Codex Local Token Counter Fix - Instructional Guide for llama.cpp #18847

jbulger82 · 2026-01-14T20:53:54Z

jbulger82
Jan 14, 2026

Codex Local Token Counter Fix - Instructional Guide

Author: Claude (Anthropic) with Jeff Bulger
Date: 2026-01-14
For: OpenAI Codex TUI (v0.80.0) running with local llama.cpp backend
GitHub: Feel free to use and share - MIT License

🎯 What This Fixes

The stock OpenAI Codex TUI doesn't display token usage when using local llama.cpp. After this fix:

╭───────────────────────────────────────────────────────────────╮
│  Token usage:      34.5K total  (33K input + 1.45K output)    │
│  Context window:   93% left (28.2K used / 249K)               │
╰───────────────────────────────────────────────────────────────╯

Before: Always shows 0 total (0 input + 0 output)
After: Real-time token tracking + auto-compact when you hit your limit

📋 Prerequisites

OpenAI Codex TUI (Rust version, v0.80.0 or similar)
llama.cpp server running locally
Using the Chat wire API (not Responses API)
Rust toolchain for rebuilding

🔧 The Fix (4 Files to Modify)

File 1: `codex-api/src/requests/chat.rs`

What: Tell llama.cpp to include usage data in the streaming response.

Find the payload construction (around line 303-310):

let payload = json!({
    "model": self.model,
    "messages": messages,
    "stream": true,
    "tools": self.tools,
});

Replace with:

let payload = json!({
    "model": self.model,
    "messages": messages,
    "stream": true,
    "stream_options": {"include_usage": true},
    "tools": self.tools,
});

File 2: `codex-api/src/sse/chat.rs`

What: Extract the token usage from llama.cpp's response and pass it to the TUI.

Step 2a: Add Import

At the top of the file, add:

use codex_protocol::protocol::TokenUsage;

Step 2b: Add Tracking Variable

In the spawn_stream_handler function, find these lines:

let mut assistant_item: Option<ResponseItem> = None;
let mut reasoning_item: Option<ResponseItem> = None;
let mut completed_sent = false;

Add below them:

let mut captured_usage: Option<TokenUsage> = None;

Step 2c: Modify `flush_and_complete` Function

Find the flush_and_complete function and change it to accept and use token_usage:

async fn flush_and_complete(
    tx_event: &mpsc::Sender<Result<ResponseEvent, ApiError>>,
    reasoning_item: &mut Option<ResponseItem>,
    assistant_item: &mut Option<ResponseItem>,
    token_usage: Option<TokenUsage>,  // NEW PARAMETER
) {
    if let Some(reasoning) = reasoning_item.take() {
        let _ = tx_event
            .send(Ok(ResponseEvent::OutputItemDone(reasoning)))
            .await;
    }

    if let Some(assistant) = assistant_item.take() {
        let _ = tx_event
            .send(Ok(ResponseEvent::OutputItemDone(assistant)))
            .await;
    }

    let _ = tx_event
        .send(Ok(ResponseEvent::Completed {
            response_id: String::new(),
            token_usage,  // WAS: None - NOW: actual usage
        }))
        .await;
}

Step 2d: Extract Usage from JSON

After the JSON is parsed (let value: serde_json::Value = serde_json::from_str(&data)?;), add:

// Extract usage from final chunk if present
if let Some(usage) = value.get("usage") {
    if let (Some(prompt), Some(completion), Some(total)) = (
        usage.get("prompt_tokens").and_then(|v| v.as_i64()),
        usage.get("completion_tokens").and_then(|v| v.as_i64()),
        usage.get("total_tokens").and_then(|v| v.as_i64()),
    ) {
        captured_usage = Some(TokenUsage {
            input_tokens: prompt,
            output_tokens: completion,
            total_tokens: total,
            cached_input_tokens: 0,
            reasoning_output_tokens: 0,
        });
    }
}

Step 2e: Update All `flush_and_complete` Calls

Find all calls to flush_and_complete and add the captured_usage parameter:

Before:

flush_and_complete(&tx_event, &mut reasoning_item, &mut assistant_item).await;

After:

flush_and_complete(&tx_event, &mut reasoning_item, &mut assistant_item, captured_usage.clone()).await;

File 3: `core/src/models_manager/model_info.rs`

What: Set the correct context window for your model so the percentage display is accurate.

Find your model's definition (or add one for local models). Example for Nemotron:

} else if slug.contains("nemotron") || slug.contains("Nemotron") || slug.contains("30b") {
    model_info!(
        slug,
        base_instructions: NEMOTRON_INSTRUCTIONS.to_string(),
        apply_patch_tool_type: Some(ApplyPatchToolType::Function),
        context_window: Some(262_144),  // MATCH YOUR llama.cpp -c VALUE
        shell_type: ConfigShellToolType::ShellCommand,
    )
}

Important: Set context_window to match your llama.cpp server's -c flag!

File 4: `config.toml`

What: Enable auto-compaction to prevent OOM when you hit your context limit.

Add at the top level of your config:

model_auto_compact_token_limit = 250000

This triggers automatic context compaction at 250k tokens (adjust based on your context window).

🔨 Build & Test

cd codex-rs
cargo build

Then launch your TUI and run /status - you should see real token counts!

🔍 How It Works

┌─────────────────────────────────────────────────────────────────┐
│  Your Prompt                                                    │
│       ↓                                                         │
│  Codex TUI builds request with stream_options: {include_usage}  │
│       ↓                                                         │
│  llama.cpp processes and streams response                       │
│       ↓                                                         │
│  Final SSE chunk includes: {"usage": {prompt_tokens, ...}}      │
│       ↓                                                         │
│  Our fix extracts usage → TokenUsage struct                     │
│       ↓                                                         │
│  TUI displays: "Token usage: 34.5K total (33K in + 1.45K out)"  │
└─────────────────────────────────────────────────────────────────┘

⚠️ Troubleshooting

Still showing 0 tokens?

Make sure you rebuilt after changes: cargo build
Verify you're using Chat wire API (check wire_api in config)
Check llama.cpp logs - should show POST /v1/chat/completions

Context percentage wrong?

Update context_window in model_info.rs to match your llama.cpp -c value

Auto-compact not triggering?

Check model_auto_compact_token_limit is set in config.toml
Value should be slightly below your context window

📊 Token Count Breakdown (For Reference)

Typical baseline overhead:

Component	Tokens	Notes
System prompt	1,400-5,200	Depends on prompt file
Tools (MCP)	2,000-3,000	Depends on tools loaded
Grammar overhead	1,200-1,500	llama.cpp peg-constructed
Request formatting	500-700	JSON structure

Optimal baseline: ~5,600 tokens with lightweight prompt
Heavy baseline: ~9,800 tokens with full prompt + features

🙏 Credits

Fix developed by Claude (Anthropic) working with Jeff Bulger on 2026-01-14.

This was a multi-hour debugging session tracing token flow from llama.cpp → SSE parser → TUI. The key insight: the Chat wire API (for local models) never implemented usage extraction, even though Responses API (cloud) had it.

The Problem: Hardcoded token_usage: None in the completion handler
The Fix: Extract usage from llama.cpp's JSON response and pass it through

📜 License

MIT - Use freely, attribution appreciated.

"When Nemo runs well, he runs GREAT. This config makes it happen." - Jeff

In the codex benchmark 12gb vram!

