feat(server): add frequency penalty (#1541)

Cargo.lock

@@ -2787,7 +2787,7 @@ dependencies = [
  "tabled",
  "text-generation-client",
  "thiserror",
- "tokenizers",
+ "tokenizers 0.14.1",
  "tokio",
  "tracing",
  "tracing-subscriber",
@@ -2850,7 +2850,7 @@ dependencies = [
  "serde_json",
  "text-generation-client",
  "thiserror",
- "tokenizers",
+ "tokenizers 0.15.1",
  "tokio",
  "tokio-stream",
  "tower-http",
@@ -2972,6 +2972,40 @@ dependencies = [
  "unicode_categories",
 ]
 
+[[package]]
+name = "tokenizers"
+version = "0.15.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "6db445cceba5dfeb0f9702be7d6bfd91801ddcbe8fe8722defe7f2e96da75812"
+dependencies = [
+ "aho-corasick",
+ "clap",
+ "derive_builder",
+ "esaxx-rs",
+ "getrandom",
+ "hf-hub",
+ "indicatif",
+ "itertools 0.11.0",
+ "lazy_static",
+ "log",
+ "macro_rules_attribute",
+ "monostate",
+ "onig",
+ "paste",
+ "rand",
+ "rayon",
+ "rayon-cond",
+ "regex",
+ "regex-syntax 0.7.5",
+ "serde",
+ "serde_json",
+ "spm_precompiled",
+ "thiserror",
+ "unicode-normalization-alignments",
+ "unicode-segmentation",
+ "unicode_categories",
+]
+
 [[package]]
 name = "tokio"
 version = "1.35.1"

benchmark/src/lib.rs

@@ -30,6 +30,7 @@ pub async fn run(
     top_p: Option<f32>,
     typical_p: Option<f32>,
     repetition_penalty: Option<f32>,
+    frequency_penalty: Option<f32>,
     watermark: bool,
     do_sample: bool,
     client: ShardedClient,
@@ -42,6 +43,7 @@ pub async fn run(
         do_sample,
         seed: 0,
         repetition_penalty: repetition_penalty.unwrap_or(1.0),
+        frequency_penalty: frequency_penalty.unwrap_or(0.0),
         watermark,
     };
 
@@ -140,6 +142,7 @@ pub async fn run(
         top_p,
         typical_p,
         repetition_penalty,
+        frequency_penalty,
         watermark,
         do_sample,
     );

benchmark/src/main.rs

@@ -84,6 +84,11 @@ struct Args {
     #[clap(long, env)]
     repetition_penalty: Option<f32>,
 
+    /// Generation parameter in case you want to specifically test/debug particular
+    /// decoding strategies, for full doc refer to the `text-generation-server`
+    #[clap(long, env)]
+    frequency_penalty: Option<f32>,
+
     /// Generation parameter in case you want to specifically test/debug particular
     /// decoding strategies, for full doc refer to the `text-generation-server`
     #[clap(long, env)]
@@ -119,6 +124,7 @@ fn main() -> Result<(), Box<dyn std::error::Error>> {
         top_p,
         typical_p,
         repetition_penalty,
+        frequency_penalty,
         watermark,
         do_sample,
         master_shard_uds_path,
@@ -187,6 +193,7 @@ fn main() -> Result<(), Box<dyn std::error::Error>> {
                 top_p,
                 typical_p,
                 repetition_penalty,
+                frequency_penalty,
                 watermark,
                 do_sample,
                 sharded_client,

benchmark/src/table.rs

@@ -15,6 +15,7 @@ pub(crate) fn parameters_table(
     top_p: Option<f32>,
     typical_p: Option<f32>,
     repetition_penalty: Option<f32>,
+    frequency_penalty: Option<f32>,
     watermark: bool,
     do_sample: bool,
 ) -> Table {
@@ -33,6 +34,7 @@ pub(crate) fn parameters_table(
     builder.push_record(["Top P", &format!("{top_p:?}")]);
     builder.push_record(["Typical P", &format!("{typical_p:?}")]);
     builder.push_record(["Repetition Penalty", &format!("{repetition_penalty:?}")]);
+    builder.push_record(["Frequency Penalty", &format!("{frequency_penalty:?}")]);
     builder.push_record(["Watermark", &watermark.to_string()]);
     builder.push_record(["Do Sample", &do_sample.to_string()]);
 

integration-tests/models/test_mamba.py

@@ -24,6 +24,7 @@ async def test_mamba(fused_kernel_mamba, response_snapshot):
     assert response.generated_text == "\n\nDeep learning is a new type of machine"
     assert response == response_snapshot
 
+
 @pytest.mark.asyncio
 @pytest.mark.private
 async def test_mamba_all_params(fused_kernel_mamba, response_snapshot):
@@ -44,13 +45,19 @@ async def test_mamba_all_params(fused_kernel_mamba, response_snapshot):
     )
 
     assert response.details.generated_tokens == 10
-    assert response.generated_text == "blue, red, yellow, \nand orange (in the order they appear in"
+    assert (
+        response.generated_text
+        == "blue, red, yellow, \nand orange (in the order they appear in"
+    )
     assert response == response_snapshot
 
