github
/
InternLM
mirror of https://github.com/InternLM/InternLM
1
0
Fork 0
Code Issues Projects Releases Wiki Activity

fix AutoModel

pull/23/head
x54-729 2023-07-10 17:43:19 +08:00
parent 3ab5c5294d
commit 59d7a1d58d
3 changed files with 11 additions and 6 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

5
tools/transformers/README-zh-Hans.md
View File

@ -17,9 +17,8 @@ python tools/transformers/convert2hf.py --src_folder origin_ckpt/ --tgt_folder h
然后可以使用 `from_pretrained` 接口加载： 然后可以使用 `from_pretrained` 接口加载：
```python ```python
from modeling_internlm import InternLMForCausalLM >>> from transformers import AutoTokenizer, AutoModel
>>> model = AutoModel.from_pretrained("hf_ckpt/", trust_remote_code=True).cuda()
model = InternForCausalLM.from_pretrained("hf_ckpt/")
``` ```

5
tools/transformers/README.md
View File

@ -16,9 +16,8 @@ python tools/transformers/convert2hf.py --src_folder origin_ckpt/ --tgt_folder h
Then, you can load it using the `from_pretrained` interface: Then, you can load it using the `from_pretrained` interface:
```python ```python
from modeling_internlm import InternLMForCausalLM >>> from transformers import AutoTokenizer, AutoModel
>>> model = AutoModel.from_pretrained("hf_ckpt/", trust_remote_code=True).cuda()
model = InternForCausalLM.from_pretrained("hf_ckpt/")
``` ```
`intern_moss_example.py` demonstrates an example of how to use LoRA for fine-tuning on the `fnlp/moss-moon-002-sft` dataset. `intern_moss_example.py` demonstrates an example of how to use LoRA for fine-tuning on the `fnlp/moss-moon-002-sft` dataset.

7
tools/transformers/convert2hf.py
View File

@ -1,5 +1,6 @@
import argparse import argparse
import math import math
import json
import os import os
import re import re
import tempfile import tempfile
@ -163,6 +164,12 @@ if __name__ == "__main__":
os.makedirs(target_folder, exist_ok=True) os.makedirs(target_folder, exist_ok=True)
model.save_pretrained(target_folder, max_shard_size="20GB") model.save_pretrained(target_folder, max_shard_size="20GB")
# TODO There should be a better way to add this.
with open(os.path.join(target_folder, "config.json")) as fp:
config_dict = json.load(fp)
config_dict["auto_map"]["AutoModel"] = "modeling_internlm.InternLMModel"
with open(os.path.join(target_folder, "config.json"), "w") as fp:
json.dump(config_dict, fp, indent=2)
tokenizer = InternLMTokenizer(args.tokenizer) tokenizer = InternLMTokenizer(args.tokenizer)
tokenizer.save_pretrained(target_folder) tokenizer.save_pretrained(target_folder)