bro if you had read the repo you would see that this implementation is for educational purpose, it's not done because it's easy. Not to mention unsloth is using trl's GRPO trainer which is super slow on cpu and does not scale for models under 500M params, I tried it both on cpu and gpu. This custom implementation cuts most of the heavy lifting allowing you to train and scale faster even on cpu, plus a bunch of custom configs with a simplified GRPO trainer in under 500 lines of code. There's a lot one can learn from it.
Jaward Sesay
Jaward
AI & ML interests
I like to train large deep neural nets too 🧠🤖💥 | First Paper (AutoAgents: A Framework for Automatic Agent Generation) Accepted @ IJCAI 2024 | Role Model Karpathy
Recent Activity
replied to
their
post
5 days ago
Finally here it is: a faster, custom, scalable GRPO trainer for smaller models with < 500M params, can train on 8gb ram cpu, also supports gpu for sanity sake (includes support for vllm + flash attention). Using smolLM2-135M/360M-instructs as ref & base models. Experience your own “aha” moment 🐳 on 8gb ram.
Code: https://github.com/Jaykef/ai-algorithms/blob/main/smollm2_360M_135M_grpo_gsm8k.ipynb
posted
an
update
7 days ago
Finally here it is: a faster, custom, scalable GRPO trainer for smaller models with < 500M params, can train on 8gb ram cpu, also supports gpu for sanity sake (includes support for vllm + flash attention). Using smolLM2-135M/360M-instructs as ref & base models. Experience your own “aha” moment 🐳 on 8gb ram.
Code: https://github.com/Jaykef/ai-algorithms/blob/main/smollm2_360M_135M_grpo_gsm8k.ipynb
liked
a model
12 days ago
HuggingFaceTB/SmolVLM-Instruct
Organizations
Jaward's activity
replied to
their
post
5 days ago
posted
an
update
7 days ago
Post
3785
Finally here it is: a faster, custom, scalable GRPO trainer for smaller models with < 500M params, can train on 8gb ram cpu, also supports gpu for sanity sake (includes support for vllm + flash attention). Using smolLM2-135M/360M-instructs as ref & base models. Experience your own “aha” moment 🐳 on 8gb ram.
Code: https://github.com/Jaykef/ai-algorithms/blob/main/smollm2_360M_135M_grpo_gsm8k.ipynb
Code: https://github.com/Jaykef/ai-algorithms/blob/main/smollm2_360M_135M_grpo_gsm8k.ipynb
-
- 2 replies
posted
an
update
19 days ago
Post
3381
ByteDance drops OmniHuman🔥
This is peak SOTA performance - flawless natural gestures with perfect lip sync and facial expressions. This is the second time they've released SOTA level talking-heads only this time with hands and body motion.
Project: https://omnihuman-lab.github.io/
This is peak SOTA performance - flawless natural gestures with perfect lip sync and facial expressions. This is the second time they've released SOTA level talking-heads only this time with hands and body motion.
Project: https://omnihuman-lab.github.io/
- 3 replies
posted
an
update
23 days ago
Post
1481
The beauty in GRPO is the fact that it doesn’t care if the rewards are rule-based or learned, the hack: let the data self-normalize— trajectories in a batch compete against their mean, no value model, no extra params, just clean, efficient RL that cuts memory usage by 50%, while maintaining SOTA performance. btw it was introduced 9months prior to R1: arxiv.org/pdf/2402.03300
- 1 reply
reacted to
mlabonne's
post with 🧠
about 1 month ago
Post
5009
🆕 LLM Course 2025 edition!
I updated the LLM Scientist roadmap and added a ton of new information and references. It covers training, datasets, evaluation, quantization, and new trends like test-time compute scaling.
The LLM Course has been incredibly popular (41.3k stars!) and I've been touched to receive many, many messages about how it helped people in their careers.
I know how difficult this stuff can be, so I'm super proud of the impact it had. I want to keep updating it in 2025, especially with the LLM Engineer roadmap.
Thanks everyone, hope you'll enjoy it!
💻 LLM Course: https://huggingface.co/blog/mlabonne/llm-course
I updated the LLM Scientist roadmap and added a ton of new information and references. It covers training, datasets, evaluation, quantization, and new trends like test-time compute scaling.
The LLM Course has been incredibly popular (41.3k stars!) and I've been touched to receive many, many messages about how it helped people in their careers.
I know how difficult this stuff can be, so I'm super proud of the impact it had. I want to keep updating it in 2025, especially with the LLM Engineer roadmap.
Thanks everyone, hope you'll enjoy it!