+
 @pytest.mark.asyncio
 @pytest.mark.private
 async def test_mamba_load(fused_kernel_mamba, generate_load, response_snapshot):
-    responses = await generate_load(fused_kernel_mamba, "What is Deep Learning?", max_new_tokens=10, n=4)
+    responses = await generate_load(
+        fused_kernel_mamba, "What is Deep Learning?", max_new_tokens=10, n=4
+    )
 
     assert len(responses) == 4
     assert all([r.generated_text == responses[0].generated_text for r in responses])

proto/generate.proto

@@ -66,6 +66,8 @@ message NextTokenChooserParameters {
     uint64 seed = 6;
     /// repetition penalty
     float repetition_penalty = 7;
+    /// frequency penalty
+    float frequency_penalty = 9;
     /// token watermarking using "A Watermark for Large Language Models"
     bool watermark = 8;
 }

router/client/src/client.rs

@@ -125,6 +125,7 @@ impl Client {
                     do_sample: false,
                     seed: 0,
                     repetition_penalty: 1.2,
+                    frequency_penalty: 0.1,
                     watermark: true,
                 }),
                 stopping_parameters: Some(StoppingCriteriaParameters {

router/src/health.rs

@@ -43,6 +43,7 @@ impl Health {
                     do_sample: false,
                     seed: 0,
                     repetition_penalty: 1.0,
+                    frequency_penalty: 0.0,
                     watermark: false,
                 }),
                 stopping_parameters: Some(StoppingCriteriaParameters {

router/src/lib.rs

@@ -106,6 +106,14 @@ pub(crate) struct GenerateParameters {
     )]
     pub repetition_penalty: Option<f32>,
     #[serde(default)]
+    #[schema(
+        exclusive_minimum = -2.0,
+        nullable = true,
+        default = "null",
+        example = 0.1
+    )]
+    pub frequency_penalty: Option<f32>,
+    #[serde(default)]
     #[schema(exclusive_minimum = 0, nullable = true, default = "null", example = 10)]
     pub top_k: Option<i32>,
     #[serde(default)]
@@ -172,6 +180,7 @@ fn default_parameters() -> GenerateParameters {
         best_of: None,
         temperature: None,
         repetition_penalty: None,
+        frequency_penalty: None,
         top_k: None,
         top_p: None,
         typical_p: None,
@@ -205,10 +214,71 @@ pub(crate) struct ChatCompletion {
 pub(crate) struct ChatCompletionComplete {
     pub index: u32,
     pub message: Message,
-    pub logprobs: Option<Vec<f32>>,
+    pub logprobs: Option<ChatCompletionLogprobs>,
     pub finish_reason: String,
 }
 
+#[derive(Clone, Deserialize, Serialize, ToSchema)]
+pub(crate) struct ChatCompletionLogprobs {
+    content: Vec<ChatCompletionLogprob>,
+}
+
+impl From<(Token, Vec<Token>)> for ChatCompletionLogprobs {
+    fn from(value: (Token, Vec<Token>)) -> Self {
+        let (token, top_tokens) = value;
+
+        Self {
+            content: vec![ChatCompletionLogprob {
+                token: token.text,
+                logprob: token.logprob,
+                top_logprobs: top_tokens
+                    .into_iter()
+                    .map(|t| ChatCompletionTopLogprob {
+                        token: t.text,
+                        logprob: t.logprob,
+                    })
+                    .collect(),
+            }],
+        }
+    }
+}
+
+impl From<(Vec<Token>, Vec<Vec<Token>>)> for ChatCompletionLogprobs {
+    fn from(value: (Vec<Token>, Vec<Vec<Token>>)) -> Self {
+        let (tokens, top_tokens) = value;
+        Self {
+            content: tokens
+                .into_iter()
+                .zip(top_tokens)
+                .map(|(t, top_t)| ChatCompletionLogprob {
+                    token: t.text,
+                    logprob: t.logprob,
+                    top_logprobs: top_t
+                        .into_iter()
+                        .map(|t| ChatCompletionTopLogprob {
+                            token: t.text,
+                            logprob: t.logprob,
+                        })
+                        .collect(),
+                })
+                .collect(),
+        }
+    }
+}
+
+#[derive(Clone, Deserialize, Serialize, ToSchema)]
+pub(crate) struct ChatCompletionLogprob {
+    token: String,
+    logprob: f32,
+    top_logprobs: Vec<ChatCompletionTopLogprob>,
+}
+
+#[derive(Clone, Deserialize, Serialize, ToSchema)]
+pub(crate) struct ChatCompletionTopLogprob {
+    token: String,
+    logprob: f32,
+}
+
 #[derive(Clone, Deserialize, Serialize)]
 pub(crate) struct Usage {
     pub prompt_tokens: u32,
@@ -238,7 +308,7 @@ impl ChatCompletion {
                     content: output,
                 },
                 logprobs: return_logprobs
-                    .then(|| details.tokens.iter().map(|t| t.logprob).collect()),
+                    .then(|| ChatCompletionLogprobs::from((details.tokens, details.top_tokens))),
                 finish_reason: details.finish_reason.to_string(),
             }],
             usage: Usage {
@@ -266,7 +336,7 @@ pub(crate) struct ChatCompletionChunk {
 pub(crate) struct ChatCompletionChoice {
     pub index: u32,
     pub delta: ChatCompletionDelta,
-    pub logprobs: Option<f32>,
+    pub logprobs: Option<ChatCompletionLogprobs>,
     pub finish_reason: Option<String>,
 }
 
@@ -285,7 +355,7 @@ impl ChatCompletionChunk {
         delta: String,
         created: u64,
         index: u32,
-        logprobs: Option<f32>,
+        logprobs: Option<ChatCompletionLogprobs>,
         finish_reason: Option<String>,
     ) -> Self {
         Self {
@@ -319,8 +389,8 @@ pub(crate) struct ChatRequest {
     /// UNUSED
     #[schema(example = "mistralai/Mistral-7B-Instruct-v0.2")]
     /// ID of the model to use. See the model endpoint compatibility table for details on which models work with the Chat API.
-    pub model: String, /* NOTE: UNUSED */
+    pub model: String,
-
+    /* NOTE: UNUSED */
     /// A list of messages comprising the conversation so far.
     #[serde(default = "default_request_messages")]
     pub messages: Vec<Message>,
@@ -346,7 +416,6 @@ pub(crate) struct ChatRequest {
     #[schema(example = "false")]
     pub logprobs: Option<bool>,
 
