About running a model
Hi, I just try to run your steps and try to do a forcast. I've git clone 'https://github.com/shiyu-coder/Kronos.git' and hopefully get your model.
Then after I install all requirements, I do as README says, and get across such an error：
[TypeError:KronosTokenizer.__init__() missing 16 required positional arguments:'d_in','d_model',''](TypeError: KronosTokenizer.__init__() missing 16 required positional arguments: 'd_in', 'd_model', 'n_heads', 'ff_dim', 'n_enc_layers', 'n_dec_layers', 'ffn_dropout_p', 'attn_dropout_p', 'resid_dropout_p', 's1_bits', 's2_bits', 'beta', 'gamma0', 'gamma', 'zeta', and 'group_size')
I tried to plug in config.json, but python tell me that 'learn_te' is an unexpectd argument.
what's wrong with it?
Anyone can help me?
Hello, this issue is most likely caused by an incorrect local path for the model. You can explicitly specify the cache_dir as the local path of the model. For example:
KronosTokenizer.from_pretrained("NeoQuasar/Kronos-Tokenizer-base", cache_dir="the path to the model on your server")
Then ensure that there are model parameters in the cache_dir path
Doeesn't work. May be try again this weekend.(cry)