DeepXR
/

Helion-OSC

+"""
+Helion-OSC Model Architecture
+PyTorch implementation of the Helion-OSC transformer with MoE
+"""
+import math
+from typing import Optional, Tuple, List
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+from transformers import PreTrainedModel, PretrainedConfig
+from transformers.modeling_outputs import CausalLMOutputWithPast
+class HelionOSCConfig(PretrainedConfig):
+    """Configuration class for Helion-OSC model"""
+    model_type = "helion-osc"
+    def __init__(
+        self,
+        vocab_size=128256,
+        hidden_size=8192,
+        intermediate_size=28672,
+        num_hidden_layers=80,
+        num_attention_heads=64,
+        num_key_value_heads=8,
+        hidden_act="swiglu",
+        max_position_embeddings=262144,
+        initializer_range=0.02,
+        rms_norm_eps=1e-5,
+        use_cache=True,
+        pad_token_id=0,
+        bos_token_id=1,
+        eos_token_id=2,
+        tie_word_embeddings=False,
+        rope_theta=10000000.0,
+        rope_scaling=None,
+        attention_bias=False,
+        attention_dropout=0.0,
+        # MoE parameters
+        num_experts=160,
+        num_experts_per_tok=6,
+        num_shared_experts=2,
+        moe_intermediate_size=4096,
+        shared_expert_intermediate_size=14336,
+        **kwargs
+    ):
+        self.vocab_size = vocab_size
+        self.hidden_size = hidden_size
+        self.intermediate_size = intermediate_size
+        self.num_hidden_layers = num_hidden_layers
+        self.num_attention_heads = num_attention_heads
+        self.num_key_value_heads = num_key_value_heads
+        self.hidden_act = hidden_act
+        self.max_position_embeddings = max_position_embeddings
+        self.initializer_range = initializer_range
+        self.rms_norm_eps = rms_norm_eps
+        self.use_cache = use_cache
+        self.rope_theta = rope_theta
+        self.rope_scaling = rope_scaling
+        self.attention_bias = attention_bias
+        self.attention_dropout = attention_dropout
+        # MoE
+        self.num_experts = num_experts
+        self.num_experts_per_tok = num_experts_per_tok
+        self.num_shared_experts = num_shared_experts
+        self.moe_intermediate_size = moe_intermediate_size
+        self.shared_expert_intermediate_size = shared_expert_intermediate_size
+        super().__init__(
+            pad_token_id=pad_token_id,
+            bos_token_id=bos_token_id,
+            eos_token_id=eos_token_id,
+            tie_word_embeddings=tie_word_embeddings,
+            **kwargs
+        )
+class HelionOSCRMSNorm(nn.Module):
+    """RMS Normalization"""
+    def __init__(self, hidden_size, eps=1e-5):
+        super().__init__()
+        self.weight = nn.Parameter(torch.ones(hidden_size))
+        self.variance_epsilon = eps
+    def forward(self, hidden_states):
+        input_dtype = hidden_states.dtype
+        hidden_states = hidden_states.to(torch.float32)
+        variance = hidden_states.pow(2).mean(-1, keepdim=True)
+        hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)
+        return self.weight * hidden_states.to(input_dtype)
+class HelionOSCRotaryEmbedding(nn.Module):
+    """Rotary Position Embedding (RoPE)"""
+    def __init__(self, dim, max_position_embeddings=262144, base=10000000.0):
+        super().__init__()
+        self.dim = dim
+        self.max_position_embeddings = max_position_embeddings
+        self.base = base
+        # Compute frequency tensor
+        inv_freq = 1.0 / (self.base ** (torch.arange(0, self.dim, 2).float() / self.dim))
+        self.register_buffer("inv_freq", inv_freq, persistent=False)
+    def forward(self, x, seq_len=None):
+        if seq_len is None:
+            seq_len = x.shape[-2]
+        t = torch.arange(seq_len, device=x.device).type_as(self.inv_freq)
+        freqs = torch.outer(t, self.inv_freq)
+        emb = torch.cat((freqs, freqs), dim=-1)
+        return emb.cos().to(x.dtype), emb.sin().to(x.dtype)
+def rotate_half(x):
+    """Rotates half the hidden dims of the input."""
