528
EasyControl Ghibli
๐ฆ
New Ghibli EasyControl model is now released!!
Wok
Wok
AI & ML interests
๐ฆ
Recent Activity
liked
a Space
1 day ago
jamesliu1217/EasyControl_Ghibli
liked
a Space
5 days ago
black-forest-labs/FLUX.1-Redux-dev
liked
a Space
about 1 month ago
linoyts/scribble-sdxl-flash
Organizations
Wok's activity
liked
a
Space
1 day ago
liked
a
Space
5 days ago
liked
a
Space
about 1 month ago
liked
2
Spaces
2 months ago
reacted to
m-ric's
post with ๐ฅ
7 months ago
Post
3398
๐ฅ ๐๐ฐ๐๐ง ๐ซ๐๐ฅ๐๐๐ฌ๐๐ฌ ๐ญ๐ก๐๐ข๐ซ ๐.๐ ๐๐๐ฆ๐ข๐ฅ๐ฒ ๐จ๐ ๐ฆ๐จ๐๐๐ฅ๐ฌ: ๐๐๐ฐ ๐๐๐๐ ๐๐จ๐ซ ๐๐ฅ๐ฅ ๐ฌ๐ข๐ณ๐๐ฌ ๐ฎ๐ฉ ๐ญ๐จ ๐๐๐!
The Chinese LLM maker just dropped a flurry of different models, ensuring there will be a Qwen SOTA model for every application out there:
Qwen2.5: 0.5B, 1.5B, 3B, 7B, 14B, 32B, and 72B
Qwen2.5-Coder: 1.5B, 7B, and 32B on the way
Qwen2.5-Math: 1.5B, 7B, and 72B.
And they didn't sleep: the performance is top of the game for each weight category!
๐๐๐ฒ ๐ข๐ง๐ฌ๐ข๐ ๐ก๐ญ๐ฌ:
๐ All models have ๐ญ๐ฎ๐ด๐ธ ๐๐ผ๐ธ๐ฒ๐ป ๐ฐ๐ผ๐ป๐๐ฒ๐ ๐ ๐น๐ฒ๐ป๐ด๐๐ต
๐ Models pre-trained on 18T tokens, even longer than the 15T of Llama-3
๐ช The flagship ๐ค๐๐ฒ๐ป๐ฎ.๐ฑ-๐ณ๐ฎ๐ ๐ถ๐ ~๐ฐ๐ผ๐บ๐ฝ๐ฒ๐๐ถ๐๐ถ๐๐ฒ ๐๐ถ๐๐ต ๐๐น๐ฎ๐บ๐ฎ-๐ฏ.๐ญ-๐ฐ๐ฌ๐ฑ๐, ๐ฎ๐ป๐ฑ ๐ต๐ฎ๐ ๐ฎ ๐ฏ-๐ฑ% ๐บ๐ฎ๐ฟ๐ด๐ถ๐ป ๐ผ๐ป ๐๐น๐ฎ๐บ๐ฎ-๐ฏ.๐ญ-๐ณ๐ฌ๐ ๐ผ๐ป ๐บ๐ผ๐๐ ๐ฏ๐ฒ๐ป๐ฐ๐ต๐บ๐ฎ๐ฟ๐ธ๐.
๐ซ๐ท On top of this, it ๐๐ฎ๐ธ๐ฒ๐ ๐๐ต๐ฒ #๐ญ ๐๐ฝ๐ผ๐ ๐ผ๐ป ๐บ๐๐น๐๐ถ๐น๐ถ๐ป๐ด๐๐ฎ๐น ๐๐ฎ๐๐ธ๐ so it might become my standard for French
๐ป Qwen2.5-Coder is only 7B but beats competing models up to 33B (DeeSeek-Coder 33B-Instruct). Let's wait for their 32B to come out!
๐งฎ Qwen2.5-Math sets a new high in the ratio of MATH benchmark score to # of parameters. They trained it by "aggregating more high-quality mathematical data, particularly in Chinese, from web sources, books, and codes across multiple recall cycles."
