Olmo2 and olmoe support (#525)

* Support olmo2 and olmoe

* push

* push changes

* add note

configs/beaker_configs/ray_node_setup.sh

@@ -1,6 +1,7 @@
 export CURRENT_DATETIME=$(python -c "import datetime; import pytz; print(datetime.datetime.now(pytz.timezone('America/Los_Angeles')).strftime('%m%d%y_%H%M%S'))")
 export PYTHONPATH=$REPO_PATH
 export PATH="/root/.local/bin:$PATH"
+export NCCL_CUMEM_ENABLE=0
 
 
 echo CURRENT_DATETIME=$CURRENT_DATETIME

open_instruct/olmo_adapter/__init__.py

@@ -0,0 +1,226 @@
+from typing import List, Optional, Tuple, Union
+
+import torch
+from torch import nn
+from torch.nn import MSELoss, CrossEntropyLoss, BCEWithLogitsLoss
+from transformers.modeling_outputs import SequenceClassifierOutputWithPast
+from transformers.models.olmo2.modeling_olmo2 import Olmo2Config, Olmo2Model, Olmo2PreTrainedModel
+from transformers.models.olmoe.modeling_olmoe import OlmoeConfig, OlmoeModel, OlmoePreTrainedModel
+
+
+class Olmo2ForSequenceClassification(Olmo2PreTrainedModel):
+    def __init__(self, config: Olmo2Config):
+        super().__init__(config)
+        self.num_labels = config.num_labels
+        self.model = Olmo2Model(config)
+        self.score = nn.Linear(config.hidden_size, self.num_labels, bias=False)
+
+        # Initialize weights and apply final processing
+        self.post_init()
+
+    def get_input_embeddings(self):
+        return self.model.embed_tokens
+
+    def set_input_embeddings(self, value):
+        self.model.embed_tokens = value
+
+    def forward(
+        self,
+        input_ids: Optional[torch.LongTensor] = None,
+        attention_mask: Optional[torch.Tensor] = None,
+        position_ids: Optional[torch.LongTensor] = None,
+        past_key_values: Optional[Union[List[torch.FloatTensor]]] = None,
+        inputs_embeds: Optional[torch.FloatTensor] = None,
+        labels: Optional[torch.LongTensor] = None,
+        use_cache: Optional[bool] = None,
+        output_attentions: Optional[bool] = None,
+        output_hidden_states: Optional[bool] = None,
+        return_dict: Optional[bool] = None,
+    ) -> Union[Tuple, 'SequenceClassifierOutputWithPast']:
+        r"""
+        labels (`torch.LongTensor` of shape `(batch_size,)`, *optional*):
+            Labels for computing the sequence classification/regression loss. Indices should be in `[0, ...,
+            config.num_labels - 1]`. If `config.num_labels == 1` a regression loss is computed (Mean-Square loss), If
+            `config.num_labels > 1` a classification loss is computed (Cross-Entropy).
+        """
+        return_dict = return_dict if return_dict is not None else self.config.use_return_dict
+
+        transformer_outputs = self.model(
+            input_ids,
+            attention_mask=attention_mask,
+            position_ids=position_ids,
+            past_key_values=past_key_values,
+            inputs_embeds=inputs_embeds,
+            use_cache=use_cache,
+            output_attentions=output_attentions,
+            output_hidden_states=output_hidden_states,
+            return_dict=return_dict,
+        )
+        hidden_states = transformer_outputs[0]
+        logits = self.score(hidden_states)
+
+        if input_ids is not None:
+            batch_size = input_ids.shape[0]
+        else:
+            batch_size = inputs_embeds.shape[0]
+
+        if self.config.pad_token_id is None and batch_size != 1:
+            raise ValueError("Cannot handle batch sizes > 1 if no padding token is defined.")
+        if self.config.pad_token_id is None:
+            sequence_lengths = -1
+        else:
+            if input_ids is not None:
+                # if no pad token found, use modulo instead of reverse indexing for ONNX compatibility
+                sequence_lengths = torch.eq(input_ids, self.config.pad_token_id).int().argmax(-1) - 1
+                sequence_lengths = sequence_lengths % input_ids.shape[-1]
+                sequence_lengths = sequence_lengths.to(logits.device)
+            else:
+                sequence_lengths = -1
+
+        pooled_logits = logits[torch.arange(batch_size, device=logits.device), sequence_lengths]
+
+        loss = None
+        if labels is not None:
+            labels = labels.to(logits.device)
+            if self.config.problem_type is None:
+                if self.num_labels == 1:
+                    self.config.problem_type = "regression"
+                elif self.num_labels > 1 and (labels.dtype == torch.long or labels.dtype == torch.int):
+                    self.config.problem_type = "single_label_classification"
+                else:
+                    self.config.problem_type = "multi_label_classification"
+
+            if self.config.problem_type == "regression":
+                loss_fct = MSELoss()
+                if self.num_labels == 1:
+                    loss = loss_fct(pooled_logits.squeeze(), labels.squeeze())
+                else:
+                    loss = loss_fct(pooled_logits, labels)
+            elif self.config.problem_type == "single_label_classification":
+                loss_fct = CrossEntropyLoss()
+                loss = loss_fct(pooled_logits.view(-1, self.num_labels), labels.view(-1))
+            elif self.config.problem_type == "multi_label_classification":
+                loss_fct = BCEWithLogitsLoss()
+                loss = loss_fct(pooled_logits, labels)
+        if not return_dict:
+            output = (pooled_logits,) + transformer_outputs[1:]
