Trying other alternative Models.

[SauravBhowmick]

Trying: microsoft/CodeGPT-small-py, Salesforce/codegen-350M-mono for now

[ekrishnachaitanya2004]

Regarding the RAM issues on Google Colab: this is a common limitation, especially on the free tier where RAM is usually capped at ~12–13GB. These models, even though they’re relatively small, can still exceed that limit during inference.

If Supported - Load models in float16 precision - This can significantly reduce memory usage.

model = AutoModelForCausalLM.from_pretrained(
    model_id, 
    device_map="auto", 
    torch_dtype=torch.float16
)