End of training

README.md

@@ -12,33 +12,37 @@ should probably proofread and complete it, then remove this comment. -->
 
 # ProGemma2
 
+This model is a fine-tuned version of [JuIm/ProGemma2](https://huggingface.co/JuIm/ProGemma2) on an unknown dataset.
 
-### This is an updated/improved version of ProGemma. Please see below for how to use it. Compared to ProGemma, ProGemma 2 has better perplexity scores, HHblits e-values, and ptm/pLDDT/iptm scores. 
+## Model description
 
+More information needed
 
-This is a custom configuration of Google’s Gemma 2 LLM that is being pre-trained on amino acid sequences of 512 AA or less in length. Periodic updates are made to this page as training reaches new checkpoints.
+## Intended uses & limitations
 
-The purpose of this model was to investigate the differences between ProGemma and ProtGPT (GPT-2 architecture) as it pertains to sequence generation. Perplexity scores as well as AlphaFold 3’s ptm, pLDDT, and iptm scores are generally in line with ProtGPT’s scores for sequence lengths < 250, although the testing phase is still very early. I have yet to do testing for sequence lengths > 250. More robust testing is also required for lengths < 250 AA. In my very preliminary testing, HHblit e-values of ~0.1 are achieved with relatively easily and more reliably than in ProGemma..
+More information needed
 
-Controlled generation is not a capability of this model, and therefore serves as a method to significantly improve generation as, in principal, a sequence that performs a given function or resides in a particular cellular location can be generated.
+## Training and evaluation data
 
-In sequence generation, a top_k of 950 appears to work well as it prevents repetition. This is also seen in ProtGPT.
+More information needed
 
-Below is code using the Transformers library to generate sequences using ProGemma.
+## Training procedure
 
-from transformers import pipeline, AutoTokenizer, AutoModelForCausalLM
+### Training hyperparameters
 
-model = AutoModelForCausalLM.from_pretrained("JuIm/ProGemma")
+The following hyperparameters were used during training:
+- learning_rate: 0.001
+- train_batch_size: 2
+- eval_batch_size: 8
+- seed: 42
+- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
+- lr_scheduler_type: linear
+- lr_scheduler_warmup_ratio: 0.4
+- training_steps: 3500
 
-tokenizer = AutoTokenizer.from_pretrained("JuIm/Amino-Acid-Sequence-Tokenizer")
+### Training results
 
-progemma = pipeline("text-generation", model=model, tokenizer=tokenizer)
 
-sequence = progemma("", top_k=950, max_length=100, num_return_sequences=1, do_sample=True, repetition_penalty=1.2, eos_token_id=21, pad_token_id=22, bos_token_id=20)
-
-s = sequence[0]['generated_text']
-
-print(s)
 
 ### Framework versions
 

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:c4eb03746651466f8b653da95f68ebe6119f42ed3e74bcc6f5c0b47f6211e8e3
+oid sha256:22829a48b1928018c6245a93ebdbf4a154c7bca5f61bac63558dea9f60c7af69
 size 1342562152

runs/Sep04_13-10-49_9dae2c38c2b7/events.out.tfevents.1725455452.9dae2c38c2b7.248.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4035ffca1cf291f99e8cc2af96dfadc0bf07043d08b87070d8f532e7eaa58ca3
+size 743330

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:67b3bc3452cb294e27535b8d1234a89c4979b2ce0b6b6dc17712f7c28838e6e9
+oid sha256:3c389a5b1d7102245920997f1489d25ffc1a70b8007fee6a1dd75c28cfeee808
 size 5112