This model is part of the 📐
FineMath ablations, we continue pretraining
Llama-3.2-3B base on different math datasets for 60B tokens.
The model has 3.21B parameters and 4096 context length. It was trained on
60B tokens from InfiMM-WebMath-3+ subset of 📐
FineMath, tokenized using
llama3 tokenizer.
This model was trained on English math data and is not instruction-tuned, making it intended for text completion in English with a focus on math.
It is important to note that the primary intended use case of this model is to compare its performance with other models trained under the same conditions. This model is not necessarily the best possible outcome achievable with the given dataset.
1# pip install -q transformers
2from transformers import AutoModelForCausalLM, AutoTokenizer
3
4model = MODEL_HERE
5device = "cuda" # for GPU usage or "cpu" for CPU usage
6
7tokenizer = AutoTokenizer.from_pretrained(model)
8model = AutoModelForCausalLM.from_pretrained(model).to(device)
9
10inputs = tokenizer.encode("Machine Learning is", return_tensors="pt").to(device)
11outputs = model.generate(inputs)
12print(tokenizer.decode(outputs[0]))
We are releasing intermediate checkpoints for this model at intervals of every 10000 training steps (10B tokens) in separate branches. The naming convention is 10B.
1from huggingface_hub import list_repo_refs
2out = list_repo_refs(MODEL_HERE)
3print([b.name for b in out.branches])
We used the SmolLM2 setup to evaluate all our ablation models with
lighteval. You can find the details here:
https://github.com/huggingface/smollm/tree/main/evaluation#smollm2-base-modelsThis model was predominantly trained on English math data, potentially limiting its performance in other languages. Furthermore, the model's behavior is influenced by the quality and diversity of its training data, which may include biases and harmful content.