+            return ((loss,) + output) if loss is not None else output
+
+        return SequenceClassifierOutputWithPast(
+            loss=loss,
+            logits=pooled_logits,
+            past_key_values=transformer_outputs.past_key_values,
+            hidden_states=transformer_outputs.hidden_states,
+            attentions=transformer_outputs.attentions,
+        )
+
+
+class OlmoeForSequenceClassification(OlmoePreTrainedModel):
+    def __init__(self, config: Olmo2Config):
+        super().__init__(config)
+        self.num_labels = config.num_labels
+        self.model = OlmoeModel(config)
+        self.score = nn.Linear(config.hidden_size, self.num_labels, bias=False)
+
+        # Initialize weights and apply final processing
+        self.post_init()
+
+    def get_input_embeddings(self):
+        return self.model.embed_tokens
+
+    def set_input_embeddings(self, value):
+        self.model.embed_tokens = value
+
+    def forward(
+        self,
+        input_ids: Optional[torch.LongTensor] = None,
+        attention_mask: Optional[torch.Tensor] = None,
+        position_ids: Optional[torch.LongTensor] = None,
+        past_key_values: Optional[Union[List[torch.FloatTensor]]] = None,
+        inputs_embeds: Optional[torch.FloatTensor] = None,
+        labels: Optional[torch.LongTensor] = None,
+        use_cache: Optional[bool] = None,
+        output_attentions: Optional[bool] = None,
+        output_hidden_states: Optional[bool] = None,
+        return_dict: Optional[bool] = None,
+        output_router_logits: Optional[bool] = None,
+        cache_position: Optional[torch.LongTensor] = None,
+    ) -> Union[Tuple, SequenceClassifierOutputWithPast]:
+        r"""
+        labels (`torch.LongTensor` of shape `(batch_size,)`, *optional*):
+            Labels for computing the sequence classification/regression loss. Indices should be in `[0, ...,
+            config.num_labels - 1]`. If `config.num_labels == 1` a regression loss is computed (Mean-Square loss), If
+            `config.num_labels > 1` a classification loss is computed (Cross-Entropy).
+        """
+        return_dict = return_dict if return_dict is not None else self.config.use_return_dict
+
+        transformer_outputs = self.model(
+            input_ids=input_ids,
+            attention_mask=attention_mask,
+            position_ids=position_ids,
+            past_key_values=past_key_values,
+            inputs_embeds=inputs_embeds,
+            use_cache=use_cache,
+            output_attentions=output_attentions,
+            output_hidden_states=output_hidden_states,
+            output_router_logits=output_router_logits,
+            return_dict=return_dict,
+            cache_position=cache_position,
+        )
+        hidden_states = transformer_outputs[0]
+        logits = self.score(hidden_states)
+
+        if input_ids is not None:
+            batch_size = input_ids.shape[0]
+        else:
+            batch_size = inputs_embeds.shape[0]
+
+        if self.config.pad_token_id is None and batch_size != 1:
+            raise ValueError("Cannot handle batch sizes > 1 if no padding token is defined.")
+        if self.config.pad_token_id is None:
+            sequence_lengths = -1
+        else:
+            if input_ids is not None:
+                # if no pad token found, use modulo instead of reverse indexing for ONNX compatibility
+                sequence_lengths = torch.eq(input_ids, self.config.pad_token_id).int().argmax(-1) - 1
+                sequence_lengths = sequence_lengths % input_ids.shape[-1]
+                sequence_lengths = sequence_lengths.to(logits.device)
+            else:
+                sequence_lengths = -1
+
+        pooled_logits = logits[torch.arange(batch_size, device=logits.device), sequence_lengths]
+
+        loss = None
+        if labels is not None:
+            labels = labels.to(logits.device)
+            if self.config.problem_type is None:
+                if self.num_labels == 1:
+                    self.config.problem_type = "regression"
+                elif self.num_labels > 1 and (labels.dtype == torch.long or labels.dtype == torch.int):
+                    self.config.problem_type = "single_label_classification"
+                else:
+                    self.config.problem_type = "multi_label_classification"
+
+            if self.config.problem_type == "regression":
+                loss_fct = MSELoss()
+                if self.num_labels == 1:
+                    loss = loss_fct(pooled_logits.squeeze(), labels.squeeze())
+                else:
+                    loss = loss_fct(pooled_logits, labels)
+            elif self.config.problem_type == "single_label_classification":
+                loss_fct = CrossEntropyLoss()
+                loss = loss_fct(pooled_logits.view(-1, self.num_labels), labels.view(-1))
+            elif self.config.problem_type == "multi_label_classification":
+                loss_fct = BCEWithLogitsLoss()
+                loss = loss_fct(pooled_logits, labels)
+        if not return_dict:
+            output = (pooled_logits,) + transformer_outputs[1:]
+            return ((loss,) + output) if loss is not None else output
+
+        return SequenceClassifierOutputWithPast(
+            loss=loss,
+            logits=pooled_logits,
+            past_key_values=transformer_outputs.past_key_values,
+            hidden_states=transformer_outputs.hidden_states,
+            attentions=transformer_outputs.attentions,
+        )

