mradermacher

/

model_requests

like 196

It's so beautiful 💀

It's so beautiful 💀

is not
is stupid

Train it on 1T tokens and see what happens lol. Simonko or nico will queue later, Im busy for now so cant queue. We will queue on usual quants we provide, so you might get iq1 too lol

Train it on 1T tokens and see what happens lol. Simonko or nico will queue later, Im busy for now so cant queue. We will queue on usual quants we provide, so you might get iq1 too lol

im training it on a pi. how do i train it on 1T tokens

I had been summoned

plz let me know when u get to my model

I would also like another of my models to be quantized (with the same if not more quants):
https://huggingface.co/NILKNARFGonzo/floppyx3-MEGAmodel-merged
It's around 1.09M parameters.

why did it do that

nvm i think i know why

Train it on 1T tokens and see what happens lol. Simonko or nico will queue later, Im busy for now so cant queue. We will queue on usual quants we provide, so you might get iq1 too lol

@RichardErkhov
question: when will you get to it? (not to be rude but i'm kinda impatient)

I'm curious why you need so many quants of a model that is... not good? Not that I don't think it's great to have extra stuff, just genuinely curious, like are you using it for something that you specifically need ggufs for as a test or what :D

i dont actually need all of them, but I’m really curious how stupid an IQ1_S could be lol

It's queued!
You can check for progress at http://hf.tst.eu/status.html or regularly check the model
summary page at https://hf.tst.eu/model#single-floppy-346k-merged-GGUF
https://hf.tst.eu/model#floppyx3-MEGAmodel-merged-GGUF for quants to appear.

tysm!!!!!

what's the progress so far?

im training it on a pi. how do i train it on 1T tokens

optimize to oblivion and just wait, one day it will work... one day... but this day is not today

also custom arches/tokenizers are not supported, stupid llamacpp limitation

single-floppy-346k-merged       WARNING:hf-to-gguf:
single-floppy-346k-merged
single-floppy-346k-merged       WARNING:hf-to-gguf:**************************************************************************************
single-floppy-346k-merged       WARNING:hf-to-gguf:** WARNING: The BPE pre-tokenizer was not recognized!
single-floppy-346k-merged       WARNING:hf-to-gguf:**          There are 2 possible reasons for this:
single-floppy-346k-merged       WARNING:hf-to-gguf:**          - the model has not been added to convert_hf_to_gguf_update.py yet
single-floppy-346k-merged       WARNING:hf-to-gguf:**          - the pre-tokenization config has changed upstream
single-floppy-346k-merged       WARNING:hf-to-gguf:**          Check your model files and convert_hf_to_gguf_update.py and update them accordingly.
single-floppy-346k-merged       WARNING:hf-to-gguf:** ref:     https://github.com/ggml-org/llama.cpp/pull/6920
single-floppy-346k-merged       WARNING:hf-to-gguf:**
single-floppy-346k-merged       WARNING:hf-to-gguf:** chkhsh:  93671e068c61f5edea078be9b11952c00f4c56b1cd8f294f535d911b654c58ed
single-floppy-346k-merged       WARNING:hf-to-gguf:**************************************************************************************
single-floppy-346k-merged       WARNING:hf-to-gguf:
single-floppy-346k-merged
single-floppy-346k-merged       Traceback (most recent call last):
single-floppy-346k-merged         File "/llmjob/llama.cpp/convert_hf_to_gguf.py", line 308, in <module>
single-floppy-346k-merged           main()
single-floppy-346k-merged         File "/llmjob/llama.cpp/convert_hf_to_gguf.py", line 302, in main
single-floppy-346k-merged           model_instance.write()
single-floppy-346k-merged         File "/llmjob/llama.cpp-nocuda/conversion/base.py", line 1036, in write
single-floppy-346k-merged           self.prepare_metadata(vocab_only=False)
single-floppy-346k-merged         File "/llmjob/llama.cpp-nocuda/conversion/base.py", line 1203, in prepare_metadata
single-floppy-346k-merged           self.set_vocab()
single-floppy-346k-merged         File "/llmjob/llama.cpp-nocuda/conversion/base.py", line 1175, in set_vocab
single-floppy-346k-merged           self._set_vocab_gpt2()
single-floppy-346k-merged         File "/llmjob/llama.cpp-nocuda/conversion/base.py", line 1724, in _set_vocab_gpt2
single-floppy-346k-merged           tokens, toktypes, tokpre = self.get_vocab_base()
single-floppy-346k-merged         File "/llmjob/llama.cpp-nocuda/conversion/base.py", line 1361, in get_vocab_base
single-floppy-346k-merged           tokpre = self.get_vocab_base_pre(tokenizer)
single-floppy-346k-merged         File "/llmjob/llama.cpp-nocuda/conversion/base.py", line 1712, in get_vocab_base_pre
single-floppy-346k-merged           raise NotImplementedError("BPE pre-tokenizer was not recognized - update get_vocab_base_pre()")
single-floppy-346k-merged       NotImplementedError: BPE pre-tokenizer was not recognized - update get_vocab_base_pre()