Downloads · 30 days
11
19% of all-time downloads
toksuite-all-merged
moe-dtoks/toksuite-all-merged
toksuite-all-merged is a machine learning model from moe-dtoks. Use it for the machine learning task on the model card, and read the license before you ship it in a product. The card lists the license as mit.
This model is a result of parameter averaging (Model Soup) across 14 models.
All-time downloads
59
Public
Parameters
1.8B
4.7 GB on disk
Likes
0
Public
Repo makeup
Click a slice to open those files.
.safetensors4.7 GB · 100%
Parameter types
How the weights are stored.
BF161.3B · 71%
Base models
- toksuite/supertoken_models-llama_google-byt5-small
- toksuite/supertoken_models-llama_facebook-xglm-564M
- toksuite/supertoken_models-llama_bigscience-bloom
- toksuite/supertoken_models-llama_gpt2
- toksuite/supertoken_models-llama_common-pile-comma-v0.1
- toksuite/supertoken_models-llama_google-gemma-2-2b
- toksuite/supertoken_models-llama_microsoft-Phi-3-mini-4k-instruct
- toksuite/supertoken_models-llama_meta-llama-Llama-3.2-1B
- toksuite/supertoken_models-llama_CohereLabs-aya-expanse-8b
- toksuite/supertoken_models-llama_tiktoken-gpt-4o
- toksuite/supertoken_models-llama_google-bert-bert-base-multilingual-cased
- toksuite/supertoken_models-llama_Qwen-Qwen3-8B
- toksuite/supertoken_models-llama_tokenmonster-englishcode-32000-consistent-v1
- toksuite/supertoken_models-llama_mistralai-tekken
- toksuite/CohereLabs-aya-expanse-8b
- toksuite/Qwen-Qwen3-8B
- toksuite/bigscience-bloom
- toksuite/common-pile-comma-v0.1
- toksuite/facebook-xglm-564M
- toksuite/google-bert-bert-base-multilingual-cased
- toksuite/google-byt5-small
- toksuite/google-gemma-2-2b
- toksuite/gpt2
- toksuite/meta-llama-Llama-3.2-1B
- toksuite/microsoft-Phi-3-mini-4k-instruct
- toksuite/mistralai-tekken
- toksuite/tiktoken-gpt-4o
- toksuite/tokenmonster-englishcode-32000-consistent-v1
Model card
From the Hugging Face model README
Merged Model: toksuite-all-merged
This model is a result of parameter averaging (Model Soup) across 14 models.
Merged Models
The following models were included in the merge:
- toksuite/supertoken_models-llama_google-byt5-small
- toksuite/supertoken_models-llama_facebook-xglm-564M
- toksuite/supertoken_models-llama_bigscience-bloom
- toksuite/supertoken_models-llama_gpt2
- toksuite/supertoken_models-llama_common-pile-comma-v0.1
- toksuite/supertoken_models-llama_google-gemma-2-2b
- toksuite/supertoken_models-llama_microsoft-Phi-3-mini-4k-instruct
- toksuite/supertoken_models-llama_meta-llama-Llama-3.2-1B
- toksuite/supertoken_models-llama_CohereLabs-aya-expanse-8b
- toksuite/supertoken_models-llama_tiktoken-gpt-4o
- toksuite/supertoken_models-llama_google-bert-bert-base-multilingual-cased
- toksuite/supertoken_models-llama_Qwen-Qwen3-8B
- toksuite/supertoken_models-llama_tokenmonster-englishcode-32000-consistent-v1
- toksuite/supertoken_models-llama_mistralai-tekken
Merging Configuration
- Method: Weighted Parameter Averaging
- Weights: Simple average with merging lambda = 1.0.
- Excluded Layers: Embeddings and LM Head were kept from the base model (toksuite/supertoken_models-llama_meta-llama-Llama-3.2-1B).
Usage
from transformers import AutoModelForCausalLM, AutoTokenizer
model = AutoModelForCausalLM.from_pretrained("moe-dtoks/toksuite-all-merged")
tokenizer = AutoTokenizer.from_pretrained("moe-dtoks/toksuite-all-merged")
- Type
- llama
- License
- mit
- Created
- Jan 18, 2026
- Updated
- Jan 18, 2026