add starcoder2

Author: reymondzzzz

src/transformers_neuronx/__init__.py

@@ -28,6 +28,7 @@
 from transformers_neuronx.mistral.model import MistralForSampling
 from transformers_neuronx.mixtral.model import MixtralForSampling
 from transformers_neuronx.opt.model import OPTForSampling
+from transformers_neuronx.starcoder2.model import Starcoder2ForSampling
 
 from transformers_neuronx.modeling_auto import NeuronAutoModelForCausalLM
 

src/transformers_neuronx/modeling_auto.py

@@ -12,6 +12,7 @@
     "mistral":    transformers_neuronx.MistralForSampling,
     "mixtral":    transformers_neuronx.MixtralForSampling,
     "opt":        transformers_neuronx.OPTForSampling,
+    "starcoder2": transformers_neuronx.Starcoder2ForSampling,
 }
 
 
@@ -24,6 +25,7 @@
     transformers.MistralConfig: "mistral",
     transformers.MixtralConfig: "mixtral",
     transformers.OPTConfig:     "opt",
+    transformers.Starcoder2Config: "starcoder2",
 }
 
 

src/transformers_neuronx/starcoder2/config.py

@@ -0,0 +1,39 @@
+from transformers_neuronx import utils
+
+
+class Starcoder2Config:
+    def __init__(
+            self,
+            config,
+            n_positions,
+            batch_size,
+            amp,
+            tp_degree,
+            **kwargs
+    ):
+        # Extract configs used for building HLO
+        self.intermediate_size = config.intermediate_size
+        self.hidden_size = config.hidden_size
+
+        self.attention_head_size = config.hidden_size // config.num_attention_heads
+        self.num_attention_heads = config.num_attention_heads
+        self.num_key_value_heads = config.num_key_value_heads if hasattr(config,
+                                                                         "num_key_value_heads") else config.num_attention_heads
+        self.num_hidden_layers = config.num_hidden_layers
+        self.vocab_size = config.vocab_size
+        self.hidden_act = config.hidden_act
+        self.bos_token_id = config.bos_token_id
+        self.eos_token_id = config.eos_token_id
+        self.max_position_embeddings = config.max_position_embeddings
+        self.rms_norm_eps = config.norm_epsilon
+        self.rotary_percentage = getattr(config, "rotary_percentage", 1)
+        self.rope_theta = getattr(config, "rope_theta", 10000)
+        self.position_interpolation_factor = getattr(config, "position_interpolation_factor", None)
+        self.use_bias = getattr(config, "use_bias", True)
+        utils.maybe_override_attributes(self, kwargs)
+
+        # Add required Neuron configs
+        self.n_positions = n_positions
+        self.batch_size = batch_size
+        self.amp = amp
+        self.tp_degree = tp_degree