๐ Technical report to be released "very soon"
๐ All models have the most permissive license apache2.0, except the 72B models that have a custom license mentioning "you can use it for free EXCEPT if your product has over 100M users"
๐ค All models are available on the HF Hub! โก๏ธ Qwen/qwen25-66e81a666513e518adb90d9e
The Chinese LLM maker just dropped a flurry of different models, ensuring there will be a Qwen SOTA model for every application out there:
Qwen2.5: 0.5B, 1.5B, 3B, 7B, 14B, 32B, and 72B
Qwen2.5-Coder: 1.5B, 7B, and 32B on the way
Qwen2.5-Math: 1.5B, 7B, and 72B.
And they didn't sleep: the performance is top of the game for each weight category!
๐๐๐ฒ ๐ข๐ง๐ฌ๐ข๐ ๐ก๐ญ๐ฌ:
๐ All models have ๐ญ๐ฎ๐ด๐ธ ๐๐ผ๐ธ๐ฒ๐ป ๐ฐ๐ผ๐ป๐๐ฒ๐ ๐ ๐น๐ฒ๐ป๐ด๐๐ต
๐ Models pre-trained on 18T tokens, even longer than the 15T of Llama-3
๐ช The flagship ๐ค๐๐ฒ๐ป๐ฎ.๐ฑ-๐ณ๐ฎ๐ ๐ถ๐ ~๐ฐ๐ผ๐บ๐ฝ๐ฒ๐๐ถ๐๐ถ๐๐ฒ ๐๐ถ๐๐ต ๐๐น๐ฎ๐บ๐ฎ-๐ฏ.๐ญ-๐ฐ๐ฌ๐ฑ๐, ๐ฎ๐ป๐ฑ ๐ต๐ฎ๐ ๐ฎ ๐ฏ-๐ฑ% ๐บ๐ฎ๐ฟ๐ด๐ถ๐ป ๐ผ๐ป ๐๐น๐ฎ๐บ๐ฎ-๐ฏ.๐ญ-๐ณ๐ฌ๐ ๐ผ๐ป ๐บ๐ผ๐๐ ๐ฏ๐ฒ๐ป๐ฐ๐ต๐บ๐ฎ๐ฟ๐ธ๐.
๐ซ๐ท On top of this, it ๐๐ฎ๐ธ๐ฒ๐ ๐๐ต๐ฒ #๐ญ ๐๐ฝ๐ผ๐ ๐ผ๐ป ๐บ๐๐น๐๐ถ๐น๐ถ๐ป๐ด๐๐ฎ๐น ๐๐ฎ๐๐ธ๐ so it might become my standard for French
๐ป Qwen2.5-Coder is only 7B but beats competing models up to 33B (DeeSeek-Coder 33B-Instruct). Let's wait for their 32B to come out!
๐งฎ Qwen2.5-Math sets a new high in the ratio of MATH benchmark score to # of parameters. They trained it by "aggregating more high-quality mathematical data, particularly in Chinese, from web sources, books, and codes across multiple recall cycles."
๐ Technical report to be released "very soon"
๐ All models have the most permissive license apache2.0, except the 72B models that have a custom license mentioning "you can use it for free EXCEPT if your product has over 100M users"
๐ค All models are available on the HF Hub! โก๏ธ Qwen/qwen25-66e81a666513e518adb90d9e
- 2 replies
reacted to
Tonic's
post with ๐ฅ
7 months ago
Post
2750
๐๐ปโโ๏ธHey there folks ,
@ucaslcl released a new OCR model , that's๐๐ป๐๐ป fantastic : https://huggingface.co/ucaslcl/GOT-OCR2_0
GPU : Tonic/GOT-OCR
Gradio Demo (Image Edit) : Tonic1/ImageEdit-GOT-OCR
Model : https://huggingface.co/ucaslcl/GOT-OCR2_0
Official demo : https://huggingface.co/spaces/ucaslcl/GOT_online
github : https://github.com/Ucas-HaoranWei/GOT-OCR2.0
@ucaslcl released a new OCR model , that's๐๐ป๐๐ป fantastic : https://huggingface.co/ucaslcl/GOT-OCR2_0
GPU : Tonic/GOT-OCR
Gradio Demo (Image Edit) : Tonic1/ImageEdit-GOT-OCR
Model : https://huggingface.co/ucaslcl/GOT-OCR2_0
Official demo : https://huggingface.co/spaces/ucaslcl/GOT_online
github : https://github.com/Ucas-HaoranWei/GOT-OCR2.0
-
- 4 replies
reacted to
Tonic's
post with ๐
7 months ago
Post
1100
๐๐ปโโ๏ธHey there folks,
Nvidia just released a small 4B Nemotron-mini model , and it works surprisingly well !