💻 LLM Course: https://huggingface.co/blog/mlabonne/llm-course
posted
an
update
about 1 month ago
Post
1874
minimal single script implementation of knowledge distillation in LLMs. In this implementation, we use GPT-2 (124M) as student model and GPT-2 Medium (340M) as teacher via reverse Kullback-Leibler (KL) divergence, trained on a small chunk of openwebtext.
Code: https://github.com/Jaykef/ai-algorithms/blob/main/llm_knowledge_distillation.ipynb
Code: https://github.com/Jaykef/ai-algorithms/blob/main/llm_knowledge_distillation.ipynb
posted
an
update
about 2 months ago
Post
1370
Huge AI win in medicine👏
"Large language of life model" just dropped!!
Full paper: https://www.nature.com/articles/s41586-024-08391-z
"Large language of life model" just dropped!!
Full paper: https://www.nature.com/articles/s41586-024-08391-z
- 1 reply
posted
an
update
about 2 months ago
Post
2324
damn I love nvidia's bullish stance on taking AI to the edge - from being the overlord of compute to cutting-edge physical AI with SOTA multiverse simulation engines that brings the scaling laws under your control!!
My favorite: Cosmos - fully opensourced, open-weight physics based video gen platform, what an incredible way to start off the year✨
Code: https://github.com/NVIDIA/Cosmos
Models: nvidia/cosmos-6751e884dc10e013a0a0d8e6
Paper: https://d1qx31qr3h6wln.cloudfront.net/publications/NVIDIA%20Cosmos_2.pdf
My favorite: Cosmos - fully opensourced, open-weight physics based video gen platform, what an incredible way to start off the year✨
Code: https://github.com/NVIDIA/Cosmos
Models: nvidia/cosmos-6751e884dc10e013a0a0d8e6
Paper: https://d1qx31qr3h6wln.cloudfront.net/publications/NVIDIA%20Cosmos_2.pdf
posted
an
update
about 2 months ago
Post
3067
nanoBLT: Simplified lightweight implementation of a character-level Byte Latent Transformer model (under 500 lines of code). The model is 2x4x2 (n_layers_encoder, n_layers_latent, n_layers_decoder) layer deep trained on ~1M bytes of tiny Shakespeare with a patch size of 4.
Code: https://github.com/Jaykef/ai-algorithms/blob/main/byte_latent_transformer.ipynb
Code: https://github.com/Jaykef/ai-algorithms/blob/main/byte_latent_transformer.ipynb
replied to
their
post
2 months ago
btw the background songs in the videos are actually what I listen to during implementation
posted
an
update
2 months ago
Post
1818
Implements from first-principle a discrete flow matching model for code generation- trained a small sized 2D dfm model on two variations of code for binary search. The result was amazing, code in comment:
Code: https://github.com/Jaykef/ai-algorithms/blob/main/dfm.ipynb
Code: https://github.com/Jaykef/ai-algorithms/blob/main/dfm.ipynb
-
- 1 reply
posted
an
update
3 months ago
Post
607
In Honour of This Year's NeurIPs Test of Time Paper Awardees
This year's NIPs Test of Time Paper Awards went to two groundbreaking papers:
1. Generative Adversarial Nets (Goodfellow et al)
2. Sequence to Sequence Learning with Neural Networks (Ilya et al)
Let's explore how these papers helped pioneered breakthroughs in today's AI:
Full Article: https://huggingface.co/blog/Jaward/nip
This year's NIPs Test of Time Paper Awards went to two groundbreaking papers:
1. Generative Adversarial Nets (Goodfellow et al)
2. Sequence to Sequence Learning with Neural Networks (Ilya et al)
Let's explore how these papers helped pioneered breakthroughs in today's AI:
Full Article: https://huggingface.co/blog/Jaward/nip
posted
an
update
3 months ago
Post
648
Lightweight implementation of the seminal paper “Sequence to Sequence Learning with Neural Networks”
Built, trained and eval a 2 layer deep seq2seq LSTM-based model (~10M params) on German-English corpus of Multi30K dataset. In honor of
ilya sutskever et al for winning this year’s NeurIPSConf Test of Time paper award 🫡
Code: https://github.com/Jaykef/ai-algorithms/blob/main/seq2seq.ipynb
Built, trained and eval a 2 layer deep seq2seq LSTM-based model (~10M params) on German-English corpus of Multi30K dataset. In honor of
ilya sutskever et al for winning this year’s NeurIPSConf Test of Time paper award 🫡
Code: https://github.com/Jaykef/ai-algorithms/blob/main/seq2seq.ipynb
posted
an
update
3 months ago
Post
490
Rethinking Backpropagation: Thoughts on What's Wrong with Backpropagation
As a young researcher, I've often pondered the limitations of backpropagation, especially when mapped with how learning occurs in the human brain. While backpropagation has been the workhorse of deep learning, it isn't without flaws. In this post, I aim to share some thoughts on these shortcomings from first principles.