+    x1 = x[..., : x.shape[-1] // 2]
+    x2 = x[..., x.shape[-1] // 2 :]
+    return torch.cat((-x2, x1), dim=-1)
+def apply_rotary_pos_emb(q, k, cos, sin, position_ids=None):
+    """Apply rotary position embedding"""
+    if position_ids is None:
+        cos = cos.unsqueeze(0).unsqueeze(0)
+        sin = sin.unsqueeze(0).unsqueeze(0)
+    else:
+        cos = cos[position_ids].unsqueeze(1)
+        sin = sin[position_ids].unsqueeze(1)
+    q_embed = (q * cos) + (rotate_half(q) * sin)
+    k_embed = (k * cos) + (rotate_half(k) * sin)
+    return q_embed, k_embed
+class HelionOSCAttention(nn.Module):
+    """Multi-head attention with GQA"""
+    def __init__(self, config: HelionOSCConfig, layer_idx: Optional[int] = None):
+        super().__init__()
+        self.config = config
+        self.layer_idx = layer_idx
+        self.hidden_size = config.hidden_size
+        self.num_heads = config.num_attention_heads
+        self.head_dim = self.hidden_size // self.num_heads
+        self.num_key_value_heads = config.num_key_value_heads
+        self.num_key_value_groups = self.num_heads // self.num_key_value_heads
+        self.max_position_embeddings = config.max_position_embeddings
+        self.rope_theta = config.rope_theta
+        self.q_proj = nn.Linear(self.hidden_size, self.num_heads * self.head_dim, bias=config.attention_bias)
+        self.k_proj = nn.Linear(self.hidden_size, self.num_key_value_heads * self.head_dim, bias=config.attention_bias)
+        self.v_proj = nn.Linear(self.hidden_size, self.num_key_value_heads * self.head_dim, bias=config.attention_bias)
+        self.o_proj = nn.Linear(self.num_heads * self.head_dim, self.hidden_size, bias=config.attention_bias)
+        self.rotary_emb = HelionOSCRotaryEmbedding(
+            self.head_dim,
+            max_position_embeddings=self.max_position_embeddings,
+            base=self.rope_theta,
+        )
+    def forward(
+        self,
+        hidden_states: torch.Tensor,
+        attention_mask: Optional[torch.Tensor] = None,
+        position_ids: Optional[torch.LongTensor] = None,
+        past_key_value: Optional[Tuple[torch.Tensor]] = None,
+        output_attentions: bool = False,
+        use_cache: bool = False,
+    ) -> Tuple[torch.Tensor, Optional[torch.Tensor], Optional[Tuple[torch.Tensor]]]:
+        bsz, q_len, _ = hidden_states.size()
+        query_states = self.q_proj(hidden_states)
+        key_states = self.k_proj(hidden_states)
+        value_states = self.v_proj(hidden_states)
+        query_states = query_states.view(bsz, q_len, self.num_heads, self.head_dim).transpose(1, 2)
+        key_states = key_states.view(bsz, q_len, self.num_key_value_heads, self.head_dim).transpose(1, 2)
+        value_states = value_states.view(bsz, q_len, self.num_key_value_heads, self.head_dim).transpose(1, 2)
+        kv_seq_len = key_states.shape[-2]
+        if past_key_value is not None:
+            kv_seq_len += past_key_value[0].shape[-2]
+        cos, sin = self.rotary_emb(value_states, seq_len=kv_seq_len)
+        query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin, position_ids)
+        if past_key_value is not None:
+            key_states = torch.cat([past_key_value[0], key_states], dim=2)
+            value_states = torch.cat([past_key_value[1], value_states], dim=2)
+        past_key_value = (key_states, value_states) if use_cache else None
+        # Repeat k/v for GQA
+        key_states = key_states.repeat_interleave(self.num_key_value_groups, dim=1)
+        value_states = value_states.repeat_interleave(self.num_key_value_groups, dim=1)
+        attn_weights = torch.matmul(query_states, key_states.transpose(2, 3)) / math.sqrt(self.head_dim)
+        if attention_mask is not None:
+            attn_weights = attn_weights + attention_mask
+        attn_weights = nn.functional.softmax(attn_weights, dim=-1, dtype=torch.float32).to(query_states.dtype)
+        attn_weights = nn.functional.dropout(attn_weights, p=self.config.attention_dropout, training=self.training)
+        attn_output = torch.matmul(attn_weights, value_states)
+        attn_output = attn_output.transpose(1, 2).contiguous()
+        attn_output = attn_output.reshape(bsz, q_len, self.hidden_size)
+        attn_output = self.o_proj(attn_output)
