FlexQwen Collection

Checkpoints, training logs, and other artifacts from training the FlexQwen models.

Checkout the FlexQwen repository for more details on the models and training process.

Instructions

repo_id = "mmcarpi/flexqwen-collection"
config_name = "gold-exponential/best" # dataset-strategy/checkpoint, eg "matched-shift/step_00500"

# Load tokenizer from the root
tokenizer = AutoTokenizer.from_pretrained(repo_id, trust_remote_code=True)

# Load model from the specific configuration subfolder
model = AutoModelForCausalLM.from_pretrained(
    repo_id,
    subfolder=config_name,
    trust_remote_code=True
)