Added One hot fix for MOE model with subfunction (quic#777)

Signed-off-by: Abhishek Kumar Singh <sabhis@qti.qualcomm.com>

Author: abhishek-singh591

QEfficient/transformers/models/granitemoe/modeling_granitemoe.py

@@ -8,22 +8,20 @@
 from typing import List, Optional, Tuple, Type, Union
 
 import torch
-import torch.nn.functional as F
 from torch import nn
 from transformers.cache_utils import Cache, StaticCache
 from transformers.modeling_attn_mask_utils import AttentionMaskConverter
 from transformers.modeling_outputs import BaseModelOutputWithPast, MoeCausalLMOutputWithPast, MoeModelOutputWithPast
 from transformers.models.granitemoe.modeling_granitemoe import (
     GraniteMoeAttention,
     GraniteMoeConfig,
+    GraniteMoeDecoderLayer,
     GraniteMoeForCausalLM,
     GraniteMoeModel,
     GraniteMoeMoE,
     GraniteMoeParallelExperts,
     GraniteMoeRotaryEmbedding,
     GraniteMoeTopKGating,
-    load_balancing_loss_func,
-    logger,
     repeat_kv,
     rotate_half,
 )
@@ -198,6 +196,88 @@ def eager_attention_forward(
     return attn_output, attn_weights
 
 
+class QEffGraniteMoeDecoderLayer(GraniteMoeDecoderLayer):
+    """
+    Copied from GraniteForCausalLM: https://github.com/huggingface/transformers/blob/main/src/transformers/models/granite/modeling_granite.py
+    The only differences are:
+    - add new args batch idx for the CB models although its not supported yet.
+    """
+
+    def forward(
+        self,
+        hidden_states: torch.Tensor,
+        attention_mask: Optional[torch.Tensor] = None,
+        position_ids: Optional[torch.LongTensor] = None,
+        past_key_value: Optional[Cache] = None,
+        output_attentions: Optional[bool] = False,
+        use_cache: Optional[bool] = False,
+        cache_position: Optional[torch.LongTensor] = None,
+        output_router_logits: Optional[bool] = False,
+        position_embeddings: Optional[tuple[torch.Tensor, torch.Tensor]] = None,  # necessary, but kept here for BC
+        **kwargs,
+    ) -> tuple[torch.FloatTensor, Optional[tuple[torch.FloatTensor, torch.FloatTensor]]]:
+        """
+        Args:
+            hidden_states (`torch.FloatTensor`): input to the layer of shape `(batch, seq_len, embed_dim)`
+            attention_mask (`torch.FloatTensor`, *optional*):
+                attention mask of size `(batch_size, sequence_length)` if flash attention is used or `(batch_size, 1,
+                query_sequence_length, key_sequence_length)` if default attention is used.
+            output_attentions (`bool`, *optional*):
+                Whether or not to return the attentions tensors of all attention layers. See `attentions` under
+                returned tensors for more detail.
+            use_cache (`bool`, *optional*):
+                If set to `True`, `past_key_values` key value states are returned and can be used to speed up decoding
+                (see `past_key_values`).
+            past_key_value (`Tuple(torch.FloatTensor)`, *optional*): cached past key and value projection states
+            cache_position (`torch.LongTensor` of shape `(sequence_length)`, *optional*):
+                Indices depicting the position of the input sequence tokens in the sequence
+            output_router_logits (`bool`, *optional*):
+                Whether or not to return the logits of all the routers. They are useful for computing the router loss, and
+                should not be returned during inference.
+            position_embeddings (`tuple[torch.FloatTensor, torch.FloatTensor]`, *optional*):
+                Tuple containing the cosine and sine positional embeddings of shape `(batch_size, seq_len, head_dim)`,
+                with `head_dim` being the embedding dimension of each attention head.
+            kwargs (`dict`, *optional*):
+                Arbitrary kwargs to be ignored, used for FSDP and other methods that injects code
+                into the model
+        """
+        residual = hidden_states
+
+        hidden_states = self.input_layernorm(hidden_states)
+
+        # Self Attention
+        hidden_states, self_attn_weights = 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,
+            cache_position=cache_position,
+            position_embeddings=position_embeddings,
+            **kwargs,
+        )
+
+        hidden_states = residual + hidden_states * self.residual_multiplier
+
+        # Fully Connected
+        residual = hidden_states
+        hidden_states = self.post_attention_layernorm(hidden_states)
+        hidden_states, router_logits = self.block_sparse_moe(hidden_states)
+
+        hidden_states = residual + hidden_states * self.residual_multiplier
+
+        outputs = (hidden_states,)
+
+        if output_attentions:
+            outputs += (self_attn_weights,)
+
+        if output_router_logits:
+            outputs += (router_logits,)
+
+        return outputs
+
+
 class QEffGraniteMoeModel(GraniteMoeModel):
     """Copied from GraniteMoeModel: https://github.com/huggingface/transformers/blob/main/src/transformers/models/granitemoe/modeling_granitemoe.py
      The only differences are:
@@ -227,39 +307,19 @@ def forward(
             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 None) ^ (inputs_embeds is not None):
             raise ValueError("You must specify exactly one of input_ids or inputs_embeds")
 
