Generate modeling_mixtral.py from modular_mixtral.py

- Auto-generate modeling_mixtral.py with the same fix
- Apply black formatting
- Fix repository consistency check

Author: akacmazz

src/transformers/models/mixtral/modeling_mixtral.py

@@ -67,7 +67,9 @@ def __init__(self, config: MixtralConfig):
         self.act_fn = ACT2FN[config.hidden_act]
 
     def forward(self, hidden_states):
-        current_hidden_states = self.act_fn(self.w1(hidden_states)) * self.w3(hidden_states)
+        current_hidden_states = self.act_fn(self.w1(hidden_states)) * self.w3(
+            hidden_states
+        )
         current_hidden_states = self.w2(current_hidden_states)
         return current_hidden_states
 
@@ -94,7 +96,9 @@ def __init__(self, config):
         # gating
         self.gate = nn.Linear(self.hidden_dim, self.num_experts, bias=False)
 
-        self.experts = nn.ModuleList([MixtralBlockSparseTop2MLP(config) for _ in range(self.num_experts)])
+        self.experts = nn.ModuleList(
+            [MixtralBlockSparseTop2MLP(config) for _ in range(self.num_experts)]
+        )
 
         # Jitter parameters
         self.jitter_noise = config.router_jitter_noise
@@ -103,39 +107,53 @@ def forward(self, hidden_states: torch.Tensor) -> torch.Tensor:
         """ """
         batch_size, sequence_length, hidden_dim = hidden_states.shape
         if self.training and self.jitter_noise > 0:
-            hidden_states *= torch.empty_like(hidden_states).uniform_(1.0 - self.jitter_noise, 1.0 + self.jitter_noise)
+            hidden_states *= torch.empty_like(hidden_states).uniform_(
+                1.0 - self.jitter_noise, 1.0 + self.jitter_noise
+            )
         hidden_states = hidden_states.view(-1, hidden_dim)
         # router_logits: (batch * sequence_length, n_experts)
         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, selected_experts = torch.topk(
+            routing_weights, self.top_k, dim=-1
+        )
         routing_weights /= routing_weights.sum(dim=-1, keepdim=True)
         # we cast back to the input dtype
         routing_weights = routing_weights.to(hidden_states.dtype)
 
         final_hidden_states = torch.zeros(
-            (batch_size * sequence_length, hidden_dim), dtype=hidden_states.dtype, device=hidden_states.device
+            (batch_size * sequence_length, hidden_dim),
+            dtype=hidden_states.dtype,
+            device=hidden_states.device,
         )
 
         # One hot encode the selected experts to create an expert mask
         # this will be used to easily index which expert is going to be sollicitated
-        expert_mask = torch.nn.functional.one_hot(selected_experts, num_classes=self.num_experts).permute(2, 1, 0)
+        expert_mask = torch.nn.functional.one_hot(
+            selected_experts, num_classes=self.num_experts
+        ).permute(2, 1, 0)
 
-        expert_hit = torch.greater(expert_mask.sum(dim=(-1, -2)), 0).nonzero()
-        for expert_idx in expert_hit:
+        # Loop over all available experts in the model and perform the computation on each expert
+        for expert_idx in range(self.num_experts):
             expert_layer = self.experts[expert_idx]
-            idx, top_x = torch.where(expert_mask[expert_idx].squeeze(0))
+            idx, top_x = torch.where(expert_mask[expert_idx])
             # Index the correct hidden states and compute the expert hidden state for
             # the current expert. We need to make sure to multiply the output hidden
             # states by `routing_weights` on the corresponding tokens (top-1 and top-2)
             current_state = hidden_states[None, top_x].reshape(-1, hidden_dim)
-            current_hidden_states = expert_layer(current_state) * routing_weights[top_x, idx, None]
+            current_hidden_states = (
+                expert_layer(current_state) * routing_weights[top_x, idx, None]
+            )
 
             # However `index_add_` only support torch tensors for indexing so we'll use
             # the `top_x` tensor here.
-            final_hidden_states.index_add_(0, top_x, current_hidden_states.to(hidden_states.dtype))
-        final_hidden_states = final_hidden_states.reshape(batch_size, sequence_length, hidden_dim)
+            final_hidden_states.index_add_(
+                0, top_x, current_hidden_states.to(hidden_states.dtype)
+            )
+        final_hidden_states = final_hidden_states.reshape(
+            batch_size, sequence_length, hidden_dim
+        )
         return final_hidden_states, router_logits
 