+        return attn_output, attn_weights if output_attentions else None, past_key_value
+class HelionOSCMLP(nn.Module):
+    """Standard MLP (for shared experts or non-MoE layers)"""
+    def __init__(self, config: HelionOSCConfig, intermediate_size: Optional[int] = None):
+        super().__init__()
+        self.config = config
+        self.hidden_size = config.hidden_size
+        self.intermediate_size = intermediate_size or config.intermediate_size
+        self.gate_proj = nn.Linear(self.hidden_size, self.intermediate_size, bias=False)
+        self.up_proj = nn.Linear(self.hidden_size, self.intermediate_size, bias=False)
+        self.down_proj = nn.Linear(self.intermediate_size, self.hidden_size, bias=False)
+    def forward(self, x):
+        return self.down_proj(F.silu(self.gate_proj(x)) * self.up_proj(x))
+class HelionOSCMoE(nn.Module):
+    """Mixture of Experts layer"""
+    def __init__(self, config: HelionOSCConfig):
+        super().__init__()
+        self.config = config
+        self.num_experts = config.num_experts
+        self.top_k = config.num_experts_per_tok
+        self.hidden_size = config.hidden_size
+        # Router
+        self.gate = nn.Linear(config.hidden_size, config.num_experts, bias=False)
+        # Experts
+        self.experts = nn.ModuleList([
+            HelionOSCMLP(config, intermediate_size=config.moe_intermediate_size)
+            for _ in range(config.num_experts)
+        ])
+        # Shared experts
+        if config.num_shared_experts > 0:
+            self.shared_experts = HelionOSCMLP(
+                config,
+                intermediate_size=config.shared_expert_intermediate_size
+            )
+        else:
+            self.shared_experts = None
+    def forward(self, hidden_states: torch.Tensor) -> torch.Tensor:
+        batch_size, sequence_length, hidden_dim = hidden_states.shape
+        hidden_states = hidden_states.view(-1, hidden_dim)
+        # Router logits
+        router_logits = self.gate(hidden_states)
+        routing_weights = F.softmax(router_logits, dim=1, dtype=torch.float)
+        routing_weights, selected_experts = torch.topk(routing_weights, self.top_k, dim=-1)
+        routing_weights = routing_weights / routing_weights.sum(dim=-1, keepdim=True)
+        # Expert computation
+        final_hidden_states = torch.zeros(
+            (batch_size * sequence_length, hidden_dim),
+            dtype=hidden_states.dtype,
+            device=hidden_states.device
+        )
+        # Process each expert
+        expert_mask = torch.nn.functional.one_hot(selected_experts, num_classes=self.num_experts).permute(2, 1, 0)
+        for expert_idx in range(self.num_experts):
+            expert_layer = self.experts[expert_idx]
+            idx, top_x = torch.where(expert_mask[expert_idx])
+            if top_x.shape[0] == 0:
+                continue
+            top_x_list = top_x.tolist()
+            idx_list = idx.tolist()
+            current_state = hidden_states[None, top_x_list].reshape(-1, hidden_dim)
+            current_hidden_states = expert_layer(current_state) * routing_weights[top_x_list, idx_list, None]
+            final_hidden_states.index_add_(0, top_x, current_hidden_states.to(hidden_states.dtype))
+        # Add shared expert output
+        if self.shared_experts is not None:
+            final_hidden_states = final_hidden_states + self.shared_experts(hidden_states)
+        final_hidden_states = final_hidden_states.reshape(batch_size, sequence_length, hidden_dim)
+        return final_hidden_states
+class HelionOSCDecoderLayer(nn.Module):
+    """Single transformer decoder layer"""
+    def __init__(self, config: HelionOSCConfig, layer_idx: int):
+        super().__init__()
+        self.hidden_size = config.hidden_size
+        self.self_attn = HelionOSCAttention(config, layer_idx)
+        # Use MoE for most layers
+        self.mlp = HelionOSCMoE(config)
+        self.input_layernorm = HelionOSCRMSNorm(config.hidden_size, eps=config.rms_norm_eps)
+        self.post_attention_layernorm = HelionOSCRMSNorm(config.hidden_size, eps=config.rms_norm_eps)
+    def forward(
+        self,
+        hidden_states: torch.Tensor,
+        attention_mask: Optional[torch.Tensor] = None,
+        position_ids: Optional[torch.LongTensor] = None,
