easydel.modules.internlm2.modeling_internlm2

class easydel.modules.internlm2.modeling_internlm2.InternLM2Attention(*args: Any, **kwargs: Any)

Bases: UnifiedAttention

InternLM2 Attention with full RoPE.

Inherits from UnifiedAttention. Uses combined QKV projection (wqkv) instead of separate projections. Overrides forward_standard to efficiently handle fused QKV with GQA layout.

projection_mapping: ClassVar[dict[str, str]] = {'output_projection': 'wo', 'query_key_value_projection': 'wqkv'}

class easydel.modules.internlm2.modeling_internlm2.InternLM2Block(*args: Any, **kwargs: Any)

Bases: Module

InternLM2 Transformer Block.

This module combines the self-attention layer and the MLP layer with residual connections and layer normalization.

config

Configuration object for the model.

Type

InternLM2Config

dtype

Data type for computation. Default is jnp.float32.

Type

jnp.dtype

param_dtype

Data type for parameters. Default is jnp.float32.

Type

jnp.dtype

precision

Precision setting for JAX operations. Default is None.

Type

jax.lax.PrecisionLike

attention

The self-attention module.

Type

InternLM2Attention

feed_forward

The feed-forward (MLP) module.

Type

InternLM2MLP

attention_norm

Layer normalization before the attention layer.

Type

RMSNorm

ffn_norm

Layer normalization before the MLP layer.

Type

RMSNorm

class easydel.modules.internlm2.modeling_internlm2.InternLM2ForCausalLM(*args: Any, **kwargs: Any)

Bases: BaseCausalLMModule[InternLM2Model, InternLM2Config]

InternLM2 model with a Causal Language Modeling head.

get_decoder()

Returns the decoder part of the model’s graph definition.

get_embedding()

Returns the embedding layer of the module.

get_encoder()

Returns the encoder part of the model’s graph definition. Decoder-Only models don’t have an encoder.

get_lm_head()

Returns the language model head of the module.

class easydel.modules.internlm2.modeling_internlm2.InternLM2ForSequenceClassification(*args: Any, **kwargs: Any)

Bases: EasyDeLBaseModule

InternLM2 model with a Sequence Classification head.

This model consists of the base InternLM2 transformer (InternLM2Model) followed by a linear layer (score) that projects the transformer’s output hidden states (typically the hidden state of the first token) to the number of classes for classification.

config

Configuration object for the model.

Type

InternLM2Config

dtype

Data type for computation. Default is jnp.float32.

Type

jnp.dtype

param_dtype

Data type for parameters. Default is jnp.float32.

Type

jnp.dtype

precision

Precision setting for JAX operations. Default is None.

Type

jax.lax.PrecisionLike

rngs

Random number generators.

Type

nn.Rngs

module

The core InternLM2 transformer model.

Type

InternLM2Model

score

The linear layer for classification.

Type

ParallelLinear

get_decoder()

Returns the decoder part of the model’s graph definition.

get_embedding()

Returns the embedding layer of the module.

get_encoder()

Returns the encoder part of the model’s graph definition. Decoder-Only models don’t have an encoder.

get_lm_head()

Returns the language model head of the module. This model has a sequence classification head, not an LM Head.

class easydel.modules.internlm2.modeling_internlm2.InternLM2MLP(*args: Any, **kwargs: Any)

Bases: Module

InternLM2 MLP module.

config

Configuration object for the model.

Type

InternLM2Config

dtype

Data type for computation. Default is jnp.float32.

Type

jnp.dtype

param_dtype

Data type for parameters. Default is jnp.float32.

Type

jnp.dtype

precision

Precision setting for JAX operations. Default is None.

Type

jax.lax.PrecisionLike

w1

First linear transformation (gate projection).

Type

ParallelLinear

w3

Second linear transformation (up projection).

Type

ParallelLinear

w2

Third linear transformation (down projection).

Type

ParallelLinear

act_fn

Activation function (e.g., SiLU).

Type

callable

class easydel.modules.internlm2.modeling_internlm2.InternLM2Model(*args: Any, **kwargs: Any)

Bases: EasyDeLBaseModule

The base InternLM2 model transformer.

This class represents the core transformer architecture of the InternLM2 model, consisting of embedding layers, multiple transformer blocks, and a final layer normalization.

config

Configuration object for the model.

Type

InternLM2Config

dtype

Data type for computation. Default is jnp.float32.

Type

jnp.dtype

param_dtype

Data type for parameters. Default is jnp.float32.

Type

jnp.dtype

precision

Precision setting for JAX operations. Default is None.

Type

jax.lax.PrecisionLike

embed_tokens

Embedding layer for input tokens.

Type

nn.Embed

layers

Sequence of transformer blocks.

Type

tp.Sequence[InternLM2Block]

norm

Final layer normalization.

Type

RMSNorm

gradient_checkpointing

Gradient checkpointing configuration.

Type

EasyDeLGradientCheckPointers

scan_layers

Whether to use JAX scan for layer processing.

Type

bool

blocks_class

The class used for the transformer blocks.

Type

InternLM2Block

get_decoder()

Returns the decoder part of the model’s graph definition.

get_embedding()

Returns the embedding layer of the module.

get_encoder()

Returns the encoder part of the model’s graph definition. Decoder-Only models don’t have an encoder.

get_lm_head()

Returns the language model head of the module. Base Models don’t have a Language Model Head.