Codex Benchmark (12GB VRAM)

Rank	Ctx	b	ub	Prompt tok/s	Gen tok/s	Overall tok/s	Notes (only what’s visible in your logs)
1	393,216	3072	6144	763.00	26.82	243.80	-fa on, --cache-type-k/v q4_0, -kvu, --parallel 1
2	393,216	3072	6144	757.87	27.11	318.94	same config as above (different run)
3	393,216	3072	3072	704.56	30.31	355.13	-fa on, --cache-type-k/v q4_0, -kvu, --parallel 1
4	393,216	512	4198	691.22	29.58	296.39	--cache-ram 8198, --context-shift, -kvu, --cache-type-k/v q4_0
5	393,216	2048*	(not shown)	656.12	28.27	240.01	*batch inferred from batch.n_tokens = 2048 steps; ub not shown
6	393,216	2048	3072	638.23	30.73	227.10	-fa on, --cache-type-k/v q4_0, -kvu, --parallel 1
7	393,216	1024	3076	541.93	33.56	335.13	-fa auto, --context-shift, -kvu, --cache-type-k/v q4_0
8	1,048,576	1000	4000 → 1000	420.47	27.34	204.38	-fa auto, --context-shift, -kvu, --cache-type-k/v q4_0

llama log from a small run for those interested...I can run this model at a million context as well on 12gvram thanks to llama.cpp

jeff@jeff-STGAUBRON 10:20 ~/Desktop/Test_llama_build/test_llamaRoc.cpp
$ cd /home/jeff/Desktop/Test_llama_build/test_llamaRoc.cpp && export HSA_OVERRIDE_GFX_VERSION=10.3.0 && export HIP_VISIBLE_DEVICES=0 && export HSA_ENABLE_SDMA=0 && export HSA_DISABLE_FRAGMENT_ALLOCATOR=1 && export LD_LIBRARY_PATH="/home/jeff/Desktop/Test_llama_build/rocm-local/opt/rocm-7.1.1/lib:/home/jeff/Desktop/Test_llama_build/rocm-local/opt/rocm-7.1.1/lib64${LD_LIBRARY_PATH:+:$LD_LIBRARY_PATH}" && MODEL="/home/jeff/Desktop/models/Nemotron-3-Nano-30B-A3B-IQ4_NL.gguf" && ./build-rocm-b7622-hipcc/bin/llama-server -m "$MODEL" -c 262144 -b 1625 -ub 20000 --cache-type-k q4_0 --cache-type-v q4_0 -fa on
ggml_cuda_init: found 1 ROCm devices:
Device 0: AMD Radeon RX 6700 XT, gfx1030 (0x1030), VMM: no, Wave Size: 32
main: n_parallel is set to auto, using n_parallel = 4 and kv_unified = true
build: 7622 (c69c7eb) with Clang 20.0.0 for Linux x86_64
system info: n_threads = 8, n_threads_batch = 8, total_threads = 16