+        past_key_value: Optional[Tuple[torch.Tensor]] = None,
+        output_attentions: Optional[bool] = False,
+        use_cache: Optional[bool] = False,
+    ) -> Tuple[torch.FloatTensor, Optional[Tuple[torch.FloatTensor, torch.FloatTensor]]]:
+        residual = hidden_states
+        hidden_states = self.input_layernorm(hidden_states)
+        # Self Attention
+        hidden_states, self_attn_weights, present_key_value = self.self_attn(
+            hidden_states=hidden_states,
+            attention_mask=attention_mask,
+            position_ids=position_ids,
+            past_key_value=past_key_value,
+            output_attentions=output_attentions,
+            use_cache=use_cache,
+        )
+        hidden_states = residual + hidden_states
+        # MLP
+        residual = hidden_states
+        hidden_states = self.post_attention_layernorm(hidden_states)
+        hidden_states = self.mlp(hidden_states)
+        hidden_states = residual + hidden_states
+        outputs = (hidden_states,)
+        if output_attentions:
+            outputs += (self_attn_weights,)
+        if use_cache:
+            outputs += (present_key_value,)
+        return outputs
+class HelionOSCPreTrainedModel(PreTrainedModel):
+    """Pretrained model base class"""
+    config_class = HelionOSCConfig
+    base_model_prefix = "model"
+    supports_gradient_checkpointing = True
+    _no_split_modules = ["HelionOSCDecoderLayer"]
+    def _init_weights(self, module):
+        std = self.config.initializer_range
+        if isinstance(module, nn.Linear):
+            module.weight.data.normal_(mean=0.0, std=std)
+            if module.bias is not None:
+                module.bias.data.zero_()
+        elif isinstance(module, nn.Embedding):
+            module.weight.data.normal_(mean=0.0, std=std)
+            if module.padding_idx is not None:
+                module.weight.data[module.padding_idx].zero_()
+class HelionOSCModel(HelionOSCPreTrainedModel):
+    """Helion-OSC transformer model"""
+    def __init__(self, config: HelionOSCConfig):
+        super().__init__(config)
+        self.padding_idx = config.pad_token_id
+        self.vocab_size = config.vocab_size
+        self.embed_tokens = nn.Embedding(config.vocab_size, config.hidden_size, self.padding_idx)
+        self.layers = nn.ModuleList([
+            HelionOSCDecoderLayer(config, layer_idx)
+            for layer_idx in range(config.num_hidden_layers)
+        ])
+        self.norm = HelionOSCRMSNorm(config.hidden_size, eps=config.rms_norm_eps)
+        self.gradient_checkpointing = False
+        self.post_init()
+    def forward(
+        self,
+        input_ids: torch.LongTensor = None,
+        attention_mask: Optional[torch.Tensor] = None,
+        position_ids: Optional[torch.LongTensor] = None,
+        past_key_values: Optional[List[torch.FloatTensor]] = None,
+        inputs_embeds: Optional[torch.FloatTensor] = None,
+        use_cache: Optional[bool] = None,
+        output_attentions: Optional[bool] = None,
+        output_hidden_states: Optional[bool] = None,
+        return_dict: Optional[bool] = None,
+    ):
+        output_attentions = output_attentions if output_attentions is not None else self.config.output_attentions
+        output_hidden_states = output_hidden_states if output_hidden_states is not None else self.config.output_hidden_states
+        use_cache = use_cache if use_cache is not None else self.config.use_cache
+        return_dict = return_dict if return_dict is not None else self.config.use_return_dict
+        if input_ids is not None and inputs_embeds is not None:
+            raise ValueError("You cannot specify both input_ids and inputs_embeds at the same time")
+        elif input_ids is not None:
+            batch_size, seq_length = input_ids.shape[:2]
+        elif inputs_embeds is not None:
+            batch_size, seq_length = inputs_embeds.shape[:2]
+        else:
+            raise ValueError("You have to specify either input_ids or inputs_embeds")
+        if position_ids is None:
+            device = input_ids.device if input_ids is not None else inputs_embeds.device
+            position_ids = torch.arange(seq_length, dtype=torch.long, device=device)
+            position_ids = position_ids.unsqueeze(0)
+        if inputs_embeds is None:
