configuration_gear.py

# coding=utf-8
# Copyright 2025 OptAI Inc. HuggingFace Inc. team. All rights reserved.
#
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
from transformers.configuration_utils import PretrainedConfig, layer_type_validation
from transformers.modeling_rope_utils import rope_config_validation
from transformers.utils import logging


class GearConfig(PretrainedConfig):
    
    model_type = "gear"
    keys_to_ignore_at_inference = ["past_key_values"]
    base_model_tp_plan = {
        "layers.*.self_attn.q_proj": "colwise",
        "layers.*.self_attn.k_proj": "colwise",
        "layers.*.self_attn.v_proj": "colwise",
        "layers.*.self_attn.o_proj": "rowwise",
        "layers.*.mlp.gate_proj": "colwise",
        "layers.*.mlp.up_proj": "colwise",
        "layers.*.mlp.down_proj": "rowwise",
    }
    base_model_pp_plan = {
        "embed_tokens": (["input_ids"], ["inputs_embeds"]),
        "layers": (["hidden_states", "attention_mask"], ["hidden_states"]),
        "norm": (["hidden_states"], ["hidden_states"]),
    }

    def __init__(
        self,
        vocab_size=125_184,
        hidden_size=1152,
        intermediate_size=6912,
        num_hidden_layers=26,
        num_attention_heads=4,
        num_key_value_heads=1,
        head_dim=256,
        hidden_activation="gelu_pytorch_tanh",
        max_position_embeddings=131_072,
        initializer_range=0.02,
        rms_norm_eps=1e-6,
        use_cache=True,
        pad_token_id=125032,
        eos_token_id=125031,
        bos_token_id=125030,
        tie_word_embeddings=True,
        rope_theta=1_000_000.0,
        attention_bias=False,
        attention_dropout=0.0,
        query_pre_attn_scalar=256,
        sliding_window=512,
        layer_types=None,
        final_logit_softcapping=None,
        attn_logit_softcapping=None,
        rope_scaling=None,
        rope_local_base_freq=10_000.0,
        use_bidirectional_attention=False,
        conv_L_cache=3,
        **kwargs,
    ):
        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,
        )
        self.vocab_size = vocab_size
        self.max_position_embeddings = max_position_embeddings
        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.head_dim = head_dim
        self.num_key_value_heads = num_key_value_heads
        self.initializer_range = initializer_range
        self.rms_norm_eps = rms_norm_eps
        self.use_cache = use_cache
        self.rope_theta = rope_theta
        self.attention_bias = attention_bias
        self.attention_dropout = attention_dropout
        self.hidden_activation = hidden_activation
        self.query_pre_attn_scalar = query_pre_attn_scalar
        self.sliding_window = sliding_window
        self.final_logit_softcapping = final_logit_softcapping
        self.attn_logit_softcapping = attn_logit_softcapping
        self.layer_types = layer_types
        self.use_bidirectional_attention = use_bidirectional_attention
        if use_bidirectional_attention:
            self.sliding_window = (self.sliding_window // 2) + 1  # due to fa we set exclusive bounds

        self.rope_local_base_freq = rope_local_base_freq
        self.rope_scaling = rope_scaling
        rope_config_validation(self)

        # BC -> the pattern used to be a simple int, and it's still present in configs on the Hub
        self._sliding_window_pattern = kwargs.get("sliding_window_pattern", 6)

        self.conv_L_cache = conv_L_cache


__all__ = ["GearConfig"]