init: using 15 threads for HTTP server
start: binding port with default address family
main: loading model
srv load_model: loading model '/home/jeff/Desktop/models/Nemotron-3-Nano-30B-A3B-IQ4_NL.gguf'
common_init_result: fitting params to device memory, for bugs during this step try to reproduce them with -fit off, or provide --verbose logs if the bug only occurs with -fit on
llama_params_fit_impl: projected to use 20221 MiB of device memory vs. 11844 MiB of free device memory
llama_params_fit_impl: cannot fulfill margin of 1024 MiB, need to reduce device memory by 9401 MiB
llama_params_fit_impl: context size set by user to 262144 -> no change
llama_params_fit_impl: with only dense weights in device memory there is a total surplus of 5985 MiB
llama_params_fit_impl: filling dense-only layers back-to-front:
llama_params_fit_impl: - ROCm0 (AMD Radeon RX 6700 XT): 53 layers, 4866 MiB used, 6977 MiB free
llama_params_fit_impl: converting dense-only layers to full layers and filling them front-to-back with overflow to next device/system memory:
llama_params_fit_impl: - ROCm0 (AMD Radeon RX 6700 XT): 53 layers (31 overflowing), 10611 MiB used, 1232 MiB free
llama_params_fit: successfully fit params to free device memory
llama_params_fit: fitting params to free memory took 2.46 seconds
llama_model_load_from_file_impl: using device ROCm0 (AMD Radeon RX 6700 XT) (0000:03:00.0) - 12068 MiB free
llama_model_loader: loaded meta data with 53 key-value pairs and 401 tensors from /home/jeff/Desktop/models/Nemotron-3-Nano-30B-A3B-IQ4_NL.gguf (version GGUF V3 (latest))
llama_model_loader: Dumping metadata keys/values. Note: KV overrides do not apply in this output.
llama_model_loader: - kv 0: general.architecture str = nemotron_h_moe
llama_model_loader: - kv 1: general.type str = model
llama_model_loader: - kv 2: general.sampling.top_p f32 = 1.000000
llama_model_loader: - kv 3: general.sampling.temp f32 = 1.000000
llama_model_loader: - kv 4: general.name str = Nemotron-3-Nano-30B-A3B
llama_model_loader: - kv 5: general.basename str = Nemotron-3-Nano-30B-A3B
llama_model_loader: - kv 6: general.quantized_by str = Unsloth
llama_model_loader: - kv 7: general.size_label str = 30B-A3B
llama_model_loader: - kv 8: general.repo_url str = https://huggingface.co/unsloth
llama_model_loader: - kv 9: nemotron_h_moe.block_count u32 = 52
llama_model_loader: - kv 10: nemotron_h_moe.context_length u32 = 1048576
llama_model_loader: - kv 11: nemotron_h_moe.embedding_length u32 = 2688
llama_model_loader: - kv 12: nemotron_h_moe.feed_forward_length arr[i32,52] = [0, 1856, 0, 1856, 0, 0, 1856, 0, 185...
llama_model_loader: - kv 13: nemotron_h_moe.attention.head_count u32 = 32
llama_model_loader: - kv 14: nemotron_h_moe.attention.head_count_kv arr[i32,52] = [0, 0, 0, 0, 0, 2, 0, 0, 0, 0, 0, 0, ...
llama_model_loader: - kv 15: nemotron_h_moe.rope.freq_base f32 = 10000.000000
llama_model_loader: - kv 16: nemotron_h_moe.attention.layer_norm_rms_epsilon f32 = 0.000010
llama_model_loader: - kv 17: nemotron_h_moe.attention.layer_norm_epsilon f32 = 0.000010
llama_model_loader: - kv 18: nemotron_h_moe.expert_used_count u32 = 6
llama_model_loader: - kv 19: nemotron_h_moe.expert_group_count u32 = 1
llama_model_loader: - kv 20: nemotron_h_moe.expert_group_used_count u32 = 1
llama_model_loader: - kv 21: nemotron_h_moe.vocab_size u32 = 131072
llama_model_loader: - kv 22: nemotron_h_moe.rope.dimension_count u32 = 84
llama_model_loader: - kv 23: nemotron_h_moe.ssm.conv_kernel u32 = 4
llama_model_loader: - kv 24: nemotron_h_moe.ssm.state_size u32 = 128
llama_model_loader: - kv 25: nemotron_h_moe.ssm.group_count u32 = 8
llama_model_loader: - kv 26: nemotron_h_moe.ssm.inner_size u32 = 4096
llama_model_loader: - kv 27: nemotron_h_moe.ssm.time_step_rank u32 = 64
llama_model_loader: - kv 28: nemotron_h_moe.rope.scaling.finetuned bool = false
llama_model_loader: - kv 29: nemotron_h_moe.attention.key_length u32 = 128
llama_model_loader: - kv 30: nemotron_h_moe.attention.value_length u32 = 128
llama_model_loader: - kv 31: nemotron_h_moe.expert_feed_forward_length u32 = 1856
llama_model_loader: - kv 32: nemotron_h_moe.expert_shared_feed_forward_length u32 = 3712
llama_model_loader: - kv 33: nemotron_h_moe.expert_count u32 = 128
llama_model_loader: - kv 34: nemotron_h_moe.expert_shared_count u32 = 1
llama_model_loader: - kv 35: nemotron_h_moe.expert_weights_norm bool = true
llama_model_loader: - kv 36: nemotron_h_moe.expert_weights_scale f32 = 2.500000
llama_model_loader: - kv 37: tokenizer.ggml.model str = gpt2
llama_model_loader: - kv 38: tokenizer.ggml.pre str = pixtral
llama_model_loader: - kv 39: tokenizer.ggml.tokens arr[str,131072] = ["", "~~", "~~", "[INST]", "[...
llama_model_loader: - kv 40: tokenizer.ggml.token_type arr[i32,131072] = [3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, ...
llama_model_loader: - kv 41: tokenizer.ggml.merges arr[str,269443] = ["Ġ Ġ", "Ġ t", "e r", "i n", "Ġ �...
llama_model_loader: - kv 42: tokenizer.ggml.bos_token_id u32 = 1
llama_model_loader: - kv 43: tokenizer.ggml.eos_token_id u32 = 11
llama_model_loader: - kv 44: tokenizer.ggml.unknown_token_id u32 = 0
llama_model_loader: - kv 45: tokenizer.ggml.padding_token_id u32 = 999
llama_model_loader: - kv 46: tokenizer.chat_template str = {# Unsloth template fixes #}\n{% macro...
llama_model_loader: - kv 47: general.quantization_version u32 = 2
llama_model_loader: - kv 48: general.file_type u32 = 25
llama_model_loader: - kv 49: quantize.imatrix.file str = Nemotron-3-Nano-30B-A3B-GGUF/imatrix_...
llama_model_loader: - kv 50: quantize.imatrix.dataset str = unsloth_calibration_Nemotron-3-Nano-3...
llama_model_loader: - kv 51: quantize.imatrix.entries_count u32 = 185
llama_model_loader: - kv 52: quantize.imatrix.chunks_count u32 = 80
llama_model_loader: - type f32: 237 tensors
llama_model_loader: - type q5_0: 1 tensors
llama_model_loader: - type q5_1: 23 tensors
llama_model_loader: - type q8_0: 24 tensors
llama_model_loader: - type iq4_nl: 116 tensors
print_info: file format = GGUF V3 (latest)
print_info: file type = IQ4_NL - 4.5 bpw
print_info: file size = 16.92 GiB (4.60 BPW)
load: 0 unused tokens
load: printing all EOG tokens:
load: - 11 ('<|im_end|>')
load: special tokens cache size = 1000
load: token to piece cache size = 0.8499 MB
print_info: arch = nemotron_h_moe
print_info: vocab_only = 0
print_info: no_alloc = 0
print_info: n_ctx_train = 1048576
print_info: n_embd = 2688
print_info: n_embd_inp = 2688
print_info: n_layer = 52
print_info: n_head = 32
print_info: n_head_kv = [0, 0, 0, 0, 0, 2, 0, 0, 0, 0, 0, 0, 2, 0, 0, 0, 0, 0, 0, 2, 0, 0, 0, 0, 0, 0, 2, 0, 0, 0, 0, 0, 0, 2, 0, 0, 0, 0, 0, 0, 0, 0, 2, 0, 0, 0, 0, 0, 0, 0, 0, 0]
print_info: n_rot = 84
print_info: n_swa = 0
print_info: is_swa_any = 0
print_info: n_embd_head_k = 128
print_info: n_embd_head_v = 128
print_info: n_gqa = [0, 0, 0, 0, 0, 16, 0, 0, 0, 0, 0, 0, 16, 0, 0, 0, 0, 0, 0, 16, 0, 0, 0, 0, 0, 0, 16, 0, 0, 0, 0, 0, 0, 16, 0, 0, 0, 0, 0, 0, 0, 0, 16, 0, 0, 0, 0, 0, 0, 0, 0, 0]
print_info: n_embd_k_gqa = [0, 0, 0, 0, 0, 256, 0, 0, 0, 0, 0, 0, 256, 0, 0, 0, 0, 0, 0, 256, 0, 0, 0, 0, 0, 0, 256, 0, 0, 0, 0, 0, 0, 256, 0, 0, 0, 0, 0, 0, 0, 0, 256, 0, 0, 0, 0, 0, 0, 0, 0, 0]
print_info: n_embd_v_gqa = [0, 0, 0, 0, 0, 256, 0, 0, 0, 0, 0, 0, 256, 0, 0, 0, 0, 0, 0, 256, 0, 0, 0, 0, 0, 0, 256, 0, 0, 0, 0, 0, 0, 256, 0, 0, 0, 0, 0, 0, 0, 0, 256, 0, 0, 0, 0, 0, 0, 0, 0, 0]
print_info: f_norm_eps = 0.0e+00
print_info: f_norm_rms_eps = 1.0e-05
print_info: f_clamp_kqv = 0.0e+00
print_info: f_max_alibi_bias = 0.0e+00
print_info: f_logit_scale = 0.0e+00
print_info: f_attn_scale = 0.0e+00
print_info: n_ff = [0, 1856, 0, 1856, 0, 0, 1856, 0, 1856, 0, 1856, 0, 0, 1856, 0, 1856, 0, 1856, 0, 0, 1856, 0, 1856, 0, 1856, 0, 0, 1856, 0, 1856, 0, 1856, 0, 0, 1856, 0, 1856, 0, 1856, 0, 1856, 0, 0, 1856, 0, 1856, 0, 1856, 0, 1856, 0, 1856]
print_info: n_expert = 128
print_info: n_expert_used = 6
print_info: n_expert_groups = 1
print_info: n_group_used = 1
print_info: causal attn = 1
print_info: pooling type = 0
print_info: rope type = -1
print_info: rope scaling = linear
print_info: freq_base_train = 10000.0
print_info: freq_scale_train = 1
print_info: n_ctx_orig_yarn = 1048576
print_info: rope_yarn_log_mul= 0.0000
print_info: rope_finetuned = unknown
print_info: ssm_d_conv = 4
print_info: ssm_d_inner = 4096
print_info: ssm_d_state = 128
print_info: ssm_dt_rank = 64
print_info: ssm_n_group = 8
print_info: ssm_dt_b_c_rms = 0
print_info: model type = 31B.A3.5B
print_info: model params = 31.58 B
print_info: general.name = Nemotron-3-Nano-30B-A3B
print_info: f_embedding_scale = 0.000000
print_info: f_residual_scale = 0.000000
print_info: f_attention_scale = 0.000000
print_info: n_ff_shexp = 3712
print_info: vocab type = BPE
print_info: n_vocab = 131072
print_info: n_merges = 269443
print_info: BOS token = 1 ''
print_info: EOS token = 11 '<|im_end|>'
print_info: EOT token = 11 '<|im_end|>'
print_info: UNK token = 0 ''
print_info: PAD token = 999 '<SPECIAL_999>'
print_info: LF token = 1010 'Ċ'
print_info: EOG token = 11 '<|im_end|>'
print_info: max token length = 150
load_tensors: loading model tensors, this can take a while... (mmap = true)
load_tensors: offloading output layer to GPU
load_tensors: offloading 51 repeating layers to GPU
load_tensors: offloaded 53/53 layers to GPU
load_tensors: CPU_Mapped model buffer size = 16965.10 MiB
load_tensors: ROCm0 model buffer size = 7487.97 MiB
.......................................................
common_init_result: added <|im_end|> logit bias = -inf
llama_context: constructing llama_context
llama_context: n_seq_max = 4
llama_context: n_ctx = 262144
llama_context: n_ctx_seq = 262144
llama_context: n_batch = 1625
llama_context: n_ubatch = 1625
llama_context: causal_attn = 1
llama_context: flash_attn = enabled
llama_context: kv_unified = true
llama_context: freq_base = 10000.0
llama_context: freq_scale = 1
llama_context: n_ctx_seq (262144) < n_ctx_train (1048576) -- the full capacity of the model will not be utilized
llama_context: ROCm_Host output buffer size = 2.00 MiB
llama_kv_cache: ROCm0 KV buffer size = 432.00 MiB
llama_kv_cache: size = 432.00 MiB (262144 cells, 6 layers, 4/1 seqs), K (q4_0): 216.00 MiB, V (q4_0): 216.00 MiB
llama_memory_recurrent: ROCm0 RS buffer size = 190.47 MiB
llama_memory_recurrent: size = 190.47 MiB ( 4 cells, 52 layers, 4 seqs), R (f32): 6.47 MiB, S (f32): 184.00 MiB
llama_context: ROCm0 compute buffer size = 2501.13 MiB
llama_context: ROCm_Host compute buffer size = 1644.73 MiB
llama_context: graph nodes = 2188
llama_context: graph splits = 33 (with bs=1625), 34 (with bs=1)
common_init_from_params: warming up the model with an empty run - please wait ... (--no-warmup to disable)
srv load_model: initializing slots, n_slots = 4
slot load_model: id 0 | task -1 | new slot, n_ctx = 262144
slot load_model: id 1 | task -1 | new slot, n_ctx = 262144
slot load_model: id 2 | task -1 | new slot, n_ctx = 262144
slot load_model: id 3 | task -1 | new slot, n_ctx = 262144
srv load_model: prompt cache is enabled, size limit: 8192 MiB
srv load_model: use --cache-ram 0 to disable the prompt cache
srv load_model: for more info see https://github.com//pull/16391
srv load_model: thinking = 1
load_model: chat template, chat_template: {# Unsloth template fixes #}
{% macro render_extra_keys(json_dict, handled_keys) %}
{%- if json_dict is mapping %}
{%- for json_key in json_dict if json_key not in handled_keys %}
{%- if json_dict[json_key] is mapping or (json_dict[json_key] is sequence and json_dict[json_key] is not string) %}
{{- '\n<' ~ json_key ~ '>' ~ (json_dict[json_key] | tojson | safe) ~ '</' ~ json_key ~ '>' }}
{%- else %}
{{-'\n<' ~ json_key ~ '>' ~ (json_dict[json_key] | string) ~ '</' ~ json_key ~ '>' }}
{%- endif %}
{%- endfor %}
{%- endif %}
{% endmacro %}
{%- set enable_thinking = enable_thinking if enable_thinking is defined else True %}
{%- set truncate_history_thinking = truncate_history_thinking if truncate_history_thinking is defined else True %}

