Universal Assisted Generation: Faster Decoding with Any Assistant Model
ā¢
52
Lewis Tunstall PRO
lewtun
orion-penner's profile picture
bird-of-paradise's profile picture
AmirhosseinSafari's profile picture
Ā·
AI & ML interests
LLMs, LLMs, LLMs
Recent Activity
published
a dataset
about 14 hours ago
HuggingFaceH4/aime_2024
updated
a dataset
about 14 hours ago
HuggingFaceH4/aime_2024
posted
an
update
1 day ago
We are reproducing the full DeepSeek R1 data and training pipeline so everybody can use their recipe. Instead of doing it in secret we can do it together in the open!
š§Ŗ Step 1: replicate the R1-Distill models by distilling a high-quality reasoning corpus from DeepSeek-R1.
š§ Step 2: replicate the pure RL pipeline that DeepSeek used to create R1-Zero. This will involve curating new, large-scale datasets for math, reasoning, and code.
š„ Step 3: show we can go from base model -> SFT -> RL via multi-stage training.
Follow along: https://github.com/huggingface/open-r1
Articles
Organizations
Posts
6
Post
2828
We are reproducing the full DeepSeek R1 data and training pipeline so everybody can use their recipe. Instead of doing it in secret we can do it together in the open!
š§Ŗ Step 1: replicate the R1-Distill models by distilling a high-quality reasoning corpus from DeepSeek-R1.
š§ Step 2: replicate the pure RL pipeline that DeepSeek used to create R1-Zero. This will involve curating new, large-scale datasets for math, reasoning, and code.
š„ Step 3: show we can go from base model -> SFT -> RL via multi-stage training.
Follow along: https://github.com/huggingface/open-r1
š§Ŗ Step 1: replicate the R1-Distill models by distilling a high-quality reasoning corpus from DeepSeek-R1.
š§ Step 2: replicate the pure RL pipeline that DeepSeek used to create R1-Zero. This will involve curating new, large-scale datasets for math, reasoning, and code.
š„ Step 3: show we can go from base model -> SFT -> RL via multi-stage training.
Follow along: https://github.com/huggingface/open-r1
Post
3417
I was initially pretty sceptical about Meta's Coconut paper [1] because the largest perf gains were reported on toy linguistic problems. However, these results on machine translation are pretty impressive!
https://x.com/casper_hansen_/status/1875872309996855343
Together with the recent PRIME method [2] for scaling RL, reasoning for open models is looking pretty exciting for 2025!
[1] Training Large Language Models to Reason in a Continuous Latent Space (2412.06769)
[2] https://huggingface.co./blog/ganqu/prime
https://x.com/casper_hansen_/status/1875872309996855343
Together with the recent PRIME method [2] for scaling RL, reasoning for open models is looking pretty exciting for 2025!
[1] Training Large Language Models to Reason in a Continuous Latent Space (2412.06769)
[2] https://huggingface.co./blog/ganqu/prime
Collections
4
A curated collection of datasets, models, Spaces, and papers on Reinforcement Learning from Human Feedback (RLHF).
A collection of Mistral 7B fine-tunes on UltraChat and Arithmo to boost the math capabilities of chat models. See https://x.com/_lewtun/status/1715652
-
lewtun/mistral-7b-sft-ultrachat-arithmo-full
Text Generation ā¢ Updated ā¢ 110 ā¢ 1 -
lewtun/mistral-7b-sft-ultrachat-arithmo-50
Text Generation ā¢ Updated ā¢ 13 ā¢ 1 -
lewtun/mistral-7b-sft-ultrachat-arithmo-25
Text Generation ā¢ Updated ā¢ 7 -
stingning/ultrachat
Viewer ā¢ Updated ā¢ 774k ā¢ 1.31k ā¢ 429
spaces
20
models
260
lewtun/Qwen2-0.5B-SFT
Updated
lewtun/Qwen2.5-0.5B-SFT-LoRA
Updated
lewtun/Llama-3.1-8B-SFT-LoRA-packing-no-lm-head
Updated
lewtun/Llama-3.1-8B-SFT-LoRA-no-packing
Updated
lewtun/Llama-3.1-8B-SFT-QLoRA-packing
Updated
lewtun/Llama-3.1-8B-SFT-LoRA-packing-no-saved-modules
Updated
lewtun/Llama-3.1-8B-SFT-LoRA-packing
Updated
lewtun/Llama-3.1-8B-SFT-LoRA-packing-pad-token-eos
Updated
lewtun/Llama-3.1-8B-SFT-QLoRA-packing-pad-token-eos
Updated
lewtun/Llama-3.1-8B-SFT-full-packing
Text Generation
ā¢
Updated
ā¢
6
datasets
61
lewtun/details_Qwen__Qwen2.5-Math-1.5B-Instruct
Viewer
ā¢
Updated
ā¢
11k
ā¢
31
lewtun/test-eval
Viewer
ā¢
Updated
ā¢
1
ā¢
14
lewtun/Llama-3.2-1B-Instruct-best_of_n-prm-completions
Viewer
ā¢
Updated
ā¢
10
ā¢
42
lewtun/math-bon-debug
Preview
ā¢
Updated
ā¢
520
lewtun/test-fast-parser-l1b-v3
Viewer
ā¢
Updated
ā¢
509
ā¢
45
lewtun/test-fast-parser-l1b-v2
Viewer
ā¢
Updated
ā¢
509
ā¢
62
lewtun/test-fast-parser-l1b
Viewer
ā¢
Updated
ā¢
509
ā¢
52
lewtun/test-fast-parser
Viewer
ā¢
Updated
ā¢
25
ā¢
54
lewtun/bon-prm-serverless-batched
Viewer
ā¢
Updated
ā¢
240
ā¢
69
lewtun/ultrafeedback_gkd
Viewer
ā¢
Updated
ā¢
62.1k
ā¢
45