Full article
https://huggingface.co/blog/Jaward/rethinking-backpropagation
As a young researcher, I've often pondered the limitations of backpropagation, especially when mapped with how learning occurs in the human brain. While backpropagation has been the workhorse of deep learning, it isn't without flaws. In this post, I aim to share some thoughts on these shortcomings from first principles.
Full article
https://huggingface.co/blog/Jaward/rethinking-backpropagation
posted
an
update
3 months ago
Post
2440
Implements compute-efficient DeepPCR algorithm which parallelizes sequential operations thus speeding up inference and training of neural networks. DeepPCR can significantly reduce the time complexity in operations such as denoising in latent diffusion space from O(L) to O(log2 L).
Code: https://github.com/Jaykef/ai-algorithms/blob/main/deep_pcr.ipynb
Code: https://github.com/Jaykef/ai-algorithms/blob/main/deep_pcr.ipynb
posted
an
update
3 months ago
Post
1243
This is supercool!!
Explores o1-like multimodal reasoning.
Multi-agents with DPO is a nice touch 👍
Paper: https://arxiv.org/pdf/2411.14432
Code: https://github.com/dongyh20/Insight-V
Explores o1-like multimodal reasoning.
Multi-agents with DPO is a nice touch 👍
Paper: https://arxiv.org/pdf/2411.14432
Code: https://github.com/dongyh20/Insight-V
posted
an
update
3 months ago
Post
1630
Ok RNNs can rap too:)
Here we implement the seminal RNN paper “Generating Text with Recurrent Neural Networks"- we train a character-level multiplicative recurrent neural network model (~250k params) for 1000 epochs with Adam opt on 2pac's "Hit 'em Up", sample was fun lol.
Code: https://github.com/Jaykef/ai-algorithms/blob/main/generating_texts_with_rnns.ipynb
Here we implement the seminal RNN paper “Generating Text with Recurrent Neural Networks"- we train a character-level multiplicative recurrent neural network model (~250k params) for 1000 epochs with Adam opt on 2pac's "Hit 'em Up", sample was fun lol.
Code: https://github.com/Jaykef/ai-algorithms/blob/main/generating_texts_with_rnns.ipynb
posted
an
update
3 months ago
Post
1743
Interesting Work on Reasoning 🤔
- explores a new take on few-shot reasoning while challenging assumptions that program synthesis is necessary for abstract reasoning.
- shows test-time training + smart inference tricks can match human-average performance, though at high computational cost. Key insight: proper compute allocation matters more than method (whether symbolic or neural).
Paper: https://ekinakyurek.github.io/papers/ttt.pdf
- explores a new take on few-shot reasoning while challenging assumptions that program synthesis is necessary for abstract reasoning.
- shows test-time training + smart inference tricks can match human-average performance, though at high computational cost. Key insight: proper compute allocation matters more than method (whether symbolic or neural).
Paper: https://ekinakyurek.github.io/papers/ttt.pdf
posted
an
update
4 months ago
Post
2115
It's work like this that in some way signal the eventual “dominance” of AI over all the sciences.
“We train our model on the six-dimensional N-body phase space, predicting particle velocities as the time derivative of the model’s displacement outputs”
The emulator is capable of predicting
the nonlinear displacement and velocity fields for 128^3 particles in half a second on a single GPU🤯
“We train our model on the six-dimensional N-body phase space, predicting particle velocities as the time derivative of the model’s displacement outputs”
The emulator is capable of predicting
the nonlinear displacement and velocity fields for 128^3 particles in half a second on a single GPU🤯
- 1 reply
posted
an
update
4 months ago
Post
1759
Triton nanoGPT now has a custom cross entropy loss kernel 🚀
Next: matmul, gradually overthrowing all major PyTorch ops:)
Simplified pseudo for parallel cross-entropy loss compute:
- init program: get pid, compute offsets, load targets.
- init row_max and row_sum.
- for-loop1 (find max logits): update row_max with max logits.
- for-loop2 (compute softmax and loss): compute row_sum, update loss.
- add log(row_sum) and store loss.
Code: https://github.com/Jaykef/ai-algorithms/blob/main/triton_nanoGPT.ipynb
Next: matmul, gradually overthrowing all major PyTorch ops:)
Simplified pseudo for parallel cross-entropy loss compute:
- init program: get pid, compute offsets, load targets.
- init row_max and row_sum.
- for-loop1 (find max logits): update row_max with max logits.
- for-loop2 (compute softmax and loss): compute row_sum, update loss.
- add log(row_sum) and store loss.
Code: https://github.com/Jaykef/ai-algorithms/blob/main/triton_nanoGPT.ipynb