-        if self.gradient_checkpointing and self.training and use_cache:
-            logger.warning_once(
-                "`use_cache=True` is incompatible with gradient checkpointing. Setting `use_cache=False`."
-            )
-            use_cache = False
-
         if inputs_embeds is None:
             inputs_embeds = self.embed_tokens(input_ids)
 
         inputs_embeds = inputs_embeds * self.embedding_multiplier  # main diff with Llama
 
-        # TODO (joao): remove this exception in v4.56 -- it exists for users that try to pass a legacy cache
-        # if not isinstance(past_key_values, (type(None), Cache)):
-        #    raise ValueError("The `past_key_values` should be either a `Cache` object or `None`.")
-
-        # if use_cache and past_key_values is None:
-        #    past_key_values = QEffDynamicCache()
-
+        return_legacy_cache = False
         if use_cache and not isinstance(past_key_values, Cache):
-            if past_key_values is None:
-                past_key_values = QEffDynamicCache()
-            else:
-                past_key_values = QEffDynamicCache.from_legacy_cache(past_key_values)
-                logger.warning_once(
-                    "We detected that you are passing `past_key_values` as a tuple of tuples. This is deprecated and "
-                    "will be removed in v4.47. Please convert your cache or use an appropriate `Cache` class "
-                    "(https://huggingface.co/docs/transformers/kv_cache#legacy-cache-format)"
-                )
+            return_legacy_cache = True
+            past_key_values = QEffDynamicCache.from_legacy_cache(past_key_values)
 
         if cache_position is None:
             past_seen_tokens = past_key_values.get_seq_length() if past_key_values is not None else 0
@@ -321,18 +381,15 @@ def forward(
         if output_hidden_states:
             all_hidden_states += (hidden_states,)
 
-        if not return_dict:
-            return tuple(
-                v for v in [hidden_states, past_key_values, all_hidden_states, all_self_attns] if v is not None
-            )
+        if return_legacy_cache:
+            past_key_values = past_key_values.to_legacy_cache()
 
-        output = MoeModelOutputWithPast(
+        return MoeModelOutputWithPast(
             last_hidden_state=hidden_states,
             past_key_values=past_key_values,
             hidden_states=all_hidden_states,
             attentions=all_self_attns,
         )
-        return output if return_dict else output.to_tuple()
 
     def _update_causal_mask(
         self,
@@ -435,7 +492,13 @@ def forward(self, hidden_states):
         logits = self.layer(hidden_states).float()
         top_k_logits, top_k_indices = torch.topk(logits, self.top_k, dim=1)  # [num_tokens, top_k]
         top_k_gates = torch.softmax(top_k_logits, dim=1).type_as(hidden_states)  # [num_tokens, top_k]
-        expert_mask = F.one_hot(top_k_indices, num_classes=self.num_experts).permute(2, 1, 0)
+
+        B, K = top_k_indices.shape
+        E = int(self.num_experts)
+        flat = top_k_indices.reshape(-1)
+        mask = torch.zeros((B * K, E), dtype=torch.int64, device=top_k_indices.device)
+        mask[torch.arange(B * K, device=flat.device), flat] = 1
+        expert_mask = mask.view(B, K, E).permute(2, 1, 0)
         return top_k_gates, expert_mask, logits, self.num_experts
 
 
@@ -511,14 +574,9 @@ def forward(
         comp_ctx_lengths: Optional[torch.LongTensor] = None,
         batch_index: Optional[torch.LongTensor] = 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,
-        output_router_logits: Optional[bool] = None,
-        return_dict: Optional[bool] = None,
         cache_position: Optional[torch.LongTensor] = None,
-        logits_to_keep: Union[int, torch.Tensor] = 0,
         **kwargs,
     ) -> Union[Tuple, MoeCausalLMOutputWithPast]:
         r"""
@@ -551,11 +609,9 @@ def forward(
         >>> tokenizer.batch_decode(generate_ids, skip_special_tokens=True, clean_up_tokenization_spaces=False)[0]
         "Hey, are you conscious? Can you talk to me?\nI'm not conscious, but I can talk to you."
         ```"""
-        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
         )
-        return_dict = return_dict if return_dict is not None else self.config.use_return_dict
 
         # decoder outputs consists of (dec_features, layer_state, dec_hidden, dec_attn)
         outputs = self.model(
@@ -567,57 +623,21 @@ def forward(
             batch_index=batch_index,
             inputs_embeds=inputs_embeds,
             use_cache=use_cache,
-            output_attentions=output_attentions,
             output_hidden_states=output_hidden_states,
-            return_dict=return_dict,
             cache_position=cache_position,
             **kwargs,
         )
 