src/transformers_neuronx/starcoder2/hlo.py

@@ -0,0 +1,264 @@
+from typing import Optional
+
+from transformers_neuronx import constants
+from transformers_neuronx import hlo
+from transformers_neuronx import utils
+from transformers_neuronx.config import NeuronConfig
+from transformers_neuronx.constants import LAYOUT_HSB, LAYOUT_BSH
+from transformers_neuronx.hlo import mlp, mlp_bsh
+from transformers_neuronx.layers import transformer, rotary, attention, attention_utils, flash_decoding
+from transformers_neuronx.starcoder2.config import Starcoder2Config
+
+
+class Starcoder2ForSamplingNoEmbeddingHlo:
+
+    def __init__(self,
+                 config: Starcoder2Config,
+                 neuron_config: Optional[NeuronConfig] = None
+                 ):
+        self.config = config
+        self.neuron_config = neuron_config
+        self.n_positions = None
+
+    @property
+    def shard_over_batch(self):
+        # Property access allows fallback configuration to be enabled after construction
+        return (
+                self.neuron_config is not None
+                and self.neuron_config.group_query_attention == constants.GQA.SHARD_OVER_BATCH
+        )
+
+    def inputs(self, scribe, dtype, n_active_tokens, batch_size):
+        tensors, dims = transformer.inputs(
+            scribe, dtype, batch_size, n_active_tokens, self.config.hidden_size, self.neuron_config)
+
+        return tensors, dims
+
+    def embedding(self, input_ids, cache_ids, start_ids, last_token_id, embed_weight):
+        dtype = getattr(input_ids.scribe, self.config.amp)
+        hidden = hlo.embedding(embed_weight, input_ids, tp_degree=self.config.tp_degree, dtype=dtype)
+        if self.config.hidden_size % self.config.tp_degree != 0:
+            hidden = hlo.slice_along(hidden, dim=-1, limit=self.config.hidden_size, start=0)
+        if self.neuron_config.attention_layout == LAYOUT_HSB:
+            hidden = hlo.transpose210(hidden)
+        return hidden
+
+    def pre_layer(self, hidden, cache_ids, start_ids, last_token_id, *weights):
+        head_dim = self.config.attention_head_size
+        pos_embed = rotary.hlo_rotary_embedding(
+            hidden.dtype, int(head_dim * self.config.rotary_percentage), cache_ids,
+            base=self.config.rope_theta,
+            interpolation_factor=self.config.position_interpolation_factor
+        )
+        mask, active_mask = hlo.attention_mask(cache_ids, start_ids, self.n_positions)
+        core_id = None
+        if self.neuron_config.shard_over_sequence:
+            core_id, *rst = weights
+            n_kv_heads = self.config.num_key_value_heads if self.config.num_attention_heads else self.config.num_attention_heads
+            cores_per_kv_head = self.config.tp_degree // n_kv_heads
+            self.cores_per_kv_head = cores_per_kv_head if cores_per_kv_head > 1 else self.config.tp_degree
+            cache_ids, mask, active_mask = flash_decoding.convert_attn_mask_and_cache_id(cache_ids,
+                                                                                         core_id, self.n_positions,
+                                                                                         cores_per_kv_head=self.cores_per_kv_head)
+
+        return hidden, last_token_id, pos_embed, cache_ids, start_ids, mask, active_mask, core_id
+
+    def layer(
+            self, hidden, last_token_id, pos_embed, cache_ids, start_ids, mask, active_mask, core_id,
+            attn_k_cache, attn_v_cache,
+            pre_attn_ln_weight, pre_attn_ln_bias,
+            attn_q_weight, attn_q_scales, attn_q_bias,
+            attn_k_weight, attn_k_scales, attn_k_bias,
+            attn_v_weight, attn_v_scales, attn_v_bias,
+            attn_out_weight, attn_out_scales, attn_out_bias,
+            post_attn_ln_weight, post_attn_ln_bias,
+            pre_mlp_ln_weight, pre_mlp_ln_bias,
+            mlp_in_weight, mlp_in_scales, mlp_in_bias,
+            mlp_out_weight, mlp_out_scales, mlp_out_bias,
+            post_mlp_ln_weight, post_mlp_ln_bias,
+    ):
+        is_bsh = self.neuron_config and self.neuron_config.attention_layout == LAYOUT_BSH
+        layer_norm_ = hlo.layer_norm_bsh if is_bsh else hlo.layer_norm
+        ln_hidden = layer_norm_(hidden, pre_attn_ln_weight, pre_attn_ln_bias)
+
+        attn_output, out_attn_k_cache, out_attn_v_cache = self.attention(
+            ln_hidden, cache_ids, start_ids, pos_embed, mask, active_mask, core_id,
+            attn_k_cache, attn_v_cache,
+            attn_q_weight, attn_q_scales, attn_q_bias,
+            attn_k_weight, attn_k_scales, attn_k_bias,
+            attn_v_weight, attn_v_scales, attn_v_bias,
+            attn_out_weight, attn_out_scales, attn_out_bias
+        )
+        hidden = hlo.add(attn_output, hidden)
+
+        norm_hidden = layer_norm_(hidden, pre_mlp_ln_weight, pre_mlp_ln_bias)
+        mlp_ = mlp_bsh if is_bsh else mlp