+            inputs_embeds = self.embed_tokens(input_ids)
+        hidden_states = inputs_embeds
+        # Decoder layers
+        all_hidden_states = () if output_hidden_states else None
+        all_self_attns = () if output_attentions else None
+        next_decoder_cache = () if use_cache else None
+        for idx, decoder_layer in enumerate(self.layers):
+            if output_hidden_states:
+                all_hidden_states += (hidden_states,)
+            past_key_value = past_key_values[idx] if past_key_values is not None else None
+            layer_outputs = decoder_layer(
+                hidden_states,
+                attention_mask=attention_mask,
+                position_ids=position_ids,
+                past_key_value=past_key_value,
+                output_attentions=output_attentions,
+                use_cache=use_cache,
+            )
+            hidden_states = layer_outputs[0]
+            if use_cache:
+                next_decoder_cache += (layer_outputs[2 if output_attentions else 1],)
+            if output_attentions:
+                all_self_attns += (layer_outputs[1],)
+        hidden_states = self.norm(hidden_states)
+        if output_hidden_states:
+            all_hidden_states += (hidden_states,)
+        return tuple(v for v in [hidden_states, next_decoder_cache, all_hidden_states, all_self_attns] if v is not None)
+class HelionOSCForCausalLM(HelionOSCPreTrainedModel):
+    """Helion-OSC model with language modeling head"""
+    def __init__(self, config):
+        super().__init__(config)
+        self.model = HelionOSCModel(config)
+        self.vocab_size = config.vocab_size
+        self.lm_head = nn.Linear(config.hidden_size, config.vocab_size, bias=False)
+        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 get_output_embeddings(self):
+        return self.lm_head
+    def set_output_embeddings(self, new_embeddings):
+        self.lm_head = new_embeddings
+    def forward(
+        self,
+        input_ids: torch.LongTensor = None,
+        attention_mask: Optional[torch.Tensor] = None,
+        position_ids: Optional[torch.LongTensor] = None,
+        past_key_values: Optional[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,
+    ):
+        return_dict = return_dict if return_dict is not None else self.config.use_return_dict
+        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,
+            return_dict=return_dict,
+        )
+        hidden_states = outputs[0]
+        logits = self.lm_head(hidden_states)
+        loss = None
+        if labels is not None:
+            shift_logits = logits[..., :-1, :].contiguous()
+            shift_labels = labels[..., 1:].contiguous()
+            loss_fct = nn.CrossEntropyLoss()
+            loss = loss_fct(shift_logits.view(-1, self.config.vocab_size), shift_labels.view(-1))
+        if not return_dict:
+            output = (logits,) + outputs[1:]
+            return (loss,) + output if loss is not None else output
+        return CausalLMOutputWithPast(
+            loss=loss,
+            logits=logits,
+            past_key_values=outputs[1] if len(outputs) > 1 else None,
+            hidden_states=outputs[2] if len(outputs) > 2 else None,
+            attentions=outputs[3] if len(outputs) > 3 else None,
+        )
+    def prepare_inputs_for_generation(
+        self, input_ids, past_key_values=None, attention_mask=None, inputs_embeds=None, **kwargs
+    ):
+        if past_key_values:
+            input_ids = input_ids[:, -1:]
+        position_ids = kwargs.get("position_ids", None)
+        if attention_mask is not None and position_ids is None:
+            position_ids = attention_mask.long().cumsum(-1) - 1
+            position_ids.masked_fill_(attention_mask == 0, 1)
+            if past_key_values:
+                position_ids = position_ids[:, -1].unsqueeze(-1)
+        if inputs_embeds is not None and past_key_values is None:
+            model_inputs = {"inputs_embeds": inputs_embeds}
+        else:
+            model_inputs = {"input_ids": input_ids}
+        model_inputs.update({
+            "position_ids": position_ids,
+            "past_key_values": past_key_values,
+            "use_cache": kwargs.get("use_cache"),
+            "attention_mask": attention_mask,
+        })
+        return model_inputs