you can check it out here :
base : nvidia/Minitron-4B-Base
instruct : nvidia/Nemotron-Mini-4B-Instruct
demo : https://huggingface.co/spaces/Tonic/Nemotron-Mini-4B
hoep you like it ๐ค๐ค
Nvidia just released a small 4B Nemotron-mini model , and it works surprisingly well !
you can check it out here :
base : nvidia/Minitron-4B-Base
instruct : nvidia/Nemotron-Mini-4B-Instruct
demo : https://huggingface.co/spaces/Tonic/Nemotron-Mini-4B
hoep you like it ๐ค๐ค
reacted to
KingNish's
post with ๐ฅ
7 months ago
Post
3916
I am experimenting with Flux and trying to push it to its limits without training (as I am GPU-poor ๐
).
I found some flaws in the pipelines, which I resolved, and now I am able to generate an approx similar quality image as Flux Schnell 4 steps in just 1 step.
Demo Link:
KingNish/Realtime-FLUX
I found some flaws in the pipelines, which I resolved, and now I am able to generate an approx similar quality image as Flux Schnell 4 steps in just 1 step.
Demo Link:
KingNish/Realtime-FLUX
- 1 reply
reacted to
fffiloni's
post with ๐
7 months ago
Post
19541
๐ซ๐ท
Quel impact de lโIA sur les filiรจres du cinรฉma, de lโaudiovisuel et du jeu vidรฉo?
Etude prospective ร destination des professionnels
โ CNC & BearingPoint | 09/04/2024
Si lโIntelligence Artificielle (IA) est utilisรฉe de longue date dans les secteurs du cinรฉma, de lโaudiovisuel et du jeu vidรฉo, les nouvelles applications de lโIA gรฉnรฉrative bousculent notre vision de ce dont est capable une machine et possรจdent un potentiel de transformation inรฉdit. Elles impressionnent par la qualitรฉ de leurs productions et suscitent par consรฉquent de nombreux dรฉbats, entre attentes et apprรฉhensions.
Le CNC a donc dรฉcider de lancer un nouvel Observatoire de lโIA Afin de mieux comprendre les usages de lโIA et ses impacts rรฉels sur la filiรจre de lโimage. Dans le cadre de cet Observatoire, le CNC a souhaitรฉ dresser un premier รฉtat des lieux ร travers la cartographie des usages actuels ou potentiels de lโIA ร chaque รฉtape du processus de crรฉation et de diffusion dโune ลuvre, en identifiant les opportunitรฉs et risques associรฉs, notamment en termes de mรฉtiers et dโemploi. Cette รฉtude CNC / Bearing Point en a prรฉsentรฉ les principaux enseignements le 6 mars, lors de la journรฉe CNC ยซ Crรฉer, produire, diffuser ร lโheure de lโintelligence artificielle ยป.
Le CNC publie la version augmentรฉe de la cartographie des usages de lโIA dans les filiรจres du cinรฉma, de lโaudiovisuel et du jeu vidรฉo.
Lien vers la cartographie complรจte: https://www.cnc.fr/documents/36995/2097582/Cartographie+des+usages+IA_rapport+complet.pdf/96532829-747e-b85e-c74b-af313072cab7?t=1712309387891
Quel impact de lโIA sur les filiรจres du cinรฉma, de lโaudiovisuel et du jeu vidรฉo?
