some fixes

Signed-off-by: Yu Chin Fabian Lim <[email protected]>

plugins/fused-ops-and-kernels/src/fms_acceleration_foak/kernels/liger/fused_linear_cross_entropy_loss.py

@@ -30,22 +30,6 @@
 import torch.nn.functional as F
 from torch.nn import CrossEntropyLoss
 from transformers.modeling_outputs import CausalLMOutputWithPast
-from transformers.models.llama.modeling_llama import (
-    _CONFIG_FOR_DOC,
-    LLAMA_INPUTS_DOCSTRING,
-)
-from transformers.models.mixtral.modeling_mixtral import (
-    _CONFIG_FOR_DOC,
-    MIXTRAL_INPUTS_DOCSTRING,
-)
-from transformers.modeling_outputs import (
-    MoeCausalLMOutputWithPast,
-    MoeModelOutputWithPast,
-)
-from transformers.utils import (
-    add_start_docstrings_to_model_forward,
-    replace_return_docstrings,
-)
 
 from .cross_entropy import (
     element_mul_kernel,
@@ -297,11 +281,6 @@ def forward(self, lin_weight, _input, target, bias=None):
             self.reduction,
         )
 
-# TODO: how to add diff docstrings for diff model types? what if the loss functions aren't the same across models?
-# @add_start_docstrings_to_model_forward(LLAMA_INPUTS_DOCSTRING)
-@replace_return_docstrings(
-    output_type=CausalLMOutputWithPast, config_class=_CONFIG_FOR_DOC
-)
 def lce_forward(
     self,
     input_ids: torch.LongTensor = None,
@@ -435,143 +414,4 @@ def lce_forward(
         past_key_values=outputs.past_key_values,
         hidden_states=outputs.hidden_states,
         attentions=outputs.attentions,
-    )
-
-# TODO: is adding a separate copy of lce_forward() the right path or should the additional logic for Moe models be in the single lce_forward?
-@add_start_docstrings_to_model_forward(MIXTRAL_INPUTS_DOCSTRING)
-@replace_return_docstrings(output_type=MoeCausalLMOutputWithPast, config_class=_CONFIG_FOR_DOC)
-# Ignore copy
-def lce_forward_mixtral(
-    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,
-    output_router_logits: Optional[bool] = None,
-    return_dict: Optional[bool] = None,
-    cache_position: Optional[torch.LongTensor] = None,
-    num_logits_to_keep: int = 0,
-) -> Union[Tuple, MoeCausalLMOutputWithPast]:
-    r"""
-    Args:
-        labels (`torch.LongTensor` of shape `(batch_size, sequence_length)`, *optional*):
-            Labels for computing the masked language modeling loss. Indices should either be in `[0, ...,
-            config.vocab_size]` or -100 (see `input_ids` docstring). Tokens with indices set to `-100` are ignored
-            (masked), the loss is only computed for the tokens with labels in `[0, ..., config.vocab_size]`.
-
-        num_logits_to_keep (`int`, *optional*):
-            Calculate logits for the last `num_logits_to_keep` tokens. If `0`, calculate logits for all
-            `input_ids` (special case). Only last token logits are needed for generation, and calculating them only for that
-            token can save memory, which becomes pretty significant for long sequences or large vocabulary size.
-
-    Returns:
-
-    Example:
-
-    ```python
-    >>> from transformers import AutoTokenizer, MixtralForCausalLM
-
-    >>> model = MixtralForCausalLM.from_pretrained("mistralai/Mixtral-8x7B-v0.1")
-    >>> tokenizer = AutoTokenizer.from_pretrained("mistralai/Mixtral-8x7B-v0.1")
-
-    >>> prompt = "Hey, are you conscious? Can you talk to me?"
-    >>> inputs = tokenizer(prompt, return_tensors="pt")
-
-    >>> # Generate
-    >>> generate_ids = model.generate(inputs.input_ids, max_length=30)
-    >>> 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_router_logits = (
-        output_router_logits if output_router_logits is not None else self.config.output_router_logits
-    )
-
-    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(
-        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 = outputs[0]
-
-    loss = None
-    logits = None
-
-    # patch change
-    if self.training and (labels is not None):
-        shift_hidden_states = hidden_states[..., :-1, :].contiguous()
-        shift_labels = labels[..., 1:].contiguous()
-
-        # flatten tokens
-        shift_hidden_states = shift_hidden_states.view(-1, self.config.hidden_size)
-        shift_labels = shift_labels.view(-1)
-
-        lce = LigerFusedLinearCrossEntropyLoss()
-        loss = lce(self.lm_head.weight, shift_hidden_states, shift_labels)
-    else:
-        # Only compute necessary logits, and do not upcast them to float if we are not computing the loss
-        logits = self.lm_head(hidden_states[:, -num_logits_to_keep:, :])
-
-        if labels is not None:
-            # Upcast to float if we need to compute the loss to avoid potential precision issues
-            logits = logits.float()
-            # Shift so that tokens < n predict n
-            shift_logits = logits[..., :-1, :].contiguous()
-            shift_labels = labels[..., 1:].contiguous()
-            # Flatten the tokens
-            loss_fct = CrossEntropyLoss()