-        hidden_states = outputs[0]
         # Cast to INT32 to avoid issue while running in ONNXRT
         logit_index = position_ids.to(torch.int32).argmax(1, keepdim=True)
-        hidden_states = outputs[0][torch.arange(position_ids.shape[0]).view(-1, 1), logit_index]
-
-        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, :])
-        logits = logits / self.config.logits_scaling
-
-        loss = None
-        if labels is not None:
-            # Upcast to float if we need to compute the loss to avoid potential precision issues
-            logits = logits.float()
-            # Flatten the tokens
-            loss = self.loss_function(
-                logits,
-                labels,
-                vocab_size=self.config.vocab_size,
-                **kwargs,
-            )
-
-        aux_loss = None
-        if output_router_logits:
-            aux_loss = load_balancing_loss_func(
-                outputs.router_logits if return_dict else outputs[-1],
-                self.num_experts,
-                self.num_experts_per_tok,
-                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
-
-        if not return_dict:
-            output = (logits,) + outputs[1:]
-            if output_router_logits:
-                output = (aux_loss,) + output
-            return (loss,) + output if loss is not None else output
+        hidden_states = outputs.last_hidden_state[torch.arange(position_ids.shape[0]).view(-1, 1), logit_index]
+        logits = self.lm_head(hidden_states).float()
+        # logits = logits / self.config.logits_scaling
 
         return MoeCausalLMOutputWithPast(
-            loss=loss,
-            aux_loss=aux_loss,
+            loss=None,
             logits=logits,
             past_key_values=outputs.past_key_values,
             hidden_states=outputs.hidden_states,
             attentions=outputs.attentions,
-            router_logits=outputs.router_logits,
         )

QEfficient/transformers/models/mixtral_moe/modeling_mixtral.py

@@ -219,7 +219,14 @@ def forward(self, hidden_states: torch.Tensor) -> torch.Tensor:
 
         # 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)
+        # selected_experts: [B, K]
+        B, K = selected_experts.shape
+        E = int(self.num_experts)
+        flat = selected_experts.reshape(-1)
+        mask = torch.zeros((B * K, E), dtype=torch.int64)
+        mask[torch.arange(B * K), flat] = 1
+        mask_bke = mask.view(B, K, E)
+        expert_mask = mask_bke.permute(2, 1, 0)
 
         # Loop over all available experts in the model and perform the computation on each expert
         for expert_idx in range(self.num_experts):

QEfficient/utils/torch_patches.py

@@ -11,8 +11,6 @@
 import torch.onnx.utils as onnx_utils
 from torch import _C
 
-from QEfficient.utils.logging_utils import logger
-
 # Store original references before patching
 _original_setup_trace_module_map = onnx_utils._setup_trace_module_map
 _original_get_module_attributes = getattr(onnx_utils, "_get_module_attributes", None)
@@ -43,7 +41,8 @@ def _track_module_attributes_forward_hook(module, input, output):
                 onnx_attrs = {}  # HACK: to reduce export time # TODO: study behaviour across models
                 _C._jit_pass_onnx_track_scope_attributes(graph, onnx_attrs)
             except Exception:
-                logger.warning("Failed to track ONNX scope attributes, Skipping this step.")
+                # Silently skip: scope-attribute tracking is best-effort and not required for export.
+                pass
 
         for m in model.modules():
             m.register_forward_hook(_track_module_attributes_forward_hook)

tests/transformers/models/test_subfunction.py

@@ -23,7 +23,7 @@
     ("gptj", 256, 2, 4, 128, 512, 127, {"rotary_dim": 16}),
     ("llama", 256, 2, 4, 128, 512, 127, {"num_key_value_heads": 2}),
     ("mistral", 256, 2, 4, 128, 512, 127, {"num_key_value_heads": 2}),
-    # ("mixtral", 256, 2, 4, 128, 512, 127, {"num_key_value_heads": 2}),
+    ("mixtral", 256, 2, 4, 128, 512, 127, {"num_key_value_heads": 2}),
     ("mpt", 256, 2, 4, 128, 512, 127, {}),
     ("phi", 256, 2, 4, 128, 512, 127, {}),
     ("phi3", 256, 2, 4, 128, 512, 127, {"pad_token_id": 0}),
@@ -34,6 +34,7 @@
     ("olmo2", 256, 2, 4, 128, 512, 127, {"num_key_value_heads": 2}),
     ("gpt_oss", 256, 3, 4, 128, 512, 127, {"num_key_value_heads": 2}),
     ("qwen3_moe", 256, 2, 4, 128, 512, 127, {"num_key_value_heads": 2}),
+    ("granitemoe", 256, 2, 4, 128, 512, 127, {"num_key_value_heads": 2}),
 ]
 
 configs = [