Hello I am trying to finetune a Mixtral8x7B Instruct v0.1 model using 4 bit quantization. The finetuning completes without any issue, but when I try to reload the finetuned model, it does not load it throwing the following errors :
size mismatch for model.embed_tokens.weight: copying a param with shape torch.Size([32007, 4096]) from checkpoint, the shape in current model is torch.Size([32000, 4096]).
size mismatch for lm_head.weight: copying a param with shape torch.Size([32007, 4096]) from checkpoint, the shape in current model is torch.Size([32000, 4096]).
From the error it seems as though the finetuning process added 7 new tokens without it being specified to?
Here is the tokenizer_config.json
{
"add_bos_token": true,
"add_eos_token": false,
"added_tokens_decoder": {
"0": {
"content": "<unk>",
"lstrip": false,
"normalized": false,
"rstrip": false,
"single_word": false,
"special": true
},
"1": {
"content": "<s>",
"lstrip": false,
"normalized": false,
"rstrip": false,
"single_word": false,
"special": true
},
"2": {
"content": "</s>",
"lstrip": false,
"normalized": false,
"rstrip": false,
"single_word": false,
"special": true
}
},
"additional_special_tokens": [],
"bos_token": "<s>",
"chat_template": "{{ bos_token }}{% for message in messages %}{% if (message['role'] == 'user') != (loop.index0 % 2 == 0) %}{{ raise_exception('Conversation roles must alternate user/assistant/user/assistant/...') }}{% endif %}{% if message['role'] == 'user' %}{{ '[INST] ' + message['content'] + ' [/INST]' }}{% elif message['role'] == 'assistant' %}{{ message['content'] + eos_token}}{% else %}{{ raise_exception('Only user and assistant roles are supported!') }}{% endif %}{% endfor %}",
"clean_up_tokenization_spaces": false,
"eos_token": "</s>",
"legacy": true,
"model_max_length": 1000000000000000019884624838656,
"pad_token": "</s>",
"sp_model_kwargs": {},
"spaces_between_special_tokens": false,
"tokenizer_class": "LlamaTokenizer",
"unk_token": "<unk>",
"use_default_system_prompt": false
}
I am able to load the vanilla model and use it for inference.