open_instruct/ppo_vllm_thread_ray_gtrl.py

@@ -634,6 +634,9 @@ def from_pretrained(
         self.model.train()
 
         # value model
+        from open_instruct.olmo_adapter import Olmo2Config, Olmo2ForSequenceClassification, OlmoeConfig, OlmoeForSequenceClassification
+        AutoModelForSequenceClassification.register(Olmo2Config, Olmo2ForSequenceClassification)
+        AutoModelForSequenceClassification.register(OlmoeConfig, OlmoeForSequenceClassification)
         self.value_model: PreTrainedModel = AutoModelForSequenceClassification.from_pretrained(
             args.reward_model_path,
             revision=args.reward_model_revision,
@@ -670,7 +673,9 @@ def from_pretrained(
         # reference model
         ds_config = get_eval_ds_config(
             offload=False,
-            stage=args.deepspeed_stage,
+            # inference model only has stage 3 (sharding) or stage 0 (no sharding)
+            # stage 2 is optimizer sharding which doesn't apply to inference
+            stage=args.deepspeed_stage if args.deepspeed_stage == 3 else 0,
             bf16=True,
         )
         ds_config["train_micro_batch_size_per_gpu"] = args.per_device_train_batch_size
@@ -709,7 +714,9 @@ def from_pretrained(
             disable_dropout_in_model(self.reward_model)
             ds_config = get_eval_ds_config(
                 offload=False,
-                stage=args.deepspeed_stage,
+                # inference model only has stage 3 (sharding) or stage 0 (no sharding)
+                # stage 2 is optimizer sharding which doesn't apply to inference
+                stage=args.deepspeed_stage if args.deepspeed_stage == 3 else 0,
                 bf16=True,
             )
             ds_config["train_micro_batch_size_per_gpu"] = args.per_device_train_batch_size

open_instruct/reward_modeling.py

@@ -195,6 +195,9 @@ def layer_init(layer: nn.Module, std: float):
 
 
 def main(args: Args, dataset_config: DatasetConfig, model_config: ModelConfig):
+    from open_instruct.olmo_adapter import Olmo2Config, Olmo2ForSequenceClassification, OlmoeConfig, OlmoeForSequenceClassification
+    AutoModelForSequenceClassification.register(Olmo2Config, Olmo2ForSequenceClassification)
+    AutoModelForSequenceClassification.register(OlmoeConfig, OlmoeForSequenceClassification)
     accelerator = calculate_runtime_args_and_accelerator(args, model_config)
     local_seed = args.seed + accelerator.process_index
 
@@ -241,7 +244,8 @@ def main(args: Args, dataset_config: DatasetConfig, model_config: ModelConfig):
         tokenizer.pad_token_id = 128002  # <|reserved_special_token_0|>
     else:
         tokenizer.add_special_tokens({"pad_token": "[PAD]"})  # NOTE: we do not resize the embedding
-    tokenizer.chat_template = CHAT_TEMPLATES[dataset_config.chat_template]
+    if dataset_config.chat_template is not None:
+        tokenizer.chat_template = CHAT_TEMPLATES[dataset_config.chat_template]
 
     # create the dataset
     dataset_dict = DatasetDict()