-    /// UNUSED
     /// An integer between 0 and 5 specifying the number of most likely tokens to return at each token position, each with
     /// an associated log probability. logprobs must be set to true if this parameter is used.
     #[serde(default)]
@@ -365,7 +434,6 @@ pub(crate) struct ChatRequest {
     #[schema(nullable = true, example = "2")]
     pub n: Option<u32>,
 
-    /// UNUSED
     /// Number between -2.0 and 2.0. Positive values penalize new tokens based on whether they appear in the text so far,
     /// increasing the model's likelihood to talk about new topics
     #[serde(default)]
@@ -447,7 +515,7 @@ pub struct PrefillToken {
     logprob: f32,
 }
 
-#[derive(Debug, Serialize, ToSchema)]
+#[derive(Debug, Serialize, ToSchema, Clone)]
 pub struct Token {
     #[schema(example = 0)]
     id: u32,

router/src/queue.rs

@@ -355,6 +355,7 @@ mod tests {
                     do_sample: false,
                     seed: 0,
                     repetition_penalty: 0.0,
+                    frequency_penalty: 0.0,
                     watermark: false,
                 },
                 stopping_parameters: StoppingCriteriaParameters {

router/src/server.rs

@@ -4,9 +4,10 @@ use crate::infer::{InferError, InferResponse, InferStreamResponse};
 use crate::validation::ValidationError;
 use crate::{
     BestOfSequence, ChatCompletion, ChatCompletionChoice, ChatCompletionChunk, ChatCompletionDelta,
-    ChatRequest, CompatGenerateRequest, Details, ErrorResponse, FinishReason, GenerateParameters,
+    ChatCompletionLogprobs, ChatRequest, CompatGenerateRequest, Details, ErrorResponse,
-    GenerateRequest, GenerateResponse, HubModelInfo, HubTokenizerConfig, Infer, Info, Message,
+    FinishReason, GenerateParameters, GenerateRequest, GenerateResponse, HubModelInfo,
-    PrefillToken, SimpleToken, StreamDetails, StreamResponse, Token, TokenizeResponse, Validation,
+    HubTokenizerConfig, Infer, Info, Message, PrefillToken, SimpleToken, StreamDetails,
+    StreamResponse, Token, TokenizeResponse, Validation,
 };
 use axum::extract::Extension;
 use axum::http::{HeaderMap, Method, StatusCode};
@@ -570,8 +571,8 @@ async fn chat_completions(
     let stream = req.stream;
     let max_new_tokens = req.max_tokens.or(Some(100));
     let repetition_penalty = req
-        .frequency_penalty
+        .presence_penalty
-        // rescale frequency_penalty from (-2.0, 2.0) to (0.0, 4.0)
+        // rescale repetition_penalty from (-2.0, 2.0) to (0.0, 4.0)
         .map(|x| x + 2.0);
     let logprobs = req.logprobs.unwrap_or(false);
     let seed = req.seed;
@@ -599,6 +600,7 @@ async fn chat_completions(
             best_of: None,
             temperature: req.temperature,
             repetition_penalty,
+            frequency_penalty: req.frequency_penalty,
             top_k: None,
             top_p: req.top_p,
             typical_p: None,
@@ -630,6 +632,10 @@ async fn chat_completions(
                 .unwrap_or_else(|_| std::time::Duration::from_secs(0))
                 .as_secs();
 
+            let logprobs = logprobs.then(|| {
+                ChatCompletionLogprobs::from((stream_token.token.clone(), stream_token.top_tokens))
+            });
+
             event
                 .json_data(ChatCompletionChunk::new(
                     model_id.clone(),
@@ -637,7 +643,7 @@ async fn chat_completions(
                     stream_token.token.text,
                     current_time,
                     stream_token.index,
-                    logprobs.then_some(stream_token.token.logprob),
+                    logprobs,
                     stream_token.details.map(|d| d.finish_reason.to_string()),
                 ))
                 .map_or_else(

router/src/validation.rs

@@ -170,6 +170,7 @@ impl Validation {
             best_of,
             temperature,
             repetition_penalty,
+            frequency_penalty,
             top_k,
             top_p,
             typical_p,
@@ -206,6 +207,11 @@ impl Validation {
             return Err(ValidationError::RepetitionPenalty);
         }
 
+        let frequency_penalty = frequency_penalty.unwrap_or(0.0);
+        if !(-2.0..=2.0).contains(&frequency_penalty) {
+            return Err(ValidationError::FrequencyPenalty);
+        }
+
         // Different because the proto default value is not a valid value
         // for the user
         let top_p = top_p
@@ -289,6 +295,7 @@ impl Validation {
         let parameters = NextTokenChooserParameters {
             temperature,
             repetition_penalty,
+            frequency_penalty,
             top_k,
             top_p,
             typical_p,
@@ -420,6 +427,8 @@ pub enum ValidationError {
     Temperature,
     #[error("`repetition_penalty` must be strictly positive")]
     RepetitionPenalty,
+    #[error("`frequency_penalty` must be >= -2.0 and <= 2.0")]
+    FrequencyPenalty,
     #[error("`top_p` must be > 0.0 and < 1.0")]
     TopP,
     #[error("`top_k` must be strictly positive")]

server/tests/utils/test_tokens.py

@@ -70,7 +70,7 @@ def test_batch_top_tokens():
 
     # Now let's make second member of the batch be speculated
     inp_logprobs = torch.tensor([[-1.0, -3.0, -4.0, -2.0, -3.0]] * 5 * 2)
-    accepted_ids[1]  = 2
+    accepted_ids[1] = 2
     topn_tok_ids, topn_tok_logprobs = batch_top_tokens(
         top_n_tokens, top_n_tokens_tensor, inp_logprobs, accepted_ids
     )

server/text_generation_server/models/__init__.py

@@ -86,6 +86,7 @@ except ImportError as e:
 if MAMBA_AVAILABLE:
     __all__.append(Mamba)
 
