deqing / llama-isolate-old

huggingface.co
Total runs: 9
24-hour runs: 0
7-day runs: 3
30-day runs: -1
Model's Last Updated: March 30 2026
text-generation

Introduction of llama-isolate-old

Model Details of llama-isolate-old

convergent-llama-300M-muon-isolate

A 300M-parameter language model trained from scratch on FineWeb-Edu 10BT (~9.4B tokens, 1 epoch) as part of the Convergent Evolution project, which investigates how Fourier features emerge in LLM number embeddings.

Model details
Architecture LLaMA-style Transformer (12 layers, 1024 hidden, 16 heads, GQA)
Parameters ~300M
Optimizer Muon (for 2D weights) + AdamW (for embeddings/bias/norm)
Data perturbation block-diagonal attention mask (numbers cannot attend to context)
Training data FineWeb-Edu sample-10BT (~9.4B tokens)
Context length 1024
Tokenizer Llama 3 (128K vocab)
Batch size 512 sequences
Training dynamics

Intermediate checkpoints are saved as branches: tokens-200M , tokens-400M , ..., tokens-9.6B .

from transformers import AutoModelForCausalLM

# Load final checkpoint
model = AutoModelForCausalLM.from_pretrained("deqing/convergent-llama-300M-muon-isolate")

# Load intermediate checkpoint (e.g., at 1B tokens)
model = AutoModelForCausalLM.from_pretrained("deqing/convergent-llama-300M-muon-isolate", revision="tokens-1B")
Citation

Paper forthcoming.

Runs of deqing llama-isolate-old on huggingface.co

9
Total runs
0
24-hour runs
3
3-day runs
3
7-day runs
-1
30-day runs

More Information About llama-isolate-old huggingface.co Model

More llama-isolate-old license Visit here:

https://choosealicense.com/licenses/mit

llama-isolate-old huggingface.co

llama-isolate-old huggingface.co is an AI model on huggingface.co that provides llama-isolate-old's model effect (), which can be used instantly with this deqing llama-isolate-old model. huggingface.co supports a free trial of the llama-isolate-old model, and also provides paid use of the llama-isolate-old. Support call llama-isolate-old model through api, including Node.js, Python, http.

llama-isolate-old huggingface.co Url

https://huggingface.co/deqing/llama-isolate-old

deqing llama-isolate-old online free

llama-isolate-old huggingface.co is an online trial and call api platform, which integrates llama-isolate-old's modeling effects, including api services, and provides a free online trial of llama-isolate-old, you can try llama-isolate-old online for free by clicking the link below.

deqing llama-isolate-old online free url in huggingface.co:

https://huggingface.co/deqing/llama-isolate-old

llama-isolate-old install

llama-isolate-old is an open source model from GitHub that offers a free installation service, and any user can find llama-isolate-old on GitHub to install. At the same time, huggingface.co provides the effect of llama-isolate-old install, users can directly use llama-isolate-old installed effect in huggingface.co for debugging and trial. It also supports api for free installation.

llama-isolate-old install url in huggingface.co:

https://huggingface.co/deqing/llama-isolate-old

Url of llama-isolate-old

llama-isolate-old huggingface.co Url

Provider of llama-isolate-old huggingface.co

deqing
ORGANIZATIONS

Other API from deqing