You can deploy this model and start earning money today!

cicopaper/
cico-splitter

Name: cico-splitter
Availability: InStock

safetensors

qwen2

Views

No views yet

splitter-model

This directory contains the inference-ready Hugging Face model files copied from:

../checkpoint-2253

Files

File	Purpose
`model.safetensors`	Full model weights in safetensors format.
`config.json`	Model architecture config.
`generation_config.json`	Default generation parameters.
`tokenizer.json`	Fast tokenizer file.
`tokenizer_config.json`	Tokenizer configuration and added token metadata.
`special_tokens_map.json`	Special token mapping.
`added_tokens.json`	Added token definitions.
`vocab.json`	BPE vocabulary.
`merges.txt`	BPE merge rules.
`chat_template.jinja`	Chat prompt template.

Excluded Training State

The following checkpoint/training recovery files were intentionally left in ../checkpoint-2253 and are not required for normal inference:

global_step2251/
latest
trainer_state.json
training_args.bin
scheduler.pt
rng_state_*.pth
zero_to_fp32.py

Load Test

Verified with conda environment llamaf:

torch 2.7.0+cu126
transformers 4.52.1
safetensors 0.5.3
GPU: NVIDIA GeForce RTX 3090 Ti

The model loaded with:

python

from transformers import AutoModelForCausalLM, AutoTokenizer

tokenizer = AutoTokenizer.from_pretrained("../splitter-model", local_files_only=True)
model = AutoModelForCausalLM.from_pretrained(
    "../splitter-model",
    local_files_only=True,
    torch_dtype="bfloat16",
    device_map={"": 0},
)

Test generation succeeded.

cicopaper/cico-splitter

splitter-model

Files

Excluded Training State

Load Test

cicopaper/
cico-splitter