 
@@ -202,7 +220,9 @@ def repeat_kv(hidden_states: torch.Tensor, n_rep: int) -> torch.Tensor:
     batch, num_key_value_heads, slen, head_dim = hidden_states.shape
     if n_rep == 1:
         return hidden_states
-    hidden_states = hidden_states[:, :, None, :, :].expand(batch, num_key_value_heads, n_rep, slen, head_dim)
+    hidden_states = hidden_states[:, :, None, :, :].expand(
+        batch, num_key_value_heads, n_rep, slen, head_dim
+    )
     return hidden_states.reshape(batch, num_key_value_heads * n_rep, slen, head_dim)
 
 
@@ -224,8 +244,12 @@ def eager_attention_forward(
         causal_mask = attention_mask[:, :, :, : key_states.shape[-2]]
         attn_weights = attn_weights + causal_mask
 
-    attn_weights = nn.functional.softmax(attn_weights, dim=-1, dtype=torch.float32).to(query.dtype)
-    attn_weights = nn.functional.dropout(attn_weights, p=dropout, training=module.training)
+    attn_weights = nn.functional.softmax(attn_weights, dim=-1, dtype=torch.float32).to(
+        query.dtype
+    )
+    attn_weights = nn.functional.dropout(
+        attn_weights, p=dropout, training=module.training
+    )
     attn_output = torch.matmul(attn_weights, value_states)
     attn_output = attn_output.transpose(1, 2).contiguous()
 
@@ -239,15 +263,28 @@ def __init__(self, config: MixtralConfig, layer_idx: int):
         super().__init__()
         self.config = config
         self.layer_idx = layer_idx
-        self.head_dim = getattr(config, "head_dim", None) or config.hidden_size // config.num_attention_heads
-        self.num_key_value_groups = config.num_attention_heads // config.num_key_value_heads
+        self.head_dim = (
+            getattr(config, "head_dim", None)
+            or config.hidden_size // config.num_attention_heads
+        )
+        self.num_key_value_groups = (
+            config.num_attention_heads // config.num_key_value_heads
+        )
         self.scaling = self.head_dim**-0.5
         self.attention_dropout = config.attention_dropout
         self.is_causal = True
-        self.q_proj = nn.Linear(config.hidden_size, config.num_attention_heads * self.head_dim, bias=False)
-        self.k_proj = nn.Linear(config.hidden_size, config.num_key_value_heads * self.head_dim, bias=False)
-        self.v_proj = nn.Linear(config.hidden_size, config.num_key_value_heads * self.head_dim, bias=False)
-        self.o_proj = nn.Linear(config.num_attention_heads * self.head_dim, config.hidden_size, bias=False)
+        self.q_proj = nn.Linear(
+            config.hidden_size, config.num_attention_heads * self.head_dim, bias=False
+        )
+        self.k_proj = nn.Linear(
+            config.hidden_size, config.num_key_value_heads * self.head_dim, bias=False
+        )
+        self.v_proj = nn.Linear(
+            config.hidden_size, config.num_key_value_heads * self.head_dim, bias=False
+        )
+        self.o_proj = nn.Linear(
+            config.num_attention_heads * self.head_dim, config.hidden_size, bias=False
+        )
 
     @deprecate_kwarg("past_key_value", new_name="past_key_values", version="4.58")
     def forward(
@@ -267,16 +304,22 @@ def forward(
         value_states = self.v_proj(hidden_states).view(hidden_shape).transpose(1, 2)
 
         cos, sin = position_embeddings
-        query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)
+        query_states, key_states = apply_rotary_pos_emb(
+            query_states, key_states, cos, sin
+        )
 
         if past_key_values is not None:
             # sin and cos are specific to RoPE models; cache_position needed for the static cache
             cache_kwargs = {"sin": sin, "cos": cos, "cache_position": cache_position}
-            key_states, value_states = past_key_values.update(key_states, value_states, self.layer_idx, cache_kwargs)
+            key_states, value_states = past_key_values.update(
+                key_states, value_states, self.layer_idx, cache_kwargs
+            )
 
         attention_interface: Callable = eager_attention_forward
         if self.config._attn_implementation != "eager":
-            attention_interface = ALL_ATTENTION_FUNCTIONS[self.config._attn_implementation]
+            attention_interface = ALL_ATTENTION_FUNCTIONS[
+                self.config._attn_implementation
+            ]
 
         attn_output, attn_weights = attention_interface(
             self,
@@ -286,7 +329,9 @@ def forward(
             attention_mask,
             dropout=0.0 if not self.training else self.attention_dropout,
             scaling=self.scaling,
-            sliding_window=getattr(self.config, "sliding_window", None),  # main diff with Llama
+            sliding_window=getattr(
+                self.config, "sliding_window", None
+            ),  # main diff with Llama
             **kwargs,
         )
 