+        mlp_hidden = mlp_(
+            norm_hidden,
+            mlp_in_weight, mlp_in_bias, mlp_out_weight, mlp_out_bias,
+            in_scales=mlp_in_scales,
+            out_scales=mlp_out_scales,
+            activation_function='gelu_new',  # 'gelu_pytorch_tanh',
+            tp_degree=self.config.tp_degree,
+            neuron_config=self.neuron_config,
+        )
+        res_hidden = hlo.add(mlp_hidden, hidden)
+        return res_hidden, out_attn_k_cache, out_attn_v_cache
+
+    def ln_lm_head(self, hidden, last_token_id, rms_weight, unused_bias, lm_head_weight, lm_head_bias,
+                   return_all_outputs=True):
+        logits = transformer.rms_lm_head(self.config.tp_degree, hidden, last_token_id, rms_weight, lm_head_weight,
+                                         lm_head_bias, return_all_outputs, eps=self.config.rms_norm_eps,
+                                         neuron_config=self.neuron_config)
+        return logits
+
+    def attention(
+            self,
+            hidden, cache_ids, start_ids, pos_embed, mask, active_mask, core_id,
+            cached_keys, cached_values,
+            q_weight, q_scales, q_bias,
+            k_weight, k_scales, k_bias,
+            v_weight, v_scales, v_bias,
+            out_weight, out_scales, out_bias,
+    ):
+        d_head = self.config.attention_head_size
+        tp_degree = self.config.tp_degree
+
+        # Compute the expected number of KV heads (Used in case fused QKV is used)
+        n_kv_heads_tp = None
+        if self.config.num_key_value_heads is not None:
+            n_head = self.config.num_attention_heads
+            n_kv_head = self.config.num_key_value_heads
+            _, n_kv_head_padded = utils.get_qkv_padding(n_head, n_kv_head, tp_degree, self.neuron_config)
+            n_kv_heads_tp = n_kv_head_padded // tp_degree
+
+        # Q = (hidden @ wQ) + bQ
+        # K = (hidden @ wK) + bK
+        # V = (hidden @ wV) + bV
+        query, key, value = attention.query_key_value(
+            hidden,
+            q_weight, q_scales, q_bias,
+            k_weight, k_scales, k_bias,
+            v_weight, v_scales, v_bias,
+            d_head,
+            neuron_config=self.neuron_config,
+            tp_degree=tp_degree,  # TODO: include tp_degree into neuron_config
+            shard_over_batch=self.shard_over_batch,
+            n_kv_heads_tp=n_kv_heads_tp,
+        )
+
+        # Q = Rotate(Q)
+        # K = Rotate(K)
+        query, key = rotary.rotate_half(query, key, pos_embed, self.config.rotary_percentage,
+                                        tp_degree=tp_degree, shard_over_batch=self.shard_over_batch)
+
+        # Q = Q / sqrt(d_head)
+        query = attention.scale(query, d_head)
+
+        # In BSH cache layout, the output of QKV linear projection is still kept as SBH for all QKV.
+        bsh_cache_layout = False
+        batch_dim = 1
+        if self.neuron_config is not None:
+            bsh_cache_layout = self.neuron_config.cache_layout == constants.LAYOUT_BSH
+        if bsh_cache_layout:
+            query, key, value = attention_utils.transpose_qkv(query, key, value)
+            batch_dim = 0
+
+        # Single Token Generation ("Prefetch"-style) ans speculative forward
+        if active_mask is not None:
+
+            n_active_tokens = key.sizes[1] if bsh_cache_layout else key.sizes[0]
+            if n_active_tokens > 1 and self.neuron_config and self.neuron_config.continuous_batching:
+                # For speculative forward + continuous batching, slice out samples in the batch size
+                # corresponding to the batch size of the speculative head
+                slice_sizes = [1] * len(cached_keys.sizes)
+                if cached_keys.sizes[batch_dim] == 1:
+                    # Use hlo.select for batch size 1 as index select is prohibitively slow
+                    # TODO: revert to hlo.index_select once its faster P126527643
+                    cached_keys_s = hlo.select(cached_keys, batch_dim, hlo.reshape(start_ids, slice_sizes),
+                                               keepdim=True)
+                    cached_values_s = hlo.select(cached_values, batch_dim, hlo.reshape(start_ids, slice_sizes),
+                                                 keepdim=True)
+                else:
+                    cached_keys_s = hlo.index_select(cached_keys, batch_dim, start_ids)
+                    cached_values_s = hlo.index_select(cached_values, batch_dim, start_ids)
+            else:
+                cached_keys_s = cached_keys
+                cached_values_s = cached_values
+            # Communication 1: all-gather query from cores
+            if (n_active_tokens != self.n_positions) and self.neuron_config.shard_over_sequence:
+                query = flash_decoding.gather_query_group(query, self.cores_per_kv_head,
