SWE-RL: Advancing LLM Reasoning via Reinforcement Learning on Open Software Evolution
Paper
•
2502.18449
•
Published
•
70
Meta Llama
Enterprise
company
Verified
AI & ML interests
None defined yet.
Recent Activity
meta-llama's activity
Adding notice for developers on importance of organization name
1
#97 opened 12 days ago
by
betodepaola
My Llama repo request denied, although I am a Meta engineer
3
#17 opened 17 days ago
by
adilevin-meta
Fixing typo on docs/overview link
1
#31 opened 20 days ago
by
betodepaola
Fixing typo on docs/overview link
1
#31 opened 20 days ago
by
betodepaola
JadeCopet
authored
a
paper
27 days ago
yorambac
authored
a
paper
about 1 month ago
Post
5832
SmolVLM-2 and SigLIP-2 are now part of
They're added on top of the v4.49.0 release, and can be installed from the following tags:
This marks a new beginning for the release process of transformers. For the past five years, we've been doing monthly releases featuring many models (v4.49.0, the latest release, features 9 new architectures).
Starting with SmolVLM-2 & SigLIP2, we'll now additionally release tags supporting new models on a stable branch. These models are therefore directly available for use by installing from the tag itself. These tags will continue to be updated with fixes applied to these models.
Going forward, continue expecting software releases following semantic versioning: v4.50.0 will have ~10 new architectures compared to v4.49.0, as well as a myriad of new features, improvements and bug fixes. Accompanying these software releases, we'll release tags offering brand new models as fast as possible, to make them accessible to all immediately.
transformers in dedicated releases!They're added on top of the v4.49.0 release, and can be installed from the following tags:
v4.49.0-SmolVLM-2 and v4.49.0-SigLIP-2.This marks a new beginning for the release process of transformers. For the past five years, we've been doing monthly releases featuring many models (v4.49.0, the latest release, features 9 new architectures).
Starting with SmolVLM-2 & SigLIP2, we'll now additionally release tags supporting new models on a stable branch. These models are therefore directly available for use by installing from the tag itself. These tags will continue to be updated with fixes applied to these models.
Going forward, continue expecting software releases following semantic versioning: v4.50.0 will have ~10 new architectures compared to v4.49.0, as well as a myriad of new features, improvements and bug fixes. Accompanying these software releases, we'll release tags offering brand new models as fast as possible, to make them accessible to all immediately.
- 1 reply
reach-vb
authored
a
paper
about 2 months ago
Chat template problem.
13
#23 opened 6 months ago
by
mylesgoose
Chat template problem.
13
#23 opened 6 months ago
by
mylesgoose
ValueError: `rope_scaling` must be a dictionary with two fields
52
#15 opened 8 months ago
by
jsemrau
Unexpected Output using Official Example Code
7
#71 opened 5 months ago
by
jenniferL
Model card updates
6
#28 opened 3 months ago
by
pcuenq
Base Model?
3
#32 opened 3 months ago
by
User8213
Is EU use permitted with THIS model?
2
#35 opened 3 months ago
by
ptrdvn
Update to latest version of the Llama 3.3 70B Instruct model
1
#71 opened 3 months ago
by
reach-vb
Add new model version to the model card
#42 opened 3 months ago
by
reach-vb
Post
1456
Performance leap: TGI v3 is out. Processes 3x more tokens, 13x faster than vLLM on long prompts. Zero config !
3x more tokens.
By reducing our memory footprint, we’re able to ingest many more tokens and more dynamically than before. A single L4 (24GB) can handle 30k tokens on llama 3.1-8B, while vLLM gets barely 10k. A lot of work went into reducing the footprint of the runtime and its effect are best seen on smaller constrained environments.
13x faster
On long prompts (200k+ tokens) conversation replies take 27.5s in vLLM, while it takes only 2s in TGI. How so ? We keep the initial conversation around, so when a new reply comes in, we can answer almost instantly. The overhead of the lookup is ~5us. Thanks @Dani ël de Kok for the beast data structure.
Zero config
That’s it. Remove all the flags your are using and you’re likely to get the best performance. By evaluating the hardware and model, TGI carefully selects automatic values to give best performance. In production, we don’t have any flags anymore in our deployments. We kept all existing flags around, they may come in handy in niche scenarios.
Read more: https://huggingface.co/docs/text-generation-inference/conceptual/chunking
3x more tokens.
By reducing our memory footprint, we’re able to ingest many more tokens and more dynamically than before. A single L4 (24GB) can handle 30k tokens on llama 3.1-8B, while vLLM gets barely 10k. A lot of work went into reducing the footprint of the runtime and its effect are best seen on smaller constrained environments.
13x faster
On long prompts (200k+ tokens) conversation replies take 27.5s in vLLM, while it takes only 2s in TGI. How so ? We keep the initial conversation around, so when a new reply comes in, we can answer almost instantly. The overhead of the lookup is ~5us. Thanks @Dani ël de Kok for the beast data structure.
Zero config
That’s it. Remove all the flags your are using and you’re likely to get the best performance. By evaluating the hardware and model, TGI carefully selects automatic values to give best performance. In production, we don’t have any flags anymore in our deployments. We kept all existing flags around, they may come in handy in niche scenarios.
Read more: https://huggingface.co/docs/text-generation-inference/conceptual/chunking
pcuenq
updated
a
model
3 months ago
Model card updates
6
#28 opened 3 months ago
by
pcuenq