Fine tuned Mistral-7B-Instruct-1.0 inference missing config.json

I have used spaces → docker → autotrain to finetune mistralai/Mistral-7B-Instruct-v0.1 with a csv file containing a text column and formatting that follow the following params:

[INST]Instruction[/INST]Expected reply

After training is successful I use spaces → docker → ChatUI to launch a UI for inference of my finetuned model.

The space keeps erroring out because I don’t have a config.json file.

Here’s a link to my model: ramon1992/Mistral-7B-JB-Instruct-3-0

Is there someone who could use my model and create a ChatUI space with it just to see if it works?

When running the below script:

from transformers import AutoModelForCausalLM, AutoTokenizer

tokenizer = AutoTokenizer.from_pretrained(“ramon1992/Mistral-7B-JB-Instruct-3-0”)

model = AutoModelForCausalLM.from_pretrained(“ramon1992/Mistral-7B-JB-Instruct-3-0”)

input_ids = tokenizer.encode(“Your input text here”, return_tensors=‘pt’)

output = model.generate(input_ids, max_length=50)

print(tokenizer.decode(output[0], skip_special_tokens=True))

I get the following error:

ramon1992/Mistral-7B-JB-Instruct-3-0 does not appear to have a file named pytorch_model.bin, tf_model.h5, model.ckpt or flax_model.msgpack.

SOLUTION: Missing config.json file after training using AutoTrain · Issue #299 · huggingface/autotrain-advanced · GitHub