himalaya-ai/himalayagpt-0.5b

Exported from nanochat checkpoints with custom transformers remote code.

Load

from transformers import AutoTokenizer, AutoModelForCausalLM
import torch

repo = "himalaya-ai/himalayagpt-0.5b"
tok = AutoTokenizer.from_pretrained(repo, trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained(
    repo,
    trust_remote_code=True,
    torch_dtype=torch.bfloat16 if torch.cuda.is_available() else torch.float32,
    device_map="auto" if torch.cuda.is_available() else None,
)

prompt = "नेपालको राजधानी "
ids = tok(prompt, return_tensors="pt").input_ids.to(model.device)
out = model.generate(ids, max_new_tokens=64)
print(tok.decode(out[0], skip_special_tokens=True))

Notes

• This repo uses custom model/tokenizer code (trust_remote_code=True).
• Checkpoint source: base
• Model tag: d15_harl_fulltokens_sdpa_bs32
• Step: 133632