+
 def get_model(
     model_id: str,
     revision: Optional[str],

server/text_generation_server/models/causal_lm.py

@@ -696,14 +696,17 @@ class CausalLM(Model):
 
                 if top_n_tokens > 0:
                     all_top_tokens = []
-                    for (top_token_ids, top_token_logprobs) in zip(top_token_ids, top_token_logprobs):
+                    for (top_token_ids, top_token_logprobs) in zip(
+                        top_token_ids, top_token_logprobs
+                    ):
                         toptoken_texts = self.tokenizer.batch_decode(
                             top_token_ids,
                             clean_up_tokenization_spaces=False,
                             skip_special_tokens=False,
                         )
                         special_toptokens = [
-                            token_id in self.all_special_ids for token_id in top_token_ids
+                            token_id in self.all_special_ids
+                            for token_id in top_token_ids
                         ]
                         top_tokens = Tokens(
                             top_token_ids,

server/text_generation_server/models/custom_modeling/mamba_modeling.py

@@ -19,6 +19,7 @@ from einops import rearrange
 from causal_conv1d import causal_conv1d_fn, causal_conv1d_update
 import math
 
+
 class MambaConfig(PretrainedConfig):
     def __init__(
         self,
@@ -53,6 +54,7 @@ class MambaConfig(PretrainedConfig):
             **kwargs,
         )
 
+
 class MambaBlock(nn.Module):
     def __init__(self, prefix, config, weights):
         super().__init__()
@@ -60,10 +62,14 @@ class MambaBlock(nn.Module):
         self.in_proj = FastLinear.load(config, f"{prefix}.in_proj", weights, bias=False)
         self.x_proj = FastLinear.load(config, f"{prefix}.x_proj", weights, bias=False)
         self.dt_proj = FastLinear.load(config, f"{prefix}.dt_proj", weights, bias=True)
-        self.dt_proj_no_bias = FastLinear.load(config, f"{prefix}.dt_proj", weights, bias=False)
+        self.dt_proj_no_bias = FastLinear.load(
-        self.out_proj = FastLinear.load(config, f"{prefix}.out_proj", weights, bias=False)
+            config, f"{prefix}.dt_proj", weights, bias=False
+        )
+        self.out_proj = FastLinear.load(
+            config, f"{prefix}.out_proj", weights, bias=False
+        )
         self.conv1d = FastLinear.load(config, f"{prefix}.conv1d", weights, bias=True)
-        self.negA = -torch.exp(weights.get_tensor(f"{prefix}.A_log").float()) 
+        self.negA = -torch.exp(weights.get_tensor(f"{prefix}.A_log").float())
         self.D = weights.get_tensor(f"{prefix}.D")
         self.activation = "silu"
         self.dt_rank = config.dt_rank
@@ -80,12 +86,14 @@ class MambaBlock(nn.Module):
             out, conv_state, ssm_state = self.step(hidden_states, conv_state, ssm_state)
             return out, conv_state, ssm_state
 
-        projected_states = self.in_proj(hidden_states).transpose(1,2)
+        projected_states = self.in_proj(hidden_states).transpose(1, 2)
         x, z = projected_states.chunk(2, dim=1)
         conv_state = F.pad(x, (self.d_conv - seqlen, 0))
         x = causal_conv1d_fn(
             x=x,
-            weight=self.conv1d.weight.view(self.conv1d.weight.size(0), self.conv1d.weight.size(2)),
+            weight=self.conv1d.weight.view(
+                self.conv1d.weight.size(0), self.conv1d.weight.size(2)
+            ),
             bias=self.conv1d.bias,
             activation=self.activation,
         )
@@ -94,7 +102,9 @@ class MambaBlock(nn.Module):
         # We want dt to have d as the slowest moving dimension
         # and L as the fastest moving dimension, since those are what the ssm_scan kernel expects.
         x_dbl = self.x_proj(rearrange(x, "b d l -> (b l) d"))  # (bl d)
-        dt, B, C = torch.split(x_dbl, [self.dt_rank, self.d_state, self.d_state], dim=-1)
+        dt, B, C = torch.split(
+            x_dbl, [self.dt_rank, self.d_state, self.d_state], dim=-1
+        )
         dt = self.dt_proj.weight @ dt.t()
         dt = rearrange(dt, "d (b l) -> b d l", l=seqlen)
         B = rearrange(B, "(b l) dstate -> b dstate l", l=seqlen).contiguous()
@@ -118,28 +128,39 @@ class MambaBlock(nn.Module):
     def step(self, hidden_states, conv_state, ssm_state):
         _xz = self.in_proj(hidden_states)
         _x, _z = _xz.chunk(2, dim=-1)  # (B D)
-        conv_state_new = torch.cat([conv_state, _x.transpose(1,2)], dim=-1)
+        conv_state_new = torch.cat([conv_state, _x.transpose(1, 2)], dim=-1)
-        conv_out = causal_conv1d_fn( 
+        conv_out = causal_conv1d_fn(
-            x=conv_state_new, 
+            x=conv_state_new,
-            weight=self.conv1d.weight.view(self.conv1d.weight.size(0), self.conv1d.weight.size(2)), 
+            weight=self.conv1d.weight.view(
-            bias=self.conv1d.bias, 
+                self.conv1d.weight.size(0), self.conv1d.weight.size(2)
-            activation=self.activation
+            ),
+            bias=self.conv1d.bias,
+            activation=self.activation,
         )
         conv_state = conv_state_new[:, :, 1:]
         bsz, seqlen, dim = hidden_states.shape
         output_tensor = torch.zeros(
-            (bsz, seqlen, dim),
+            (bsz, seqlen, dim), device=hidden_states.device, dtype=hidden_states.dtype
-            device=hidden_states.device, 
-            dtype=hidden_states.dtype
         )
         for i in range(0, bsz):
-            x = conv_out[i:i+1,:,-1]
+            x = conv_out[i : i + 1, :, -1]
-            z = _z[i:i+1, -1, :]
+            z = _z[i : i + 1, -1, :]
             x_db = self.x_proj(x)
-            dt, B, C = torch.split(x_db, [self.dt_rank, self.d_state, self.d_state], dim=-1)
+            dt, B, C = torch.split(
+                x_db, [self.dt_rank, self.d_state, self.d_state], dim=-1
+            )
             dt = F.linear(dt, self.dt_proj.weight)
             y = selective_state_update(
-                ssm_state[i:i+1,:,:], x, dt, self.negA, B, C, self.D, z=z, dt_bias=self.dt_proj.bias, dt_softplus=True
+                ssm_state[i : i + 1, :, :],
+                x,
+                dt,
+                self.negA,
+                B,
+                C,
+                self.D,
+                z=z,
+                dt_bias=self.dt_proj.bias,
+                dt_softplus=True,
             )
             out = self.out_proj(y)
             output_tensor[i] = out
@@ -147,48 +168,70 @@ class MambaBlock(nn.Module):
         return output_tensor, conv_state, ssm_state
 
