add 1B HF model variant and patch contexts for new paths

MaxToki-1B-HF/config.json

@@ -0,0 +1,35 @@
+{
+  "architectures": [
+    "LlamaForCausalLM"
+  ],
+  "attention_bias": false,
+  "attention_dropout": 0.0,
+  "bos_token_id": 2,
+  "eos_token_id": 3,
+  "head_dim": 144,
+  "hidden_act": "silu",
+  "hidden_size": 2304,
+  "initializer_range": 0.02,
+  "intermediate_size": 4608,
+  "max_position_embeddings": 4096,
+  "mlp_bias": false,
+  "model_type": "llama",
+  "num_attention_heads": 16,
+  "num_hidden_layers": 20,
+  "num_key_value_heads": 8,
+  "pretraining_tp": 1,
+  "rms_norm_eps": 1e-05,
+  "rope_scaling": {
+    "factor": 1.0,
+    "high_freq_factor": 4.0,
+    "low_freq_factor": 1.0,
+    "original_max_position_embeddings": 4096,
+    "rope_type": "llama3"
+  },
+  "rope_theta": 500000,
+  "tie_word_embeddings": false,
+  "torch_dtype": "float32",
+  "transformers_version": "4.53.3",
+  "use_cache": true,
+  "vocab_size": 20275
+}

MaxToki-1B-HF/generation_config.json

@@ -0,0 +1,6 @@
+{
+  "_from_model_config": true,
+  "bos_token_id": 2,
+  "eos_token_id": 3,
+  "transformers_version": "4.53.3"
+}

MaxToki-1B-HF/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:96c6c8f9d61732c9bedfc37a9cfd937359e6ce0bfa860526646d101815a242e6
+size 4196167128

MaxToki-1B-bionemo/context/model.yaml

@@ -1,7 +1,7 @@
 _target_: bionemo.llm.lightning.BionemoLightningModule
 config:
   _cpu_offloading_context: null
-  _target_: bionemo.geneformer.biollama.finetune_example.BioLlamaConfig
+  _target_: bionemo.maxtoki.model.MaxTokiConfig
   account_for_embedding_in_pipeline_split: false
   account_for_loss_in_pipeline_split: false
   activation_func:
@@ -255,7 +255,7 @@ data_step:
   _target_: nemo.collections.llm.gpt.model.base.gpt_data_step
 forward_step:
   _call_: false
-  _target_: bionemo.geneformer.biollama.finetune_example.cellformer_forward_step
+  _target_: bionemo.maxtoki.model.maxtoki_forward_step
 freeze_params_until_key_suffix: null
 model_context_managers: []
 model_transform: null
@@ -329,7 +329,7 @@ predict_loss_reduction_class:
   _target_: bionemo.llm.lightning.PassthroughLossReduction
 predict_step: null
 tokenizer:
-  _target_: bionemo.geneformer.biollama.hf_trainer.CellformerTokenizer
+  _target_: bionemo.maxtoki.tokenizer.MaxTokiTokenizer
   pad_token: <pad>
   token_dictionary:
     <boq>: 23275

MaxToki-217M-bionemo/context/model.yaml

@@ -1,7 +1,7 @@
 _target_: bionemo.llm.lightning.BionemoLightningModule
 config:
   _cpu_offloading_context: null
-  _target_: bionemo.geneformer.biollama.finetune_example.BioLlamaConfig
+  _target_: bionemo.maxtoki.model.MaxTokiConfig
   account_for_embedding_in_pipeline_split: false
   account_for_loss_in_pipeline_split: false
   activation_func:
@@ -255,7 +255,7 @@ data_step:
   _target_: nemo.collections.llm.gpt.model.base.gpt_data_step
 forward_step:
   _call_: false
-  _target_: bionemo.geneformer.biollama.finetune_example.cellformer_forward_step
+  _target_: bionemo.maxtoki.model.maxtoki_forward_step
 freeze_params_until_key_suffix: null
 model_context_managers: []
 model_transform: null
@@ -321,7 +321,7 @@ predict_loss_reduction_class:
   _target_: bionemo.llm.lightning.PassthroughLossReduction
 predict_step: null
 tokenizer:
-  _target_: bionemo.geneformer.biollama.hf_trainer.CellformerTokenizer
+  _target_: bionemo.maxtoki.tokenizer.MaxTokiTokenizer
   pad_token: <pad>
   token_dictionary:
     '-1': 21774