{%- set ns = namespace(last_user_idx = -1) %}
{%- set loop_messages = messages %}
{%- for m in loop_messages %}
{%- if m["role"] == "user" %}
{%- set ns.last_user_idx = loop.index0 %}
{%- endif %}
{%- endfor %}

{%- if messages[0]["role"] == "system" %}
{%- set system_message = messages[0]["content"] %}
{%- set loop_messages = messages[1:] %}
{%- else %}
{%- set system_message = "" %}
{%- set loop_messages = messages %}
{%- endif %}
{%- if not tools is defined %}
{%- set tools = [] %}
{%- endif %}
{# Recompute last_user_idx relative to loop_messages after handling system #}
{%- set ns = namespace(last_user_idx = -1) %}
{%- for m in loop_messages %}
{%- if m["role"] == "user" %}
{%- set ns.last_user_idx = loop.index0 %}
{%- endif %}
{%- endfor %}
{%- if system_message is defined %}
{{- "<|im_start|>system\n" + system_message }}
{%- else %}
{%- if tools is iterable and tools | length > 0 %}
{{- "<|im_start|>system\n" }}
{%- endif %}
{%- endif %}
{%- if tools is iterable and tools | length > 0 %}
{%- if system_message is defined and system_message | length > 0 %}
{{- "\n\n" }}
{%- endif %}
{{- "# Tools\n\nYou have access to the following functions:\n\n" }}
{{- "" }}
{%- for tool in tools %}
{%- if tool.function is defined %}
{%- set tool = tool.function %}
{%- endif %}
{{- "\n\n" ~ tool.name ~ "" }}
{%- if tool.description is defined %}
{{- '\n' ~ (tool.description | trim) ~ '' }}
{%- endif %}
{{- '\n' }}
{%- if tool.parameters is defined and tool.parameters is mapping and tool.parameters.properties is defined and tool.parameters.properties is mapping %}
{%- for param_name, param_fields in tool.parameters.properties|items %}
{{- '\n' }}
{{- '\n' ~ param_name ~ '' }}
{%- if param_fields.type is defined %}
{{- '\n' ~ (param_fields.type | string) ~ '' }}
{%- endif %}
{%- if param_fields.description is defined %}
{{- '\n' ~ (param_fields.description | trim) ~ '' }}
{%- endif %}
{%- if param_fields.enum is defined %}
{{- '\n' ~ (param_fields.enum | tojson | safe) ~ '' }}
{%- endif %}
{%- set handled_keys = ['name', 'type', 'description', 'enum'] %}
{{- render_extra_keys(param_fields, handled_keys) }}
{{- '\n' }}
{%- endfor %}
{%- endif %}
{% set handled_keys = ['type', 'properties', 'required'] %}
{{- render_extra_keys(tool.parameters, handled_keys) }}
{%- if tool.parameters is defined and tool.parameters.required is defined %}
{{- '\n' ~ (tool.parameters.required | tojson | safe) ~ '' }}
{%- endif %}
{{- '\n' }}
{%- set handled_keys = ['type', 'name', 'description', 'parameters'] %}
{{- render_extra_keys(tool, handled_keys) }}
{{- '\n' }}
{%- endfor %}
{{- "\n" }}

{{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}

{%- endif %}

{%- if system_message is defined %}
{{- '<|im_end|>\n' }}
{%- else %}
{%- if tools is iterable and tools | length > 0 %}
{{- '<|im_end|>\n' }}
{%- endif %}
{%- endif %}

{%- for message in loop_messages %}
{%- if message.role == "assistant" %}
{# Add reasoning content in to content field for unified processing below. #}
{%- if message.reasoning_content is defined and message.reasoning_content is string and message.reasoning_content | trim | length > 0 %}
{%- set content = "\n" ~ message.reasoning_content ~ "\n\n" ~ (message.content | default('', true)) %}
{%- else %}
{%- set content = message.content | default('', true) %}
{%- if content is string -%}
{# Allow downstream logic to to take care of broken thought, only handle coherent reasoning here. #}
{%- if '' not in content and '' not in content -%}
{%- set content = "" ~ content -%}
{%- endif -%}
{%- else -%}
{%- set content = content -%}
{%- endif -%}
{%- endif %}
{%- if message.tool_calls is defined and message.tool_calls is iterable and message.tool_calls | length > 0 %}
{# Assistant message has tool calls. #}
{{- '<|im_start|>assistant\n' }}
{%- set include_content = not (truncate_history_thinking and loop.index0 < ns.last_user_idx) %}
{%- if content is string and content | trim | length > 0 %}
{%- if include_content %}
{{- (content | trim) ~ '\n' -}}
{%- else %}
{%- set c = (content | string) %}
{%- if '' in c %}
{# Keep only content after the last closing think. Also generation prompt causes this. #}
{%- set c = (c.split('')|last) %}
{%- elif '' in c %}
{# If was opened but never closed, drop the trailing think segment #}
{%- set c = (c.split('')|first) %}
{%- endif %}
{%- set c = "" ~ c | trim %}
{%- if c | length > 0 %}
{{- c ~ '\n' -}}
{%- endif %}
{%- endif %}
{%- else %}
{{- "" -}}
{%- endif %}
{%- for tool_call in message.tool_calls %}
{%- if tool_call.function is defined %}
{%- set tool_call = tool_call.function %}
{%- endif %}
{{- '<tool_call>\n<function=' ~ tool_call.name ~ '>\n' -}}
{%- if tool_call.arguments is defined %}{%- if tool_call.arguments is mapping %}
{%- for args_name, args_value in tool_call.arguments|items %}
{{- '<parameter=' ~ args_name ~ '>\n' -}}
{%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
{{- args_value ~ '\n\n' -}}
{%- endfor %}{%- endif %}
{%- endif %}
{{- '\n</tool_call>\n' -}}
{%- endfor %}
{{- '<|im_end|>\n' }}
{%- else %}
{# Assistant message doesn't have tool calls. #}
{%- if not (truncate_history_thinking and loop.index0 < ns.last_user_idx) %}
{{- '<|im_start|>assistant\n' ~ (content | default('', true) | string | trim) ~ '<|im_end|>\n' }}
{%- else %}
{%- set c = (content | default('', true) | string) %}
{%- if '' in c and '' in c %}
{%- set c = "" ~ (c.split('')|last) %}
{%- endif %}
{%- set c = c | trim %}
{%- if c | length > 0 %}
{{- '<|im_start|>assistant\n' ~ c ~ '<|im_end|>\n' }}
{%- else %}
{{- '<|im_start|>assistant\n<|im_end|>\n' }}
{%- endif %}
{%- endif %}
{%- endif %}
{%- elif message.role == "user" or message.role == "system" %}
{{- '<|im_start|>' + message.role + '\n' }}
{%- set content = message.content | string %}
{{- content }}
{{- '<|im_end|>\n' }}
{%- elif message.role == "tool" %}
{%- if loop.previtem and loop.previtem.role != "tool" %}
{{- '<|im_start|>user\n' }}
{%- endif %}
{{- '<tool_response>\n' }}
{{- message.content }}
{{- '\n</tool_response>\n' }}
{%- if not loop.last and loop.nextitem.role != "tool" %}
{{- '<|im_end|>\n' }}
{%- elif loop.last %}
{{- '<|im_end|>\n' }}
{%- endif %}
{%- else %}
{{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>\n' }}
{%- endif %}
{%- endfor %}

