making converter additionally support deepseek-coder dense hf model #167
Conversation
add deepseek1b support
add two directions converter (hf->olmo, olmo->hf)
clean imports
replace hf to olmo conveter
| # HF_MODEL = "meta-llama/Llama-3.2-1B" | ||
| # HF_MODEL = "meta-llama/Llama-3.2-8B" | ||
| HF_MODEL = f"{os.environ['SHARE_RES_DIR']}/models/deepseek/deepseek-coder-1.3b-base" | ||
| # HF_MODEL = "/home/zliu/shared_resources/models/llama3/hf/Llama-3.2-1B" |
There was a problem hiding this comment.
Please don't put your private paths into the code base.
| # TOKENIZER_CONFIG = TokenizerConfig.from_hf(HF_MODEL) | ||
| TOKENIZER_CONFIG = TokenizerConfig.from_hf("deepseek-ai/deepseek-coder-1.3b-base") | ||
| # TOKENIZER_CONFIG = TokenizerConfig.from_hf("meta-llama/Llama-3.2-1B") |
There was a problem hiding this comment.
Same here. This is one-off stuff that breaks everybody else's workflow. Imagine someone wants to use this tool that has zero context about this part of the work, and they just want to convert some checkpoint.
| MODEL_CONFIG: TransformerConfig | ||
| if HF_MODEL == "meta-llama/Llama-3.2-1B": | ||
| if "Llama-3.2-1B" in HF_MODEL: |
There was a problem hiding this comment.
Do we also need this for bigger Llama models?
There was a problem hiding this comment.
I change it so that it works with local cache dir. We are encountering some additional weird bug related to conversion; and requires some help from you. We will submit a separate issue regarding that. Do you prefer we incorporate solution to that issue into this PR or should I separate that from this current PR?
There was a problem hiding this comment.
Many small PRs is always better, if possible.
# Conflicts: # src/examples/huggingface/convert_checkpoint_to_hf.py
|
@lingchensanwen , I merged current main into this. There were a fair number of conflicts, and I'm not sure I did it right. Can you check? |
Changes made:
@dirkgr