc4b

add ui component to allow user enabl or disable reset_clip per frame 5462ff3

add debug to show which weight files we’re using this run a459bee

use all cpu cores be5c239

show cpu count in debug message b56b6ec

add q2_k precsion weights 07f3263

add rich debug message and dedicated display ui 34cd1e5

apply in-memory encoding instead of temp files 45c2159

avoid memory leak 238a95a

use more thread for inference bd12f6b

remove interval ui doublon bdd1478

bugifx on ui about model selection e1ad065

increase interval default to 3s c9c43a8

1. add more models, 5c50991

reduce ctx and max tokens for performance 76a0b57

minor update then add todos 65b3c3a

resize frame to 384 x384 resolution c1d8038

add debug messages 36dacc6

switch to gradio implementation as streamlit + webrtc requires turn server 970f416

decouple inference from streaming 292fb3c

set default interval to 3s 2529cb3

slightly increase repeat_penalty to reduce token repetition 636baf9

bugfix on 'NoneType' object has no attribute 'caption' 7b7ed26

update 221e4b6

add debug abec2c1

update dd0d47d

initial commit ca97f63

reduce n_ctx to 512

1aba000

fix imencode call

cc08312

default to smallest model with q8 prcision, enable verbose mode, disable reset clip

65efb90

fix imencode

3be8e88

reduce jpg quality for smaller image footprint

4e5fc85

show llama cpp version

9069c3e

inject verbose message to debug window

69c8775

add vebose mode switch

2881733

increase n_ctx to 8192

4decc4b

add ui component to allow user enabl or disable reset_clip per frame

5462ff3

add debug to show which weight files we’re using this run

a459bee

use all cpu cores

be5c239

show cpu count in debug message

b56b6ec

add q2_k precsion weights

07f3263

add rich debug message and dedicated display ui

34cd1e5

apply in-memory encoding instead of temp files

45c2159

avoid memory leak

238a95a

use more thread for inference

bd12f6b

remove interval ui doublon

bdd1478

bugifx on ui about model selection

e1ad065

increase interval default to 3s

c9c43a8

1. add more models,

5c50991

reduce ctx and max tokens for performance

76a0b57

minor update then add todos

65b3c3a

resize frame to 384 x384 resolution

c1d8038

add debug messages

36dacc6

switch to gradio implementation as streamlit + webrtc requires turn server

970f416

decouple inference from streaming

292fb3c

set default interval to 3s

2529cb3

slightly increase repeat_penalty to reduce token repetition

636baf9

bugfix on 'NoneType' object has no attribute 'caption'

7b7ed26

update

221e4b6

add debug

abec2c1

update

dd0d47d

initial commit

ca97f63