 
-
 class ResidualBlock(nn.Module):
     def __init__(self, layer_id, config, weights):
         super().__init__()
-        self.mamba_block = MambaBlock(prefix=f"{layer_id}.mixer", config=config, weights=weights)
+        self.mamba_block = MambaBlock(
-        self.layer_norm = FastRMSNorm.load(prefix=f"{layer_id}.norm", weights=weights, eps=config.layer_norm_epsilon)
+            prefix=f"{layer_id}.mixer", config=config, weights=weights
+        )
+        self.layer_norm = FastRMSNorm.load(
+            prefix=f"{layer_id}.norm", weights=weights, eps=config.layer_norm_epsilon
+        )
 
     def forward(
         self,
         hidden_states: torch.Tensor,
         residual: Optional[torch.Tensor] = None,
         inference_params: Optional[Any] = None,
-    ):  
+    ):
         residual = (hidden_states + residual) if residual is not None else hidden_states
         shape = residual.shape
         hidden_states, _ = self.layer_norm(residual.view(-1, shape[-1]))
-        hidden_states, conv_state, last_ssm_state = self.mamba_block(hidden_states.view(*shape), inference_params)
+        hidden_states, conv_state, last_ssm_state = self.mamba_block(
+            hidden_states.view(*shape), inference_params
+        )
         return hidden_states, residual, conv_state, last_ssm_state
 
+
 class MambaModel(nn.Module):
     def __init__(self, config, weights):
         super().__init__()
         prefix = "backbone"
         self.embed_tokens = TensorParallelEmbedding(f"{prefix}.embedding", weights)
         self.blocks = nn.ModuleList(
-            [ResidualBlock(f"{prefix}.layers.{i}", config, weights) for i in range(config.n_layer)]
+            [
+                ResidualBlock(f"{prefix}.layers.{i}", config, weights)
+                for i in range(config.n_layer)
+            ]
+        )
+        self.norm_f = FastRMSNorm.load(
+            f"{prefix}.norm_f", weights, eps=config.layer_norm_epsilon
+        )
+        self.lm_head = FastLinear.load(
+            config, f"{prefix}.embedding", weights, bias=False
         )
-        self.norm_f = FastRMSNorm.load(f"{prefix}.norm_f", weights, eps=config.layer_norm_epsilon)
-        self.lm_head = FastLinear.load(config, f"{prefix}.embedding", weights, bias=False)
         self.config = config
 
-    def forward(self, input_ids: torch.Tensor, inference_params=None, residual=None) -> Tuple[torch.Tensor, torch.Tensor, InferenceParams]:
+    def forward(
+        self, input_ids: torch.Tensor, inference_params=None, residual=None
+    ) -> Tuple[torch.Tensor, torch.Tensor, InferenceParams]:
         hidden_states = self.embed_tokens(input_ids)
         for block in self.blocks:
-            hidden_states, residual, conv_state, ssm_state = block(hidden_states, residual, inference_params)
+            hidden_states, residual, conv_state, ssm_state = block(
-            inference_params.key_value_memory_dict[block.mamba_block.layer_idx] = (conv_state, ssm_state)
+                hidden_states, residual, inference_params
+            )
+            inference_params.key_value_memory_dict[block.mamba_block.layer_idx] = (
+                conv_state,
+                ssm_state,
+            )
 
-        hidden_states = hidden_states + residual if residual is not None else hidden_states
+        hidden_states = (
+            hidden_states + residual if residual is not None else hidden_states
+        )
         hidden_states, _ = self.norm_f(hidden_states.view(-1, hidden_states.size(-1)))
         hidden_states = hidden_states.view(residual.shape)
         logits = self.lm_head(hidden_states)
 
         # update the offset for the next inference using these params
         inference_params.seqlen_offset += input_ids.size(1)
-        return logits, input_ids, inference_params
+        return logits, input_ids, inference_params

server/text_generation_server/models/flash_causal_lm.py

@@ -842,7 +842,6 @@ class FlashCausalLM(Model):
         else:
             next_token_logits = out
 
-
         speculate = get_speculate()
         (
             next_input_ids,
@@ -1064,14 +1063,17 @@ class FlashCausalLM(Model):
 
                 if top_n_tokens > 0:
                     all_top_tokens = []
-                    for (top_token_ids, top_token_logprobs) in zip(top_token_ids, top_token_logprobs):
+                    for (top_token_ids, top_token_logprobs) in zip(
+                        top_token_ids, top_token_logprobs
+                    ):
                         toptoken_texts = self.tokenizer.batch_decode(
                             top_token_ids,
                             clean_up_tokenization_spaces=False,
                             skip_special_tokens=False,
                         )
                         special_toptokens = [
-                            token_id in self.all_special_ids for token_id in top_token_ids
+                            token_id in self.all_special_ids
+                            for token_id in top_token_ids
                         ]
                         top_tokens = Tokens(
                             top_token_ids,

server/text_generation_server/models/mamba.py

@@ -26,6 +26,7 @@ from dataclasses import dataclass
 from text_generation_server.utils import NextTokenChooser, StoppingCriteria, Sampling
 from mamba_ssm.utils.generation import InferenceParams
 