-            shift_logits = shift_logits.view(-1, self.config.vocab_size)
-            shift_labels = shift_labels.view(-1)
-            # Enable model parallelism
-            shift_labels = shift_labels.to(shift_logits.device)
-            loss = loss_fct(shift_logits, shift_labels)
-
-    # TODO: unique differing part to mixtral model forward
-    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,
-        )
-        # TODO: should this loss manipulation be indented in?? or should it be added to even the liger loss?
-        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
-
-    return MoeCausalLMOutputWithPast(
-        loss=loss,
-        aux_loss=aux_loss,
-        logits=logits,
-        past_key_values=outputs.past_key_values,
-        hidden_states=outputs.hidden_states,
-        attentions=outputs.attentions,
-        router_logits=outputs.router_logits,
     )

plugins/fused-ops-and-kernels/src/fms_acceleration_foak/models/mixtral.py

@@ -23,7 +23,6 @@
     combine_triggers,
 )
 from transformers.models.mixtral.modeling_mixtral import (
-    MixtralForCausalLM,
     MixtralAttention,
     MixtralRMSNorm,
 )
@@ -32,7 +31,6 @@
 from ..kernels.unsloth.cross_entropy_loss import FastCrossEntropyLoss
 from ..kernels.unsloth.rms_layernorm import fast_rms_layernorm
 from ..kernels.unsloth.rope_embedding import fast_rope_embedding
-from ..kernels.liger.fused_linear_cross_entropy_loss import lce_forward_mixtral
 from .utils import KEY_O, KEY_QKV, build_lora_fused_ops, trigger_fused_ops
 
 
@@ -95,11 +93,6 @@ def get_mp_rules(base_type):
                 "transformers.models.mixtral.modeling_mixtral",
             ),
         ),
-        ModelPatcherRule(
-            rule_id="mixtral-fused-lce",
-            trigger=ModelPatcherTrigger(check=MixtralForCausalLM),
-            forward=lce_forward_mixtral,
-        ),
         ModelPatcherRule(
             rule_id="mixtral-rope",
             import_and_maybe_reload=(

scripts/benchmarks/benchmark.py

@@ -723,7 +723,7 @@ def prepare_arguments(args, benchmark_dataset: BenchmarkDataset):
 
         if (
             not args.run_only_scenarios
-            and scenarios.slow
+            and scenario.slow
         ):
             # unfiltered runs omit all "slow" marked scenarios
             print(f"Skipping slow scenario '{_scn_name}' beacuse run_only_scenarios=None.")

scripts/benchmarks/scenarios-liger.yaml

@@ -38,22 +38,18 @@
 scenarios:
     -   name: full-finetuning
         framework_config: 
-            - 
             - foak-fast-kernels
             - foak-fast-kernels-liger
         arguments:
             learning_rate: 2e-5
             model_name_or_path: 
-                - 'bigcode/gpt_bigcode-santacoder'
-                - 'mistralai/Mistral-7B-v0.1'
-                - 'mistralai/Mixtral-8x7B-Instruct-v0.1'
-                - 'NousResearch/Llama-2-70b-hf'
+                # - 'mistralai/Mistral-7B-v0.1'
+                - 'meta-llama/Meta-Llama-3-8B'
             torch_dtype: bfloat16
             bf16: True
 
     -   name: standard-peft
         framework_config: 
-            - 
             - foak-fast-kernels
             - foak-fast-kernels-liger
         arguments:
@@ -66,13 +62,29 @@ scenarios:
             lora_dropout: 0.1
             target_modules: ["q_proj", "k_proj", "v_proj", "o_proj"]
             model_name_or_path: 
-                - 'mistralai/Mistral-7B-v0.1'
-                - 'mistralai/Mixtral-8x7B-Instruct-v0.1'
-                - 'NousResearch/Llama-2-70b-hf'
+                # - 'mistralai/Mistral-7B-v0.1'
+                - 'meta-llama/Meta-Llama-3-8B'
+
+    -   name: accelerated-peft-bnb
+        framework_config: 
+            - accelerated-peft-bnb-foak
+            - accelerated-peft-bnb-liger
+        arguments:
+            bf16: True
+            learning_rate: 2e-4
+            torch_dtype: bfloat16
+            peft_method: lora
+            r: 16
+            lora_alpha: 16
+            lora_dropout: 0.1
+            per_device_train_batch_size: 
+            target_modules: ["q_proj", "k_proj", "v_proj", "o_proj"]
+            model_name_or_path: 
+                # - 'mistralai/Mistral-7B-v0.1'
+                - 'meta-llama/Meta-Llama-3-8B'
 
     -   name: accelerated-peft-gptq
         framework_config: 
-            - accelerated-peft-autogptq
             - accelerated-peft-autogptq-foak
             - accelerated-peft-autogptq-foak-liger
         arguments:
@@ -85,6 +97,5 @@ scenarios:
             lora_dropout: 0.1
             target_modules: ["q_proj", "k_proj", "v_proj", "o_proj"]
             model_name_or_path: 
-                - 'TheBloke/Mistral-7B-v0.1-GPTQ'
-                - 'TheBloke/Mixtral-8x7B-Instruct-v0.1-GPTQ'
-                - 'TheBloke/Llama-2-70B-GPTQ'
+                # - 'TheBloke/Mistral-7B-v0.1-GPTQ'
+                - 'meta-llama/Meta-Llama-3-8B'