Etude prospective ร destination des professionnels
โ CNC & BearingPoint | 09/04/2024
Si lโIntelligence Artificielle (IA) est utilisรฉe de longue date dans les secteurs du cinรฉma, de lโaudiovisuel et du jeu vidรฉo, les nouvelles applications de lโIA gรฉnรฉrative bousculent notre vision de ce dont est capable une machine et possรจdent un potentiel de transformation inรฉdit. Elles impressionnent par la qualitรฉ de leurs productions et suscitent par consรฉquent de nombreux dรฉbats, entre attentes et apprรฉhensions.
Le CNC a donc dรฉcider de lancer un nouvel Observatoire de lโIA Afin de mieux comprendre les usages de lโIA et ses impacts rรฉels sur la filiรจre de lโimage. Dans le cadre de cet Observatoire, le CNC a souhaitรฉ dresser un premier รฉtat des lieux ร travers la cartographie des usages actuels ou potentiels de lโIA ร chaque รฉtape du processus de crรฉation et de diffusion dโune ลuvre, en identifiant les opportunitรฉs et risques associรฉs, notamment en termes de mรฉtiers et dโemploi. Cette รฉtude CNC / Bearing Point en a prรฉsentรฉ les principaux enseignements le 6 mars, lors de la journรฉe CNC ยซ Crรฉer, produire, diffuser ร lโheure de lโintelligence artificielle ยป.
Le CNC publie la version augmentรฉe de la cartographie des usages de lโIA dans les filiรจres du cinรฉma, de lโaudiovisuel et du jeu vidรฉo.
Lien vers la cartographie complรจte: https://www.cnc.fr/documents/36995/2097582/Cartographie+des+usages+IA_rapport+complet.pdf/96532829-747e-b85e-c74b-af313072cab7?t=1712309387891
- 4 replies
reacted to
sayakpaul's
post with ๐ฅ
8 months ago
Post
4522
Flux.1-Dev like images but in fewer steps.
Merging code (very simple), inference code, merged params: sayakpaul/FLUX.1-merged
Enjoy the Monday ๐ค
Merging code (very simple), inference code, merged params: sayakpaul/FLUX.1-merged
Enjoy the Monday ๐ค
- 4 replies
reacted to
gokaygokay's
post with ๐๐ฅ
9 months ago
Post
4684
I've made a creative version of Tile Upscaler
- gokaygokay/TileUpscalerV2
- https://github.com/gokayfem/Tile-Upscaler
- New tiling strategy
- Now it's closer to Clarity Upscaler
- It has more parameters to play and it has more room to fail because of that
- You should try different resolutions, strength and controlnet strength
Original Tile Upscaler
- gokaygokay/Tile-Upscaler
- gokaygokay/TileUpscalerV2
- https://github.com/gokayfem/Tile-Upscaler
- New tiling strategy
- Now it's closer to Clarity Upscaler
- It has more parameters to play and it has more room to fail because of that
- You should try different resolutions, strength and controlnet strength
Original Tile Upscaler
- gokaygokay/Tile-Upscaler
reacted to
gokaygokay's
post with ๐
9 months ago
Post
4854
InSPyReNet Background Removal
I've built a space for fast background removal.
- gokaygokay/Inspyrenet-Rembg
- https://github.com/plemeri/InSPyReNet
I've built a space for fast background removal.
- gokaygokay/Inspyrenet-Rembg
- https://github.com/plemeri/InSPyReNet
- 2 replies
reacted to
gokaygokay's
post with ๐
9 months ago
Post
3021
I've created a Stable Diffusion 3 (SD3) image generation space for convenience. Now you can:
1. Generate SD3 prompts from images
2. Enhance your text prompts (turn 1-2 words into full SD3 prompts)
https://huggingface.co/spaces/gokaygokay/SD3-with-VLM-and-Prompt-Enhancer
These features are based on my custom models:
- VLM captioner for prompt generation:
- gokaygokay/sd3-long-captioner
- Prompt Enhancers for SD3 Models:
- gokaygokay/Lamini-Prompt-Enchance-Long
- gokaygokay/Lamini-Prompt-Enchance
You can now simplify your SD3 workflow with these tools!