{%- if add_generation_prompt %}
{%- if enable_thinking %}
{{- '<|im_start|>assistant\n\n' }}
{%- else %}
{{- '<|im_start|>assistant\n' }}
{%- endif %}
{%- endif %}
{# Copyright 2025-present Unsloth. Apache 2.0 License. #}, example_format: '<|im_start|>system
You are a helpful assistant<|im_end|>
<|im_start|>user
Hello<|im_end|>
<|im_start|>assistant
Hi there<|im_end|>
<|im_start|>user
How are you?<|im_end|>
<|im_start|>assistant

'
main: model loaded
main: server is listening on http://127.0.0.1:8080
main: starting the main loop...
srv update_slots: all slots are idle
srv log_server_r: request: GET /api/tags 127.0.0.1 200
srv update_slots: all slots are idle
srv log_server_r: request: GET /slots 127.0.0.1 200
srv params_from_: Chat format: peg-constructed
slot get_availabl: id 3 | task -1 | selected slot by LRU, t_last = -1
slot launch_slot_: id 3 | task -1 | sampler chain: logits -> penalties -> dry -> top-n-sigma -> top-k -> typical -> top-p -> min-p -> xtc -> temp-ext -> dist
slot launch_slot_: id 3 | task 1 | processing task
slot update_slots: id 3 | task 1 | new prompt, n_ctx_slot = 262144, n_keep = 0, task.n_tokens = 9972
slot update_slots: id 3 | task 1 | n_tokens = 0, memory_seq_rm [0, end)
slot update_slots: id 3 | task 1 | prompt processing progress, n_tokens = 1625, batch.n_tokens = 1625, progress = 0.162956
slot update_slots: id 3 | task 1 | n_tokens = 1625, memory_seq_rm [1625, end)
slot update_slots: id 3 | task 1 | prompt processing progress, n_tokens = 3250, batch.n_tokens = 1625, progress = 0.325913
slot update_slots: id 3 | task 1 | n_tokens = 3250, memory_seq_rm [3250, end)
slot update_slots: id 3 | task 1 | prompt processing progress, n_tokens = 4875, batch.n_tokens = 1625, progress = 0.488869
srv stop: cancel task, id_task = 1
srv log_server_r: request: POST /v1/chat/completions 127.0.0.1 200
slot release: id 3 | task 1 | stop processing: n_tokens = 4875, truncated = 0
srv update_slots: all slots are idle
srv log_server_r: request: GET /api/tags 127.0.0.1 200
srv update_slots: all slots are idle
srv log_server_r: request: GET /slots 127.0.0.1 200
srv params_from_: Chat format: peg-constructed
slot get_availabl: id 2 | task -1 | selected slot by LRU, t_last = -1
slot launch_slot_: id 2 | task -1 | sampler chain: logits -> penalties -> dry -> top-n-sigma -> top-k -> typical -> top-p -> min-p -> xtc -> temp-ext -> dist
slot launch_slot_: id 2 | task 7 | processing task
slot update_slots: id 2 | task 7 | new prompt, n_ctx_slot = 262144, n_keep = 0, task.n_tokens = 5605
slot update_slots: id 2 | task 7 | n_tokens = 0, memory_seq_rm [0, end)
slot update_slots: id 2 | task 7 | prompt processing progress, n_tokens = 1625, batch.n_tokens = 1625, progress = 0.289920
slot update_slots: id 2 | task 7 | n_tokens = 1625, memory_seq_rm [1625, end)
slot update_slots: id 2 | task 7 | prompt processing progress, n_tokens = 3250, batch.n_tokens = 1625, progress = 0.579839
slot update_slots: id 2 | task 7 | n_tokens = 3250, memory_seq_rm [3250, end)
slot update_slots: id 2 | task 7 | prompt processing progress, n_tokens = 4875, batch.n_tokens = 1625, progress = 0.869759
slot update_slots: id 2 | task 7 | n_tokens = 4875, memory_seq_rm [4875, end)
slot update_slots: id 2 | task 7 | prompt processing progress, n_tokens = 5541, batch.n_tokens = 666, progress = 0.988582
slot update_slots: id 2 | task 7 | n_tokens = 5541, memory_seq_rm [5541, end)
slot update_slots: id 2 | task 7 | prompt processing progress, n_tokens = 5605, batch.n_tokens = 64, progress = 1.000000
slot update_slots: id 2 | task 7 | prompt done, n_tokens = 5605, batch.n_tokens = 64
slot update_slots: id 2 | task 7 | created context checkpoint 1 of 8 (pos_min = 5540, pos_max = 5540, size = 47.618 MiB)
slot print_timing: id 2 | task 7 |
prompt eval time = 8854.74 ms / 5605 tokens ( 1.58 ms per token, 632.99 tokens per second)
eval time = 19107.85 ms / 673 tokens ( 28.39 ms per token, 35.22 tokens per second)
total time = 27962.59 ms / 6278 tokens
slot release: id 2 | task 7 | stop processing: n_tokens = 6277, truncated = 0
srv update_slots: all slots are idle
srv log_server_r: request: POST /v1/chat/completions 127.0.0.1 200
srv params_from_: Chat format: peg-constructed
slot get_availabl: id 2 | task -1 | selected slot by LCP similarity, sim_best = 0.982 (> 0.100 thold), f_keep = 0.893
slot launch_slot_: id 2 | task -1 | sampler chain: logits -> penalties -> dry -> top-n-sigma -> top-k -> typical -> top-p -> min-p -> xtc -> temp-ext -> dist
slot launch_slot_: id 2 | task 685 | processing task
slot update_slots: id 2 | task 685 | new prompt, n_ctx_slot = 262144, n_keep = 0, task.n_tokens = 5706
slot update_slots: id 2 | task 685 | n_past = 5604, slot.prompt.tokens.size() = 6277, seq_id = 2, pos_min = 6276, n_swa = 1
slot update_slots: id 2 | task 685 | restored context checkpoint (pos_min = 5540, pos_max = 5540, size = 47.618 MiB)
slot update_slots: id 2 | task 685 | n_tokens = 5541, memory_seq_rm [5541, end)
slot update_slots: id 2 | task 685 | prompt processing progress, n_tokens = 5642, batch.n_tokens = 101, progress = 0.988784
slot update_slots: id 2 | task 685 | n_tokens = 5642, memory_seq_rm [5642, end)
slot update_slots: id 2 | task 685 | prompt processing progress, n_tokens = 5706, batch.n_tokens = 64, progress = 1.000000
slot update_slots: id 2 | task 685 | prompt done, n_tokens = 5706, batch.n_tokens = 64
slot update_slots: id 2 | task 685 | created context checkpoint 2 of 8 (pos_min = 5641, pos_max = 5641, size = 47.618 MiB)
slot print_timing: id 2 | task 685 |
prompt eval time = 1701.31 ms / 165 tokens ( 10.31 ms per token, 96.98 tokens per second)
eval time = 5458.47 ms / 193 tokens ( 28.28 ms per token, 35.36 tokens per second)
total time = 7159.78 ms / 358 tokens
slot release: id 2 | task 685 | stop processing: n_tokens = 5898, truncated = 0
srv update_slots: all slots are idle
srv log_server_r: request: POST /v1/chat/completions 127.0.0.1 200
srv params_from_: Chat format: peg-constructed
slot get_availabl: id 2 | task -1 | selected slot by LCP similarity, sim_best = 0.349 (> 0.100 thold), f_keep = 0.967
slot launch_slot_: id 2 | task -1 | sampler chain: logits -> penalties -> dry -> top-n-sigma -> top-k -> typical -> top-p -> min-p -> xtc -> temp-ext -> dist
slot launch_slot_: id 2 | task 880 | processing task
slot update_slots: id 2 | task 880 | new prompt, n_ctx_slot = 262144, n_keep = 0, task.n_tokens = 16333
slot update_slots: id 2 | task 880 | n_past = 5705, slot.prompt.tokens.size() = 5898, seq_id = 2, pos_min = 5897, n_swa = 1
slot update_slots: id 2 | task 880 | restored context checkpoint (pos_min = 5641, pos_max = 5641, size = 47.618 MiB)
