ProTrain: Efficient LLM Training via Memory-Aware Techniques
Paper ā¢ 2406.08334 ā¢ Published
AI & ML interests
None defined yet.
Recent Activity
View all activity
Xiaoqun2023Ā
authored a
paper 17 days ago
clefourrierĀ
authored a
paper about 1 month ago
alvarobarttĀ
posted an update about 1 month ago
Post
3599
Learn how to deploy Microsoft Research VibeVoice ASR on Microsoft Azure Foundry with Hugging Face to generate rich audio transcriptions with Who, When, and What! š„
> š 60-minute single-pass processing, no chunking or stitching
> š¤ Customized hotwords to guide recognition on domain-specific content
> š Rich transcription: joint ASR + diarization + timestamping in one pass
> š 50+ languages with automatic detection and code-switching support
> š¤ Deployed on Microsoft Foundry via an OpenAI-compatible Chat Completions API
https://huggingface.co/docs/microsoft-azure/foundry/examples/deploy-vibevoice-asr
> š 60-minute single-pass processing, no chunking or stitching
> š¤ Customized hotwords to guide recognition on domain-specific content
> š Rich transcription: joint ASR + diarization + timestamping in one pass
> š 50+ languages with automatic detection and code-switching support
> š¤ Deployed on Microsoft Foundry via an OpenAI-compatible Chat Completions API
https://huggingface.co/docs/microsoft-azure/foundry/examples/deploy-vibevoice-asr
alvarobarttĀ
posted an update 3 months ago
Post
3221
š„
š” Alternatively, you can also set the
hf-mem v0.4.1 now also estimates KV cache memory requirements for any context length and batch size with the --experimental flag!uvx hf-mem --model-id ... --experimental will automatically pull the required information from the Hugging Face Hub to include the KV cache estimation, when applicable.š” Alternatively, you can also set the
--max-model-len, --batch-size and --kv-cache-dtype arguments (Ć la vLLM) manually if preferred. - 1 reply
ankits0052Ā
authored a
paper 3 months ago
eliebakĀ
submitted a
paper to Daily Papers 4 months ago
Post
11454
deepseek-ai/DeepSeek-OCR is out! š„ my take ā¤µļø
> pretty insane it can parse and re-render charts in HTML
> it uses CLIP and SAM features concatenated, so better grounding
> very efficient per vision tokens/performance ratio
> covers 100 languages
> pretty insane it can parse and re-render charts in HTML
> it uses CLIP and SAM features concatenated, so better grounding
> very efficient per vision tokens/performance ratio
> covers 100 languages
- 4 replies
jfchiĀ
authored 6
papers 6 months ago
FFB: A Fair Fairness Benchmark for In-Processing Group Fairness Methods
Paper ā¢ 2306.09468 ā¢ Published ā¢ 1
The Llama 3 Herd of Models
Paper ā¢ 2407.21783 ā¢ Published ā¢ 118
Persistent Pre-Training Poisoning of LLMs
Paper ā¢ 2410.13722 ā¢ Published
Towards Understanding the Fragility of Multilingual LLMs against Fine-Tuning Attacks
Paper ā¢ 2410.18210 ā¢ Published
Large Reasoning Models Learn Better Alignment from Flawed Thinking
Paper ā¢ 2510.00938 ā¢ Published ā¢ 60
Shape it Up! Restoring LLM Safety during Finetuning
Paper ā¢ 2505.17196 ā¢ Published ā¢ 1
Post
6986
large AI labs open-sourced a ton of models last week š„
here's few picks, find even more here merve/sep-16-releases-68d13ea4c547f02f95842f05 š¤
> IBM released a new Docling model with 258M params based on Granite (A2.0) š ibm-granite/granite-docling-258M
> Xiaomi released 7B audio LM with base and instruct variants (MIT) XiaomiMiMo/mimo-audio-68cc7202692c27dae881cce0
> DecartAI released Lucy Edit, open Nano Banana š (NC) decart-ai/Lucy-Edit-Dev
> OpenGVLab released a family of agentic computer use models (3B/7B/32B) with the dataset š» OpenGVLab/scalecua-68c912cf56f7ff4c8e034003
> Meituan Longcat released thinking version of LongCat-Flash š meituan-longcat/LongCat-Flash-Thinking
here's few picks, find even more here merve/sep-16-releases-68d13ea4c547f02f95842f05 š¤
> IBM released a new Docling model with 258M params based on Granite (A2.0) š ibm-granite/granite-docling-258M
> Xiaomi released 7B audio LM with base and instruct variants (MIT) XiaomiMiMo/mimo-audio-68cc7202692c27dae881cce0
> DecartAI released Lucy Edit, open Nano Banana š (NC) decart-ai/Lucy-Edit-Dev
> OpenGVLab released a family of agentic computer use models (3B/7B/32B) with the dataset š» OpenGVLab/scalecua-68c912cf56f7ff4c8e034003
> Meituan Longcat released thinking version of LongCat-Flash š meituan-longcat/LongCat-Flash-Thinking
- 2 replies
Post
3521
IBM just released small swiss army knife for the document models: granite-docling-258M on Hugging Face š„
> not only a document converter but also can do document question answering, understand multiple languages š¤Æ
> best part: released with Apache 2.0 license š use it with your commercial projects!