@@ -303,8 +348,12 @@ def __init__(self, config: MixtralConfig, layer_idx: int):
         self.self_attn = MixtralAttention(config, layer_idx)
 
         self.block_sparse_moe = MixtralSparseMoeBlock(config)
-        self.input_layernorm = MixtralRMSNorm(config.hidden_size, eps=config.rms_norm_eps)
-        self.post_attention_layernorm = MixtralRMSNorm(config.hidden_size, eps=config.rms_norm_eps)
+        self.input_layernorm = MixtralRMSNorm(
+            config.hidden_size, eps=config.rms_norm_eps
+        )
+        self.post_attention_layernorm = MixtralRMSNorm(
+            config.hidden_size, eps=config.rms_norm_eps
+        )
 
     @deprecate_kwarg("past_key_value", new_name="past_key_values", version="4.58")
     def forward(
@@ -349,7 +398,9 @@ def __init__(self, config: MixtralConfig, device=None):
         super().__init__()
         # BC: "rope_type" was originally "type"
         if hasattr(config, "rope_scaling") and isinstance(config.rope_scaling, dict):
-            self.rope_type = config.rope_scaling.get("rope_type", config.rope_scaling.get("type"))
+            self.rope_type = config.rope_scaling.get(
+                "rope_type", config.rope_scaling.get("type")
+            )
         else:
             self.rope_type = "default"
         self.max_seq_len_cached = config.max_position_embeddings
@@ -365,12 +416,23 @@ def __init__(self, config: MixtralConfig, device=None):
     @torch.no_grad()
     @dynamic_rope_update  # power user: used with advanced RoPE types (e.g. dynamic rope)
     def forward(self, x, position_ids):
-        inv_freq_expanded = self.inv_freq[None, :, None].float().expand(position_ids.shape[0], -1, 1).to(x.device)
+        inv_freq_expanded = (
+            self.inv_freq[None, :, None]
+            .float()
+            .expand(position_ids.shape[0], -1, 1)
+            .to(x.device)
+        )
         position_ids_expanded = position_ids[:, None, :].float()
 
-        device_type = x.device.type if isinstance(x.device.type, str) and x.device.type != "mps" else "cpu"
+        device_type = (
+            x.device.type
+            if isinstance(x.device.type, str) and x.device.type != "mps"
+            else "cpu"
+        )
         with torch.autocast(device_type=device_type, enabled=False):  # Force float32
-            freqs = (inv_freq_expanded.float() @ position_ids_expanded.float()).transpose(1, 2)
+            freqs = (
+                inv_freq_expanded.float() @ position_ids_expanded.float()
+            ).transpose(1, 2)
             emb = torch.cat((freqs, freqs), dim=-1)
             cos = emb.cos() * self.attention_scaling
             sin = emb.sin() * self.attention_scaling
@@ -404,9 +466,14 @@ def __init__(self, config: MixtralConfig):
         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.embed_tokens = nn.Embedding(
+            config.vocab_size, config.hidden_size, self.padding_idx
+        )
         self.layers = nn.ModuleList(
-            [MixtralDecoderLayer(config, layer_idx) for layer_idx in range(config.num_hidden_layers)]
+            [
+                MixtralDecoderLayer(config, layer_idx)
+                for layer_idx in range(config.num_hidden_layers)
+            ]
         )
         self.norm = MixtralRMSNorm(config.hidden_size, eps=config.rms_norm_eps)
         self.rotary_emb = MixtralRotaryEmbedding(config=config)
@@ -429,7 +496,9 @@ def forward(
         **kwargs: Unpack[TransformersKwargs],
     ) -> MoeModelOutputWithPast:
         if (input_ids is None) ^ (inputs_embeds is not None):
-            raise ValueError("You must specify exactly one of input_ids or inputs_embeds")
+            raise ValueError(
+                "You must specify exactly one of input_ids or inputs_embeds"
+            )
 
         if use_cache and past_key_values is None:
             past_key_values = DynamicCache(config=self.config)
@@ -438,14 +507,22 @@ def forward(
             inputs_embeds = self.embed_tokens(input_ids)
 
         if cache_position is None:
-            past_seen_tokens = past_key_values.get_seq_length() if past_key_values is not None else 0
+            past_seen_tokens = (
+                past_key_values.get_seq_length() if past_key_values is not None else 0
+            )
             cache_position = torch.arange(
-                past_seen_tokens, past_seen_tokens + inputs_embeds.shape[1], device=inputs_embeds.device
+                past_seen_tokens,
+                past_seen_tokens + inputs_embeds.shape[1],
+                device=inputs_embeds.device,
             )
         if position_ids is None:
             position_ids = cache_position.unsqueeze(0)
 