slot update_slots: id 2 | task 880 | n_tokens = 5642, memory_seq_rm [5642, end)
slot update_slots: id 2 | task 880 | prompt processing progress, n_tokens = 7267, batch.n_tokens = 1625, progress = 0.444927
slot update_slots: id 2 | task 880 | n_tokens = 7267, memory_seq_rm [7267, end)
slot update_slots: id 2 | task 880 | prompt processing progress, n_tokens = 8892, batch.n_tokens = 1625, progress = 0.544419
slot update_slots: id 2 | task 880 | n_tokens = 8892, memory_seq_rm [8892, end)
slot update_slots: id 2 | task 880 | prompt processing progress, n_tokens = 10517, batch.n_tokens = 1625, progress = 0.643911
slot update_slots: id 2 | task 880 | n_tokens = 10517, memory_seq_rm [10517, end)
slot update_slots: id 2 | task 880 | prompt processing progress, n_tokens = 12142, batch.n_tokens = 1625, progress = 0.743403
slot update_slots: id 2 | task 880 | n_tokens = 12142, memory_seq_rm [12142, end)
slot update_slots: id 2 | task 880 | prompt processing progress, n_tokens = 13767, batch.n_tokens = 1625, progress = 0.842895
slot update_slots: id 2 | task 880 | n_tokens = 13767, memory_seq_rm [13767, end)
slot update_slots: id 2 | task 880 | prompt processing progress, n_tokens = 15392, batch.n_tokens = 1625, progress = 0.942387
slot update_slots: id 2 | task 880 | n_tokens = 15392, memory_seq_rm [15392, end)
slot update_slots: id 2 | task 880 | prompt processing progress, n_tokens = 16269, batch.n_tokens = 877, progress = 0.996082
slot update_slots: id 2 | task 880 | n_tokens = 16269, memory_seq_rm [16269, end)
slot update_slots: id 2 | task 880 | prompt processing progress, n_tokens = 16333, batch.n_tokens = 64, progress = 1.000000
slot update_slots: id 2 | task 880 | prompt done, n_tokens = 16333, batch.n_tokens = 64
slot update_slots: id 2 | task 880 | created context checkpoint 3 of 8 (pos_min = 16268, pos_max = 16268, size = 47.618 MiB)
slot print_timing: id 2 | task 880 |
prompt eval time = 16780.02 ms / 10691 tokens ( 1.57 ms per token, 637.13 tokens per second)
eval time = 2609.34 ms / 89 tokens ( 29.32 ms per token, 34.11 tokens per second)
total time = 19389.35 ms / 10780 tokens
slot release: id 2 | task 880 | stop processing: n_tokens = 16421, truncated = 0
srv update_slots: all slots are idle
srv log_server_r: request: POST /v1/chat/completions 127.0.0.1 200
srv params_from_: Chat format: peg-constructed
slot get_availabl: id 2 | task -1 | selected slot by LCP similarity, sim_best = 0.607 (> 0.100 thold), f_keep = 0.995
slot launch_slot_: id 2 | task -1 | sampler chain: logits -> penalties -> dry -> top-n-sigma -> top-k -> typical -> top-p -> min-p -> xtc -> temp-ext -> dist
slot launch_slot_: id 2 | task 977 | processing task
slot update_slots: id 2 | task 977 | new prompt, n_ctx_slot = 262144, n_keep = 0, task.n_tokens = 26911
slot update_slots: id 2 | task 977 | n_past = 16332, slot.prompt.tokens.size() = 16421, seq_id = 2, pos_min = 16420, n_swa = 1
slot update_slots: id 2 | task 977 | restored context checkpoint (pos_min = 16268, pos_max = 16268, size = 47.618 MiB)
slot update_slots: id 2 | task 977 | n_tokens = 16269, memory_seq_rm [16269, end)
slot update_slots: id 2 | task 977 | prompt processing progress, n_tokens = 17894, batch.n_tokens = 1625, progress = 0.664933
slot update_slots: id 2 | task 977 | n_tokens = 17894, memory_seq_rm [17894, end)
slot update_slots: id 2 | task 977 | prompt processing progress, n_tokens = 19519, batch.n_tokens = 1625, progress = 0.725317
slot update_slots: id 2 | task 977 | n_tokens = 19519, memory_seq_rm [19519, end)
slot update_slots: id 2 | task 977 | prompt processing progress, n_tokens = 21144, batch.n_tokens = 1625, progress = 0.785701
slot update_slots: id 2 | task 977 | n_tokens = 21144, memory_seq_rm [21144, end)
slot update_slots: id 2 | task 977 | prompt processing progress, n_tokens = 22769, batch.n_tokens = 1625, progress = 0.846085
slot update_slots: id 2 | task 977 | n_tokens = 22769, memory_seq_rm [22769, end)
slot update_slots: id 2 | task 977 | prompt processing progress, n_tokens = 24394, batch.n_tokens = 1625, progress = 0.906469
slot update_slots: id 2 | task 977 | n_tokens = 24394, memory_seq_rm [24394, end)
slot update_slots: id 2 | task 977 | prompt processing progress, n_tokens = 26019, batch.n_tokens = 1625, progress = 0.966854
slot update_slots: id 2 | task 977 | n_tokens = 26019, memory_seq_rm [26019, end)
slot update_slots: id 2 | task 977 | prompt processing progress, n_tokens = 26847, batch.n_tokens = 828, progress = 0.997622
slot update_slots: id 2 | task 977 | n_tokens = 26847, memory_seq_rm [26847, end)
slot update_slots: id 2 | task 977 | prompt processing progress, n_tokens = 26911, batch.n_tokens = 64, progress = 1.000000
slot update_slots: id 2 | task 977 | prompt done, n_tokens = 26911, batch.n_tokens = 64
slot update_slots: id 2 | task 977 | created context checkpoint 4 of 8 (pos_min = 26846, pos_max = 26846, size = 47.618 MiB)
slot print_timing: id 2 | task 977 |
prompt eval time = 18209.44 ms / 10642 tokens ( 1.71 ms per token, 584.42 tokens per second)
eval time = 12526.74 ms / 406 tokens ( 30.85 ms per token, 32.41 tokens per second)
total time = 30736.18 ms / 11048 tokens
slot release: id 2 | task 977 | stop processing: n_tokens = 27316, truncated = 0
srv update_slots: all slots are idle
srv log_server_r: request: POST /v1/chat/completions 127.0.0.1 200
srv params_from_: Chat format: peg-constructed
slot get_availabl: id 2 | task -1 | selected slot by LCP similarity, sim_best = 1.000 (> 0.100 thold), f_keep = 0.985
slot launch_slot_: id 2 | task -1 | sampler chain: logits -> penalties -> dry -> top-n-sigma -> top-k -> typical -> top-p -> min-p -> xtc -> temp-ext -> dist
slot launch_slot_: id 2 | task 1391 | processing task
slot update_slots: id 2 | task 1391 | new prompt, n_ctx_slot = 262144, n_keep = 0, task.n_tokens = 26919
slot update_slots: id 2 | task 1391 | n_past = 26906, slot.prompt.tokens.size() = 27316, seq_id = 2, pos_min = 27315, n_swa = 1
slot update_slots: id 2 | task 1391 | restored context checkpoint (pos_min = 26846, pos_max = 26846, size = 47.618 MiB)
slot update_slots: id 2 | task 1391 | n_tokens = 26847, memory_seq_rm [26847, end)
slot update_slots: id 2 | task 1391 | prompt processing progress, n_tokens = 26855, batch.n_tokens = 8, progress = 0.997622
slot update_slots: id 2 | task 1391 | n_tokens = 26855, memory_seq_rm [26855, end)
slot update_slots: id 2 | task 1391 | prompt processing progress, n_tokens = 26919, batch.n_tokens = 64, progress = 1.000000
slot update_slots: id 2 | task 1391 | prompt done, n_tokens = 26919, batch.n_tokens = 64
slot print_timing: id 2 | task 1391 |
prompt eval time = 995.73 ms / 72 tokens ( 13.83 ms per token, 72.31 tokens per second)