+                                                          self.config.num_attention_heads,
+                                                          tp_degree)
+
+            # Sp = Q @ Kp
+            prior_scores = attention.score(query, cached_keys_s, n_kv_heads=self.config.num_key_value_heads,
+                                           tp_degree=tp_degree, neuron_config=self.neuron_config)
+            prior_scores = attention.mask(prior_scores, mask, tp_degree=tp_degree,
+                                          shard_over_batch=self.shard_over_batch)
+
+            # Sa = Q @ Ka
+            active_score = attention.score(query, key, n_kv_heads=self.config.num_key_value_heads,
+                                           tp_degree=tp_degree, neuron_config=self.neuron_config)
+            active_score = attention.mask(active_score, active_mask, tp_degree=tp_degree,
+                                          shard_over_batch=self.shard_over_batch)
+
+            # C = softmax(Sa, Sp) @ (Va, Vp)
+            if self.neuron_config.shard_over_sequence:
+                dtype = query.dtype
+                context = flash_decoding.context(prior_scores, active_score, cached_values_s, value, core_id, mask,
+                                                 active_mask,
+                                                 n_kv_heads=self.config.num_key_value_heads,
+                                                 n_heads=self.config.num_attention_heads, dtype=dtype,
+                                                 tp_degree=tp_degree, neuron_config=self.neuron_config,
+                                                 shard_over_batch=self.shard_over_batch)
+                cache_ids, value, key = flash_decoding.select_values_within_bound(cache_ids, value, key,
+                                                                                  self.cores_per_kv_head, core_id,
+                                                                                  dim=0)
+
+            else:
+                context = attention.context(prior_scores, active_score, cached_values_s, value,
+                                            n_kv_heads=self.config.num_key_value_heads, tp_degree=tp_degree,
+                                            neuron_config=self.neuron_config)
+
+            # KCache[I], VCache[I] = K, V
+            updated_keys, updated_values = attention.fused_kv_update_cache(cached_keys, cached_values, cache_ids,
+                                                                           key, value, start_ids,
+                                                                           neuron_config=self.neuron_config)
+
+        # Multi-Token Context Encoding
+        else:
+            _, batch_size, _, _ = query.sizes
+            if self.neuron_config.lhs_aligned or batch_size == 1:
+                context = attention.flash_attention(query, key, value)
+            else:
+                # do not use flash attention for lhs padded (right aligned) batch > 1 case
+                # because it does not correctly take mask into account
+                context = None
+
+            if context is None:
+                # S = Q @ K
+
+                score = attention.score(query, key, n_kv_heads=self.config.num_key_value_heads,
+                                        tp_degree=tp_degree, neuron_config=self.neuron_config)
+                score = attention.mask(score, mask, tp_degree=tp_degree, shard_over_batch=self.shard_over_batch)
+                context = attention.context_combined(score, value, n_kv_heads=self.config.num_key_value_heads,
+                                                     tp_degree=tp_degree, neuron_config=self.neuron_config)
+
+            if self.neuron_config.shard_over_sequence:
+                cache_ids, value, key = flash_decoding.select_values_within_bound(cache_ids,
+                                                                                  value,
+                                                                                  key,
+                                                                                  self.cores_per_kv_head,
+                                                                                  core_id, dim=0)
+            # KCache, VCache = K, V
+            if cached_keys.sizes == key.sizes:
+                updated_keys, updated_values = key, value
+            else:
+                updated_keys, updated_values = attention.fused_kv_update_cache(cached_keys, cached_values, cache_ids,
+                                                                               key, value, start_ids,
+                                                                               neuron_config=self.neuron_config)
+
+        # O = (C @ wO) + bO
+        output = attention.output(context, out_weight, out_scales, out_bias, tp_degree, self.neuron_config)
+        return output, updated_keys, updated_values