GeoV-9B is a 9 billion parameter causal language model.
The GeoV model was designed by Georges Harik and uses Rotary Positional Embeddings with Relative distances (RoPER) by Georges Harik and Varuna Jayasiri.
RoPER, in addition to using relative positions in the attention score calculation by RoPE embeddings, adds relative positional information explicitly to value embeddings. Specifically, it incorporates the relative positions of the tokens paid attention to. RoPER has given better performance in some algorithmic tasks, and seems comparable to RoPE in language modeling.
Model details
- Developed by: Georges Harik
- Model type: Transformer-based Language Model
- Language: English
| Hyperparameter | Value |
|---|---|
| nparameters | 9B |
| nlayers | 32 |
| dmodel | 5120 |
| nheads | 40 |
| dhead | 128 |
| nvocab | 65500 |
| Sequence Length | 2048 |
The released weights were trained on ~70 billion tokens. We plan to continue training up to 300 billion tokens and update the weights at every 20b tokens. This training run is monolingual and uses c4en and english wikipedia datasets.
Test results
These are the results from EleutherAI/lm-evaluation-harness at 80B (tokens trained) checkpoint.
| Task | Version | Metric | Value | Stderr | |
|---|---|---|---|---|---|
| anli_r1 | 0 | acc | 0.3150 | ± | 0.0147 |
| anli_r2 | 0 | acc | 0.3380 | ± | 0.0150 |
| anli_r3 | 0 | acc | 0.3367 | ± | 0.0136 |
| hellaswag | 0 | acc | 0.4761 | ± | 0.0050 |
| acc_norm | 0.6308 | ± | 0.0048 | ||
| lambada_openai | 0 | ppl | 8.9700 | ± | 0.2606 |
| acc | 0.5628 | ± | 0.0069 | ||
| mathqa | 0 | acc | 0.2318 | ± | 0.0077 |
| acc_norm | 0.2372 | ± | 0.0078 | ||
| piqa | 0 | acc | 0.7448 | ± | 0.0102 |
| acc_norm | 0.7639 | ± | 0.0099 | ||
| winogrande | 0 | acc | 0.5935 | ± | 0.0138 |
| wsc | 0 | acc | 0.4038 | ± | 0.0483 |
Installation
pip install geov
Generation
from geov import GeoVForCausalLM, GeoVTokenizer
model = GeoVForCausalLM.from_pretrained("GeoV/GeoV-9b")
tokenizer = GeoVTokenizer.from_pretrained("GeoV/GeoV-9b")
prompt = "In mathematics, topology is the study of"
input_ids = tokenizer(prompt, return_tensors="pt").input_ids
gen_tokens = model.generate(
input_ids,
do_sample=True,
temperature=0.9,
max_length=100,
)
gen_text = tokenizer.batch_decode(gen_tokens)[0]