eval time = 30936.83 ms / 1004 tokens ( 30.81 ms per token, 32.45 tokens per second)
total time = 31932.55 ms / 1076 tokens
slot release: id 2 | task 1391 | stop processing: n_tokens = 27922, truncated = 0
srv update_slots: all slots are idle
srv log_server_r: request: POST /v1/chat/completions 127.0.0.1 200
srv params_from_: Chat format: peg-constructed
slot get_availabl: id 2 | task -1 | selected slot by LCP similarity, sim_best = 0.982 (> 0.100 thold), f_keep = 0.964
slot launch_slot_: id 2 | task -1 | sampler chain: logits -> penalties -> dry -> top-n-sigma -> top-k -> typical -> top-p -> min-p -> xtc -> temp-ext -> dist
slot launch_slot_: id 2 | task 2397 | processing task
slot update_slots: id 2 | task 2397 | new prompt, n_ctx_slot = 262144, n_keep = 0, task.n_tokens = 27412
slot update_slots: id 2 | task 2397 | n_past = 26918, slot.prompt.tokens.size() = 27922, seq_id = 2, pos_min = 27921, n_swa = 1
slot update_slots: id 2 | task 2397 | restored context checkpoint (pos_min = 26846, pos_max = 26846, size = 47.618 MiB)
slot update_slots: id 2 | task 2397 | n_tokens = 26847, memory_seq_rm [26847, end)
slot update_slots: id 2 | task 2397 | prompt processing progress, n_tokens = 27348, batch.n_tokens = 501, progress = 0.997665
slot update_slots: id 2 | task 2397 | n_tokens = 27348, memory_seq_rm [27348, end)
slot update_slots: id 2 | task 2397 | prompt processing progress, n_tokens = 27412, batch.n_tokens = 64, progress = 1.000000
slot update_slots: id 2 | task 2397 | prompt done, n_tokens = 27412, batch.n_tokens = 64
slot update_slots: id 2 | task 2397 | created context checkpoint 5 of 8 (pos_min = 27347, pos_max = 27347, size = 47.618 MiB)
slot print_timing: id 2 | task 2397 |
prompt eval time = 2606.55 ms / 565 tokens ( 4.61 ms per token, 216.76 tokens per second)
eval time = 23907.75 ms / 773 tokens ( 30.93 ms per token, 32.33 tokens per second)
total time = 26514.30 ms / 1338 tokens
slot release: id 2 | task 2397 | stop processing: n_tokens = 28184, truncated = 0
srv update_slots: all slots are idle
srv log_server_r: request: POST /v1/chat/completions 127.0.0.1 200
srv params_from_: Chat format: peg-constructed
slot get_availabl: id 2 | task -1 | selected slot by LCP similarity, sim_best = 0.997 (> 0.100 thold), f_keep = 0.973
slot launch_slot_: id 2 | task -1 | sampler chain: logits -> penalties -> dry -> top-n-sigma -> top-k -> typical -> top-p -> min-p -> xtc -> temp-ext -> dist
slot launch_slot_: id 2 | task 3172 | processing task
slot update_slots: id 2 | task 3172 | new prompt, n_ctx_slot = 262144, n_keep = 0, task.n_tokens = 27480
slot update_slots: id 2 | task 3172 | n_past = 27411, slot.prompt.tokens.size() = 28184, seq_id = 2, pos_min = 28183, n_swa = 1
slot update_slots: id 2 | task 3172 | restored context checkpoint (pos_min = 27347, pos_max = 27347, size = 47.618 MiB)
slot update_slots: id 2 | task 3172 | n_tokens = 27348, memory_seq_rm [27348, end)
slot update_slots: id 2 | task 3172 | prompt processing progress, n_tokens = 27416, batch.n_tokens = 68, progress = 0.997671
slot update_slots: id 2 | task 3172 | n_tokens = 27416, memory_seq_rm [27416, end)
slot update_slots: id 2 | task 3172 | prompt processing progress, n_tokens = 27480, batch.n_tokens = 64, progress = 1.000000
slot update_slots: id 2 | task 3172 | prompt done, n_tokens = 27480, batch.n_tokens = 64
slot update_slots: id 2 | task 3172 | created context checkpoint 6 of 8 (pos_min = 27415, pos_max = 27415, size = 47.618 MiB)
slot print_timing: id 2 | task 3172 |
prompt eval time = 1770.92 ms / 132 tokens ( 13.42 ms per token, 74.54 tokens per second)
eval time = 101896.73 ms / 3191 tokens ( 31.93 ms per token, 31.32 tokens per second)
total time = 103667.65 ms / 3323 tokens
slot release: id 2 | task 3172 | stop processing: n_tokens = 30670, truncated = 0
srv update_slots: all slots are idle
srv log_server_r: request: POST /v1/chat/completions 127.0.0.1 200
srv params_from_: Chat format: peg-constructed
slot get_availabl: id 2 | task -1 | selected slot by LCP similarity, sim_best = 0.993 (> 0.100 thold), f_keep = 0.896
slot launch_slot_: id 2 | task -1 | sampler chain: logits -> penalties -> dry -> top-n-sigma -> top-k -> typical -> top-p -> min-p -> xtc -> temp-ext -> dist
slot launch_slot_: id 2 | task 6365 | processing task
slot update_slots: id 2 | task 6365 | new prompt, n_ctx_slot = 262144, n_keep = 0, task.n_tokens = 27680
slot update_slots: id 2 | task 6365 | n_past = 27479, slot.prompt.tokens.size() = 30670, seq_id = 2, pos_min = 30669, n_swa = 1
slot update_slots: id 2 | task 6365 | restored context checkpoint (pos_min = 27415, pos_max = 27415, size = 47.618 MiB)
slot update_slots: id 2 | task 6365 | n_tokens = 27416, memory_seq_rm [27416, end)
slot update_slots: id 2 | task 6365 | prompt processing progress, n_tokens = 27616, batch.n_tokens = 200, progress = 0.997688
slot update_slots: id 2 | task 6365 | n_tokens = 27616, memory_seq_rm [27616, end)
slot update_slots: id 2 | task 6365 | prompt processing progress, n_tokens = 27680, batch.n_tokens = 64, progress = 1.000000
slot update_slots: id 2 | task 6365 | prompt done, n_tokens = 27680, batch.n_tokens = 64
slot update_slots: id 2 | task 6365 | created context checkpoint 7 of 8 (pos_min = 27615, pos_max = 27615, size = 47.618 MiB)
slot print_timing: id 2 | task 6365 |
prompt eval time = 2179.08 ms / 264 tokens ( 8.25 ms per token, 121.15 tokens per second)
eval time = 50388.43 ms / 1625 tokens ( 31.01 ms per token, 32.25 tokens per second)
total time = 52567.51 ms / 1889 tokens
slot release: id 2 | task 6365 | stop processing: n_tokens = 29304, truncated = 0
srv update_slots: all slots are idle
srv log_server_r: request: POST /v1/chat/completions 127.0.0.1 200
srv params_from_: Chat format: peg-constructed
slot get_availabl: id 2 | task -1 | selected slot by LCP similarity, sim_best = 0.986 (> 0.100 thold), f_keep = 0.945
slot launch_slot_: id 2 | task -1 | sampler chain: logits -> penalties -> dry -> top-n-sigma -> top-k -> typical -> top-p -> min-p -> xtc -> temp-ext -> dist
slot launch_slot_: id 2 | task 7992 | processing task
slot update_slots: id 2 | task 7992 | new prompt, n_ctx_slot = 262144, n_keep = 0, task.n_tokens = 28062
slot update_slots: id 2 | task 7992 | n_past = 27679, slot.prompt.tokens.size() = 29304, seq_id = 2, pos_min = 29303, n_swa = 1
slot update_slots: id 2 | task 7992 | restored context checkpoint (pos_min = 27615, pos_max = 27615, size = 47.618 MiB)
slot update_slots: id 2 | task 7992 | n_tokens = 27616, memory_seq_rm [27616, end)
slot update_slots: id 2 | task 7992 | prompt processing progress, n_tokens = 27998, batch.n_tokens = 382, progress = 0.997719