+
 @dataclass
 class MambaBatch(Batch):
     batch_id: int
@@ -69,7 +70,7 @@ class MambaBatch(Batch):
             size=len(self),
             max_tokens=self.max_tokens,
         )
-    
+
     @classmethod
     def from_pb(
         cls,
@@ -196,7 +197,7 @@ class MambaBatch(Batch):
             new_padding_right_offset = max(
                 new_padding_right_offset, remaining_decode_tokens
             )
-    
+
         # Apply indices to input_ids, attention mask, past key values and other items that need to be cached
         input_ids = self.input_ids[keep_indices]
 
@@ -218,10 +219,13 @@ class MambaBatch(Batch):
         self.padding_right_offset = new_padding_right_offset
         self.max_tokens = max_tokens
 
-        # TODO 
+        # TODO
         # Kept it simple by just updating the state, maybe updating the other CPU values is necessary.
         key_value_memory_dict = {}
-        for i, (conv_state, ssm_state) in self.inference_params.key_value_memory_dict.items():
+        for i, (
+            conv_state,
+            ssm_state,
+        ) in self.inference_params.key_value_memory_dict.items():
             key_value_memory_dict[i] = (conv_state[indices], ssm_state[indices])
         self.inference_params.key_value_memory_dict = key_value_memory_dict
 
@@ -305,8 +309,9 @@ class MambaBatch(Batch):
 
             start_index = end_index
 
-
+        (_, d_model, d_conv) = (
-        (_, d_model, d_conv) = batches[0].inference_params.key_value_memory_dict[0][0].shape
+            batches[0].inference_params.key_value_memory_dict[0][0].shape
+        )
         (_, _, d_state) = batches[0].inference_params.key_value_memory_dict[0][1].shape
         n_blocks = len(batches[0].inference_params.key_value_memory_dict)
         dtype = batches[0].inference_params.key_value_memory_dict[0][0].dtype
@@ -344,9 +349,15 @@ class MambaBatch(Batch):
             for i in range(n_blocks):
                 conv_state, ssm_state = batch.inference_params.key_value_memory_dict[i]
                 batch_size = batch.inference_params.max_batch_size
-                inference_params.key_value_memory_dict[i][0][current_batch:current_batch + batch_size] = conv_state
+                inference_params.key_value_memory_dict[i][0][
-                inference_params.key_value_memory_dict[i][1][current_batch:current_batch + batch_size] = ssm_state
+                    current_batch : current_batch + batch_size
-                inference_params.lengths_per_sample[current_batch: current_batch + batch_size] = batch.inference_params.lengths_per_sample
+                ] = conv_state
+                inference_params.key_value_memory_dict[i][1][
+                    current_batch : current_batch + batch_size
+                ] = ssm_state
+                inference_params.lengths_per_sample[
+                    current_batch : current_batch + batch_size
+                ] = batch.inference_params.lengths_per_sample
             current_batch += batch_size
 
         return cls(
@@ -366,12 +377,13 @@ class MambaBatch(Batch):
             padding_right_offset=padding_right_offset,
             keys_head_dim_last=batches[0].keys_head_dim_last,
             max_tokens=max_tokens,
-            inference_params=inference_params
+            inference_params=inference_params,
         )
 
     def __len__(self):
         return len(self.requests)
 
+
 class Mamba(Model):
     def __init__(
         self,
@@ -428,7 +440,7 @@ class Mamba(Model):
     def warmup(self, batch) -> Optional[int]:
         # TODO: implement warmup for Mamba if needed
         return None
-    
+
     def forward(
         self,
         input_ids: torch.Tensor,
@@ -441,7 +453,9 @@ class Mamba(Model):
 
     def generate_token(self, batch) -> Tuple[List[Any], Optional[Any], Tuple[int, int]]:
         start = time.time_ns()
-        input_ids = batch.input_ids # batch.past_input_ids if batch.past_input_ids is not None else batch.input_ids
+        input_ids = (
+            batch.input_ids
+        )  # batch.past_input_ids if batch.past_input_ids is not None else batch.input_ids
 
         batch_size = input_ids.shape[0]
         max_seqlen = input_ids.shape[1]
@@ -450,8 +464,11 @@ class Mamba(Model):
         # Inference params
         seqlen_og = 0
         inf_cache = {}
-        lengths_per_sample = torch.ones(batch_size, dtype=torch.int32, device=input_ids.device) * max_seqlen
+        lengths_per_sample = (
-        
+            torch.ones(batch_size, dtype=torch.int32, device=input_ids.device)
+            * max_seqlen
+        )
+
         if batch.inference_params is None:
             inference_params = InferenceParams(
                 max_seqlen=max_seqlen,
@@ -478,11 +495,16 @@ class Mamba(Model):
                     device=block.dt_proj.weight.device,
                     dtype=block.dt_proj.weight.dtype,
                 )
-                inference_params.key_value_memory_dict[block.layer_idx] = (conv_state, ssm_state)
+                inference_params.key_value_memory_dict[block.layer_idx] = (
+                    conv_state,
+                    ssm_state,
+                )
             batch.inference_params = inference_params
-        
+
         # Forward pass
-        logits, past_input_ids, new_inference_params = self.model(input_ids, batch.inference_params)
+        logits, past_input_ids, new_inference_params = self.model(
+            input_ids, batch.inference_params
+        )
 