> it supports transformers, vLLM and MLX from the get-go! š¤
> built on SigLIP2 & granite-165M
model: ibm-granite/granite-docling-258M
demo: ibm-granite/granite-docling-258m-demo š
> not only a document converter but also can do document question answering, understand multiple languages š¤Æ
> best part: released with Apache 2.0 license š use it with your commercial projects!
> it supports transformers, vLLM and MLX from the get-go! š¤
> built on SigLIP2 & granite-165M
model: ibm-granite/granite-docling-258M
demo: ibm-granite/granite-docling-258m-demo š
Post
1277
a ton of image/video generation models and LLMs from big labs š„
> Meta released facebook/mobilellm-r1-68c4597b104fac45f28f448e, smol LLMs for on-device use š¬
> Tencent released tencent/SRPO, high res image generation model and tencent/POINTS-Reader, cutting edge OCR š
> ByteDance released bytedance-research/HuMo, video generation from any input āÆļø
find more models, datasets, demos here merve/sep-11-releases-68c7dbfa26bea8cd921fa0ac
> Meta released facebook/mobilellm-r1-68c4597b104fac45f28f448e, smol LLMs for on-device use š¬
> Tencent released tencent/SRPO, high res image generation model and tencent/POINTS-Reader, cutting edge OCR š
> ByteDance released bytedance-research/HuMo, video generation from any input āÆļø
find more models, datasets, demos here merve/sep-11-releases-68c7dbfa26bea8cd921fa0ac
Post
1070
fan-favorite vision LM Florence-2 is now officially supported in transformers š¤
find all the models in
florence-community org š«”
find all the models in
Post
1870
past week was great for open LLMs š„ merve/sep-1-releases-68bede0e729c12597eefd050
> Google released google/embeddinggemma-300m, new embedding model with 300M params
> new update to Kimi-K2 just landed moonshotai/Kimi-K2-Instruct-0905 š
> OpenBMB released a new version to MiniCPM with 8B params openbmb/MiniCPM4.1-8B
also soooo many Qwen-Image & Kontext LoRAs dropped!
> Google released google/embeddinggemma-300m, new embedding model with 300M params
> new update to Kimi-K2 just landed moonshotai/Kimi-K2-Instruct-0905 š
> OpenBMB released a new version to MiniCPM with 8B params openbmb/MiniCPM4.1-8B
also soooo many Qwen-Image & Kontext LoRAs dropped!
Post
3756
upgrade your transformers š„
it comes with insanely capable models like merve/sam2-66ac9deac6fca3bc5482fe30, microsoft/kosmos-2.5, and more š«”
I built a notebook you can run with free Colab T4 to walk through the API for new models šš»āāļø merve/smol-vision
fine-tuning will follow-up soon!
it comes with insanely capable models like merve/sam2-66ac9deac6fca3bc5482fe30, microsoft/kosmos-2.5, and more š«”
I built a notebook you can run with free Colab T4 to walk through the API for new models šš»āāļø merve/smol-vision
fine-tuning will follow-up soon!