1. Generate SD3 prompts from images
2. Enhance your text prompts (turn 1-2 words into full SD3 prompts)
https://huggingface.co/spaces/gokaygokay/SD3-with-VLM-and-Prompt-Enhancer
These features are based on my custom models:
- VLM captioner for prompt generation:
- gokaygokay/sd3-long-captioner
- Prompt Enhancers for SD3 Models:
- gokaygokay/Lamini-Prompt-Enchance-Long
- gokaygokay/Lamini-Prompt-Enchance
You can now simplify your SD3 workflow with these tools!
reacted to
gokaygokay's
post with ๐๐คฏ๐ฅ
9 months ago
Post
6202
Kolors with VLM support
I've built a space for using Kolors image generation model with captioner models and prompt enhancers.
- Space with VLM and Prompt Enhancer
gokaygokay/KolorsPlusPlus
- Original Space for model
gokaygokay/Kolors
- Captioner VLMs
- gokaygokay/sd3-long-captioner-v2
- microsoft/Florence-2-base
- Prompt Enhancers
- gokaygokay/Lamini-Prompt-Enchance-Long
- gokaygokay/Lamini-Prompt-Enchance
I've built a space for using Kolors image generation model with captioner models and prompt enhancers.
- Space with VLM and Prompt Enhancer
gokaygokay/KolorsPlusPlus
- Original Space for model
gokaygokay/Kolors
- Captioner VLMs
- gokaygokay/sd3-long-captioner-v2
- microsoft/Florence-2-base
- Prompt Enhancers
- gokaygokay/Lamini-Prompt-Enchance-Long
- gokaygokay/Lamini-Prompt-Enchance
reacted to
sanchit-gandhi's
post with ๐
about 1 year ago
Post
Why does returning timestamps help Whisper reduce hallucinations? ๐ง
Empirically, most practitioners have found that setting
But why does this work?..
My interpretation is that forcing the model to predict timestamps is contradictory to hallucinations. Suppose you have the transcription:
Where we have a repeated hallucination for โon theโ. If we ask the model to predict timestamps, then the โon theโ has to contribute to the overall segment-level timing, e.g.:
However, itโs impossible to fit 3 copies of โon theโ within the time allocation given to the segment, so the probability for this hallucinatory sequence becomes lower, and the model actually predicts the correct transcription with highest probability:
In this sense, the end timestamp is of the opposite of the initial timestamp constraint they describe in Section 4.5 of the paper Robust Speech Recognition via Large-Scale Weak Supervision (2212.04356) โ it helps the model remove extra words at the end of the sequence (rather than the initial timestamp which helps when the model ignores words at the start), but the overall principle is the same (using timestamps to improve the probability of more realistic sequences).
Leaving it open to you: why do you think timestamps reduces Whisper hallucinations?
Empirically, most practitioners have found that setting
return_timestamps=True helps reduce hallucinations, particularly when doing long-form evaluation with Transformersโ โchunkedโ algorithm. But why does this work?..
My interpretation is that forcing the model to predict timestamps is contradictory to hallucinations. Suppose you have the transcription:
The cat sat on the on the on the mat.Where we have a repeated hallucination for โon theโ. If we ask the model to predict timestamps, then the โon theโ has to contribute to the overall segment-level timing, e.g.:
<|0.00|> The cat sat on the on the on the mat.<|5.02|>However, itโs impossible to fit 3 copies of โon theโ within the time allocation given to the segment, so the probability for this hallucinatory sequence becomes lower, and the model actually predicts the correct transcription with highest probability:
<|0.00|> The cat sat on the mat.<|5.02|>In this sense, the end timestamp is of the opposite of the initial timestamp constraint they describe in Section 4.5 of the paper Robust Speech Recognition via Large-Scale Weak Supervision (2212.04356) โ it helps the model remove extra words at the end of the sequence (rather than the initial timestamp which helps when the model ignores words at the start), but the overall principle is the same (using timestamps to improve the probability of more realistic sequences).
Leaving it open to you: why do you think timestamps reduces Whisper hallucinations?
- 11 replies
Fix imports
1
#45 opened over 1 year ago
by
Wok