         batch.inference_params = new_inference_params
         # Results
@@ -564,7 +586,8 @@ class Mamba(Model):
                         prefix_offset=len(all_input_ids)
                         - stopping_criteria.current_tokens
                         - 1,
-                        read_offset=len(all_input_ids) - stopping_criteria.current_tokens,
+                        read_offset=len(all_input_ids)
+                        - stopping_criteria.current_tokens,
                         skip_special_tokens=True,
                     )
                     # Get seed

server/text_generation_server/models/seq2seq_lm.py

@@ -750,14 +750,17 @@ class Seq2SeqLM(Model):
 
                 if top_n_tokens > 0:
                     all_top_tokens = []
-                    for (top_token_ids, top_token_logprobs) in zip(top_token_ids, top_token_logprobs):
+                    for (top_token_ids, top_token_logprobs) in zip(
+                        top_token_ids, top_token_logprobs
+                    ):
                         toptoken_texts = self.tokenizer.batch_decode(
                             top_token_ids,
                             clean_up_tokenization_spaces=False,
                             skip_special_tokens=False,
                         )
                         special_toptokens = [
-                            token_id in self.all_special_ids for token_id in top_token_ids
+                            token_id in self.all_special_ids
+                            for token_id in top_token_ids
                         ]
                         top_tokens = Tokens(
                             top_token_ids,

server/text_generation_server/models/types.py

@@ -95,5 +95,7 @@ class Generation:
             generated_text=self.generated_text.to_pb()
             if self.generated_text is not None
             else None,
-            top_tokens=[top_tokens.to_pb() for top_tokens in self.top_tokens] if self.top_tokens is not None else None,
+            top_tokens=[top_tokens.to_pb() for top_tokens in self.top_tokens]
+            if self.top_tokens is not None
+            else None,
         )

server/text_generation_server/utils/logits_process.py

@@ -118,6 +118,62 @@ class HeterogeneousRepetitionPenaltyLogitsProcessor(LogitsProcessor):
         return None
 
 
+class FrequencyPenaltyLogitsProcessor(LogitsProcessor):
+    r"""
+    Frequency penalty as defined by OpenAI
+
+    Args:
+        penalty (`float`):
+            The parameter for frequency penalty. 0.0 means no penalty.
+    """
+
+    def __init__(self, penalty: float):
+        self.penalty = penalty
+
+    def __call__(
+        self, input_ids: torch.LongTensor, scores: torch.FloatTensor
+    ) -> torch.FloatTensor:
+        score = torch.gather(scores, 1, input_ids)
+        # if score < 0 then penalty has to be multiplied to reduce the previous token probability
+        score = -torch.where(
+            score < 0, score * self.penalty, score / self.penalty
+        )
+
+        return scores.scatter_add_(1, input_ids, score)
+
+
+class HeterogeneousFrequencyPenaltyLogitsProcessor(LogitsProcessor):
+    r"""
+    Frequency penalty as defined by OpenAI
+
+    Args:
+        frequency_penalty (`List[float]`):
+            The parameter for frequency penalty. 0.0 means no penalty.
+    """
+
+    def __init__(self, penalty: List[float], dtype: torch.dtype, device: torch.device):
+        self.penalty = penalty
+        self.penalty_tensor = torch.tensor(
+            penalty, dtype=dtype, device=device
+        ).unsqueeze(1)
+
+    def __call__(self, input_ids: torch.Tensor, scores: torch.Tensor) -> torch.Tensor:
+        score = torch.gather(scores, 1, input_ids)
+        # if score < 0 then penalty has to be multiplied to reduce the previous token probability
+        score = -torch.where(
+            score < 0, score * self.penalty_tensor, score / self.penalty_tensor
+        )
+
+        return scores.scatter_add_(1, input_ids, score)
+
+    def filter(self, indices):
+        self.penalty = [self.penalty[i] for i in indices]
+        if any([x != 0.0 for x in self.penalty]):
+            self.penalty_tensor = self.penalty_tensor[indices]
+            return self
+        return None
+
+
 class HeterogeneousTemperatureLogitsWarper:
     r"""
     [`LogitsWarper`] for temperature (exponential scaling output probability distribution).

server/text_generation_server/utils/tokens.py

@@ -1,12 +1,14 @@
 import re
-from typing import Callable, List, Optional, Tuple
+from typing import List, Optional, Tuple
 
 import torch
 from text_generation_server.pb import generate_pb2
 from text_generation_server.pb.generate_pb2 import FinishReason
 from text_generation_server.utils.logits_process import (
+    FrequencyPenaltyLogitsProcessor,
     HeterogeneousProcessorWrapper,
     HeterogeneousRepetitionPenaltyLogitsProcessor,
+    HeterogeneousFrequencyPenaltyLogitsProcessor,
     HeterogeneousTemperatureLogitsWarper,
     HeterogeneousTopKLogitsWarper,
     HeterogeneousTopPLogitsWarper,
@@ -23,6 +25,7 @@ class NextTokenChooser:
         watermark=False,
         temperature=1.0,
         repetition_penalty=1.0,
+        frequency_penalty=0.0,
         top_k=None,
         top_p=None,
         typical_p=None,
@@ -35,7 +38,12 @@ class NextTokenChooser:
         )
         self.repetition_processor = (
             RepetitionPenaltyLogitsProcessor(penalty=repetition_penalty)
-            if repetition_penalty
+            if repetition_penalty and repetition_penalty != 1.0
+            else None
+        )
+        self.frequency_processor = (
+            FrequencyPenaltyLogitsProcessor(penalty=frequency_penalty)
+            if frequency_penalty and frequency_penalty != 0.0
             else None
         )
 