-        mask_function = create_causal_mask if self.config.sliding_window is None else create_sliding_window_causal_mask
+        mask_function = (
+            create_causal_mask
+            if self.config.sliding_window is None
+            else create_sliding_window_causal_mask
+        )
         causal_mask = mask_function(
             config=self.config,
             input_embeds=inputs_embeds,
@@ -514,7 +591,9 @@ def load_balancing_loss_func(
 
     if isinstance(gate_logits, tuple):
         compute_device = gate_logits[0].device
-        concatenated_gate_logits = torch.cat([layer_gate.to(compute_device) for layer_gate in gate_logits], dim=0)
+        concatenated_gate_logits = torch.cat(
+            [layer_gate.to(compute_device) for layer_gate in gate_logits], dim=0
+        )
 
     routing_weights = torch.nn.functional.softmax(concatenated_gate_logits, dim=-1)
 
@@ -530,20 +609,24 @@ def load_balancing_loss_func(
         router_prob_per_expert = torch.mean(routing_weights, dim=0)
     else:
         batch_size, sequence_length = attention_mask.shape
-        num_hidden_layers = concatenated_gate_logits.shape[0] // (batch_size * sequence_length)
+        num_hidden_layers = concatenated_gate_logits.shape[0] // (
+            batch_size * sequence_length
+        )
 
         # Compute the mask that masks all padding tokens as 0 with the same shape of expert_mask
         expert_attention_mask = (
             attention_mask[None, :, :, None, None]
-            .expand((num_hidden_layers, batch_size, sequence_length, top_k, num_experts))
+            .expand(
+                (num_hidden_layers, batch_size, sequence_length, top_k, num_experts)
+            )
             .reshape(-1, top_k, num_experts)
             .to(compute_device)
         )
 
         # Compute the percentage of tokens routed to each experts
-        tokens_per_expert = torch.sum(expert_mask.float() * expert_attention_mask, dim=0) / torch.sum(
-            expert_attention_mask, dim=0
-        )
+        tokens_per_expert = torch.sum(
+            expert_mask.float() * expert_attention_mask, dim=0
+        ) / torch.sum(expert_attention_mask, dim=0)
 
         # Compute the mask that masks all padding tokens as 0 with the same shape of tokens_per_expert
         router_per_expert_attention_mask = (
@@ -554,9 +637,9 @@ def load_balancing_loss_func(
         )
 
         # Compute the average probability of routing to these experts
-        router_prob_per_expert = torch.sum(routing_weights * router_per_expert_attention_mask, dim=0) / torch.sum(
-            router_per_expert_attention_mask, dim=0
-        )
+        router_prob_per_expert = torch.sum(
+            routing_weights * router_per_expert_attention_mask, dim=0
+        ) / torch.sum(router_per_expert_attention_mask, dim=0)
 
     overall_loss = torch.sum(tokens_per_expert * router_prob_per_expert.unsqueeze(0))
     return overall_loss * num_experts
@@ -626,7 +709,9 @@ def forward(
         ```"""
 
         output_router_logits = (
-            output_router_logits if output_router_logits is not None else self.config.output_router_logits
+            output_router_logits
+            if output_router_logits is not None
+            else self.config.output_router_logits
         )
 
         # decoder outputs consists of (dec_features, layer_state, dec_hidden, dec_attn)
@@ -644,7 +729,11 @@ def forward(
 
         hidden_states = outputs.last_hidden_state
         # Only compute necessary logits, and do not upcast them to float if we are not computing the loss
-        slice_indices = slice(-logits_to_keep, None) if isinstance(logits_to_keep, int) else logits_to_keep
+        slice_indices = (
+            slice(-logits_to_keep, None)
+            if isinstance(logits_to_keep, int)
+            else logits_to_keep
+        )
         logits = self.lm_head(hidden_states[:, slice_indices, :])
 
         loss = None
@@ -660,7 +749,9 @@ def forward(
                 attention_mask,
             )
             if labels is not None:
-                loss += self.router_aux_loss_coef * aux_loss.to(loss.device)  # make sure to reside in the same device
+                loss += self.router_aux_loss_coef * aux_loss.to(
+                    loss.device
+                )  # make sure to reside in the same device
 
         return MoeCausalLMOutputWithPast(
             loss=loss,
@@ -673,11 +764,15 @@ def forward(
         )
 
 
-class MixtralForSequenceClassification(GenericForSequenceClassification, MixtralPreTrainedModel):
+class MixtralForSequenceClassification(
+    GenericForSequenceClassification, MixtralPreTrainedModel
+):
     pass
 
 
-class MixtralForTokenClassification(GenericForTokenClassification, MixtralPreTrainedModel):
+class MixtralForTokenClassification(
+    GenericForTokenClassification, MixtralPreTrainedModel
+):
     pass