deqing / convergent-llama-300M-adamw-isolate

huggingface.co
Total runs: 9
24-hour runs: 0
7-day runs: -3
30-day runs: -2
Model's Last Updated: March 30 2026
text-generation

Introduction of convergent-llama-300M-adamw-isolate

Model Details of convergent-llama-300M-adamw-isolate

convergent-llama-300M-adamw-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 AdamW
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-adamw-isolate")

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

Paper forthcoming.

Runs of deqing convergent-llama-300M-adamw-isolate on huggingface.co

9
Total runs
0
24-hour runs
-2
3-day runs
-3
7-day runs
-2
30-day runs

More Information About convergent-llama-300M-adamw-isolate huggingface.co Model

More convergent-llama-300M-adamw-isolate license Visit here:

https://choosealicense.com/licenses/mit

convergent-llama-300M-adamw-isolate huggingface.co

convergent-llama-300M-adamw-isolate huggingface.co is an AI model on huggingface.co that provides convergent-llama-300M-adamw-isolate's model effect (), which can be used instantly with this deqing convergent-llama-300M-adamw-isolate model. huggingface.co supports a free trial of the convergent-llama-300M-adamw-isolate model, and also provides paid use of the convergent-llama-300M-adamw-isolate. Support call convergent-llama-300M-adamw-isolate model through api, including Node.js, Python, http.

convergent-llama-300M-adamw-isolate huggingface.co Url

https://huggingface.co/deqing/convergent-llama-300M-adamw-isolate

deqing convergent-llama-300M-adamw-isolate online free

convergent-llama-300M-adamw-isolate huggingface.co is an online trial and call api platform, which integrates convergent-llama-300M-adamw-isolate's modeling effects, including api services, and provides a free online trial of convergent-llama-300M-adamw-isolate, you can try convergent-llama-300M-adamw-isolate online for free by clicking the link below.

deqing convergent-llama-300M-adamw-isolate online free url in huggingface.co:

https://huggingface.co/deqing/convergent-llama-300M-adamw-isolate

convergent-llama-300M-adamw-isolate install

convergent-llama-300M-adamw-isolate is an open source model from GitHub that offers a free installation service, and any user can find convergent-llama-300M-adamw-isolate on GitHub to install. At the same time, huggingface.co provides the effect of convergent-llama-300M-adamw-isolate install, users can directly use convergent-llama-300M-adamw-isolate installed effect in huggingface.co for debugging and trial. It also supports api for free installation.

convergent-llama-300M-adamw-isolate install url in huggingface.co:

https://huggingface.co/deqing/convergent-llama-300M-adamw-isolate

Url of convergent-llama-300M-adamw-isolate

convergent-llama-300M-adamw-isolate huggingface.co Url

Provider of convergent-llama-300M-adamw-isolate huggingface.co

deqing
ORGANIZATIONS

Other API from deqing