@@ -60,6 +68,8 @@ class NextTokenChooser:
             scores = self.watermark_processor(input_ids, scores)
         if self.repetition_processor is not None:
             scores = self.repetition_processor(input_ids, scores)
+        if self.frequency_processor is not None:
+            scores = self.frequency_processor(input_ids, scores)
 
         if self.static_warper is None:
             next_logprob = torch.log_softmax(scores, -1)
@@ -80,6 +90,7 @@ class NextTokenChooser:
             watermark=pb.watermark,
             temperature=pb.temperature,
             repetition_penalty=pb.repetition_penalty,
+            frequency_penalty=pb.frequency_penalty,
             top_k=pb.top_k,
             top_p=pb.top_p,
             typical_p=pb.typical_p,
@@ -184,6 +195,7 @@ class HeterogeneousNextTokenChooser:
         watermark: List[bool],
         temperature: List[float],
         repetition_penalty: List[float],
+        frequency_penalty: List[float],
         top_k: List[int],
         top_p: List[float],
         typical_p: List[float],
@@ -212,6 +224,14 @@ class HeterogeneousNextTokenChooser:
             else None
         )
 
+        self.frequency_processor = (
+            HeterogeneousFrequencyPenaltyLogitsProcessor(
+                frequency_penalty, dtype, device
+            )
+            if any([x != 0.0 for x in frequency_penalty])
+            else None
+        )
+
         if any([x != 1.0 for x in temperature]):
             do_sample = [
                 sample or x != 1.0 for x, sample in zip(temperature, do_sample)
@@ -269,6 +289,8 @@ class HeterogeneousNextTokenChooser:
                 _scores = self.watermark_processor(input_ids, _scores)
             if self.repetition_processor is not None:
                 _scores = self.repetition_processor(input_ids, _scores)
+            if self.frequency_processor is not None:
+                _scores = self.frequency_processor(input_ids, _scores)
 
             for warper in self.warpers:
                 _scores = warper(input_ids, _scores)
@@ -316,7 +338,6 @@ class HeterogeneousNextTokenChooser:
 
         next_logprobs = torch.gather(logprobs, 1, next_ids.view(-1, 1)).view(-1)
 
-
         if speculate > 0:
             if speculative_scores is not None:
                 # Medusa provided some scores
@@ -338,6 +359,9 @@ class HeterogeneousNextTokenChooser:
         if self.repetition_processor is not None:
             self.repetition_processor = self.repetition_processor.filter(indices)
 
+        if self.frequency_processor is not None:
+            self.frequency_processor = self.frequency_processor.filter(indices)
+
         filtered_warpers = []
         for warper in self.warpers:
             filtered_warper = warper.filter(indices)
@@ -366,6 +390,7 @@ class HeterogeneousNextTokenChooser:
             watermark=[pb_.watermark for pb_ in pb],
             temperature=[pb_.temperature for pb_ in pb],
             repetition_penalty=[pb_.repetition_penalty for pb_ in pb],
+            frequency_penalty=[pb_.frequency_penalty for pb_ in pb],
             top_k=[pb_.top_k for pb_ in pb],
             top_p=[pb_.top_p for pb_ in pb],
             typical_p=[pb_.typical_p for pb_ in pb],
@@ -438,7 +463,10 @@ class HeterogeneousSampling:
 
 
 def batch_top_tokens(
-    top_n_tokens: List[int], top_n_tokens_tensor: torch.Tensor, logprobs: torch.Tensor, accepted_ids: torch.Tensor
+    top_n_tokens: List[int],
+    top_n_tokens_tensor: torch.Tensor,
+    logprobs: torch.Tensor,
+    accepted_ids: torch.Tensor,
 ) -> Tuple[List[List[List[int]]], List[List[List[float]]]]:
     """Find the top n most likely tokens for a batch of generations.
 
@@ -450,12 +478,15 @@ def batch_top_tokens(
     if max_top_n == 0:
         return [[[]]] * len(top_n_tokens), [[[]]] * len(top_n_tokens)
 
-
     batch_size = accepted_ids.shape[0]
     speculate_size = logprobs.shape[0] // batch_size
     top_n_tokens_tensor = top_n_tokens_tensor.repeat_interleave(speculate_size)
     # Ensure top_n doesn't exceed vocab size
-    top_n_tokens = [min(tok, logprobs.size(-1)) for tok in top_n_tokens for _ in range(speculate_size)]
+    top_n_tokens = [
+        min(tok, logprobs.size(-1))
+        for tok in top_n_tokens
+        for _ in range(speculate_size)
+    ]
 
     # Parallel kthvalue adapted from https://discuss.pytorch.org/t/how-to-efficiently-get-the-k-th-largest-values-in-parallel/160529/2
     # Sorted topk is faster than torch.sort() since we only need a small subset
@@ -484,10 +515,10 @@ def batch_top_tokens(
     for i, n_accepted_ids in enumerate(accepted_ids_list):
         start = speculate_size * i
         stop = speculate_size * (i + 1)
-        _top_indices = top_indices[start: stop]
+        _top_indices = top_indices[start:stop]
-        _top_values = top_values[start: stop]
+        _top_values = top_values[start:stop]
-        _top_n_ishes = top_n_ishes[start: stop]
+        _top_n_ishes = top_n_ishes[start:stop]
-        _top_n_tokens = top_n_tokens[start: stop]
+        _top_n_tokens = top_n_tokens[start:stop]
 
         _top_indices = _top_indices[:n_accepted_ids]
         _top_values = _top_values[:n_accepted_ids]
@@ -497,7 +528,9 @@ def batch_top_tokens(
         row_top_token_ids = []
         row_top_token_logprobs = []
 
-        for idxs, vals, n, req_n in zip(_top_indices, _top_values, _top_n_ishes, _top_n_tokens):
+        for idxs, vals, n, req_n in zip(
+            _top_indices, _top_values, _top_n_ishes, _top_n_tokens
+        ):
             indices = idxs[:n] if req_n > 0 else []
             values = vals[:n] if req_n > 0 else []