slot update_slots: id 2 | task 7992 | n_tokens = 27998, memory_seq_rm [27998, end)
slot update_slots: id 2 | task 7992 | prompt processing progress, n_tokens = 28062, batch.n_tokens = 64, progress = 1.000000
slot update_slots: id 2 | task 7992 | prompt done, n_tokens = 28062, batch.n_tokens = 64
slot update_slots: id 2 | task 7992 | created context checkpoint 8 of 8 (pos_min = 27997, pos_max = 27997, size = 47.618 MiB)
slot print_timing: id 2 | task 7992 |
prompt eval time = 2489.96 ms / 446 tokens ( 5.58 ms per token, 179.12 tokens per second)
eval time = 20526.31 ms / 662 tokens ( 31.01 ms per token, 32.25 tokens per second)
total time = 23016.27 ms / 1108 tokens
slot release: id 2 | task 7992 | stop processing: n_tokens = 28723, truncated = 0
srv update_slots: all slots are idle
srv log_server_r: request: POST /v1/chat/completions 127.0.0.1 200
srv params_from_: Chat format: peg-constructed
slot get_availabl: id 2 | task -1 | selected slot by LCP similarity, sim_best = 0.981 (> 0.100 thold), f_keep = 0.977
slot launch_slot_: id 2 | task -1 | sampler chain: logits -> penalties -> dry -> top-n-sigma -> top-k -> typical -> top-p -> min-p -> xtc -> temp-ext -> dist
slot launch_slot_: id 2 | task 8656 | processing task
slot update_slots: id 2 | task 8656 | new prompt, n_ctx_slot = 262144, n_keep = 0, task.n_tokens = 28599
slot update_slots: id 2 | task 8656 | n_past = 28061, slot.prompt.tokens.size() = 28723, seq_id = 2, pos_min = 28722, n_swa = 1
slot update_slots: id 2 | task 8656 | restored context checkpoint (pos_min = 27997, pos_max = 27997, size = 47.618 MiB)
slot update_slots: id 2 | task 8656 | n_tokens = 27998, memory_seq_rm [27998, end)
slot update_slots: id 2 | task 8656 | prompt processing progress, n_tokens = 28535, batch.n_tokens = 537, progress = 0.997762
slot update_slots: id 2 | task 8656 | n_tokens = 28535, memory_seq_rm [28535, end)
slot update_slots: id 2 | task 8656 | prompt processing progress, n_tokens = 28599, batch.n_tokens = 64, progress = 1.000000
slot update_slots: id 2 | task 8656 | prompt done, n_tokens = 28599, batch.n_tokens = 64
slot update_slots: id 2 | task 8656 | erasing old context checkpoint (pos_min = 5540, pos_max = 5540, size = 47.618 MiB)
slot update_slots: id 2 | task 8656 | created context checkpoint 8 of 8 (pos_min = 28534, pos_max = 28534, size = 47.618 MiB)
slot print_timing: id 2 | task 8656 |
prompt eval time = 2603.42 ms / 601 tokens ( 4.33 ms per token, 230.85 tokens per second)
eval time = 21206.75 ms / 684 tokens ( 31.00 ms per token, 32.25 tokens per second)
total time = 23810.17 ms / 1285 tokens
slot release: id 2 | task 8656 | stop processing: n_tokens = 29282, truncated = 0
srv update_slots: all slots are idle
srv log_server_r: request: POST /v1/chat/completions 127.0.0.1 200
srv log_server_r: request: GET /api/tags 127.0.0.1 200
srv update_slots: all slots are idle
srv log_server_r: request: GET /slots 127.0.0.1 200
srv params_from_: Chat format: peg-constructed
slot get_availabl: id 2 | task -1 | selected slot by LCP similarity, sim_best = 0.988 (> 0.100 thold), f_keep = 0.191
srv get_availabl: updating prompt cache
srv prompt_save: - saving prompt with length 29282, total state size = 96.208 MiB
srv load: - looking for better prompt, base f_keep = 0.191, sim = 0.988
srv update: - cache state: 1 prompts, 477.150 MiB (limits: 8192.000 MiB, 262144 tokens, 502731 est)
srv update: - prompt 0x2ab563c0: 29282 tokens, checkpoints: 8, 477.150 MiB
srv get_availabl: prompt cache update took 170.31 ms
slot launch_slot_: id 2 | task -1 | sampler chain: logits -> penalties -> dry -> top-n-sigma -> top-k -> typical -> top-p -> min-p -> xtc -> temp-ext -> dist
slot launch_slot_: id 2 | task 9343 | processing task
slot update_slots: id 2 | task 9343 | new prompt, n_ctx_slot = 262144, n_keep = 0, task.n_tokens = 5655
slot update_slots: id 2 | task 9343 | n_past = 5589, slot.prompt.tokens.size() = 29282, seq_id = 2, pos_min = 29281, n_swa = 1
slot update_slots: id 2 | task 9343 | forcing full prompt re-processing due to lack of cache data (likely due to SWA or hybrid/recurrent memory, see https://github.com//pull/13194#issuecomment-2868343055)
slot update_slots: id 2 | task 9343 | erased invalidated context checkpoint (pos_min = 5641, pos_max = 5641, n_swa = 1, size = 47.618 MiB)
slot update_slots: id 2 | task 9343 | erased invalidated context checkpoint (pos_min = 16268, pos_max = 16268, n_swa = 1, size = 47.618 MiB)
slot update_slots: id 2 | task 9343 | erased invalidated context checkpoint (pos_min = 26846, pos_max = 26846, n_swa = 1, size = 47.618 MiB)
slot update_slots: id 2 | task 9343 | erased invalidated context checkpoint (pos_min = 27347, pos_max = 27347, n_swa = 1, size = 47.618 MiB)
slot update_slots: id 2 | task 9343 | erased invalidated context checkpoint (pos_min = 27415, pos_max = 27415, n_swa = 1, size = 47.618 MiB)
slot update_slots: id 2 | task 9343 | erased invalidated context checkpoint (pos_min = 27615, pos_max = 27615, n_swa = 1, size = 47.618 MiB)
slot update_slots: id 2 | task 9343 | erased invalidated context checkpoint (pos_min = 27997, pos_max = 27997, n_swa = 1, size = 47.618 MiB)
slot update_slots: id 2 | task 9343 | erased invalidated context checkpoint (pos_min = 28534, pos_max = 28534, n_swa = 1, size = 47.618 MiB)
slot update_slots: id 2 | task 9343 | n_tokens = 0, memory_seq_rm [0, end)
slot update_slots: id 2 | task 9343 | prompt processing progress, n_tokens = 1625, batch.n_tokens = 1625, progress = 0.287356
slot update_slots: id 2 | task 9343 | n_tokens = 1625, memory_seq_rm [1625, end)
slot update_slots: id 2 | task 9343 | prompt processing progress, n_tokens = 3250, batch.n_tokens = 1625, progress = 0.574713
slot update_slots: id 2 | task 9343 | n_tokens = 3250, memory_seq_rm [3250, end)
slot update_slots: id 2 | task 9343 | prompt processing progress, n_tokens = 4875, batch.n_tokens = 1625, progress = 0.862069
slot update_slots: id 2 | task 9343 | n_tokens = 4875, memory_seq_rm [4875, end)
slot update_slots: id 2 | task 9343 | prompt processing progress, n_tokens = 5591, batch.n_tokens = 716, progress = 0.988683
slot update_slots: id 2 | task 9343 | n_tokens = 5591, memory_seq_rm [5591, end)
slot update_slots: id 2 | task 9343 | prompt processing progress, n_tokens = 5655, batch.n_tokens = 64, progress = 1.000000
slot update_slots: id 2 | task 9343 | prompt done, n_tokens = 5655, batch.n_tokens = 64
slot update_slots: id 2 | task 9343 | created context checkpoint 1 of 8 (pos_min = 5590, pos_max = 5590, size = 47.618 MiB)
slot print_timing: id 2 | task 9343 |
prompt eval time = 8991.38 ms / 5655 tokens ( 1.59 ms per token, 628.94 tokens per second)
eval time = 9852.93 ms / 347 tokens ( 28.39 ms per token, 35.22 tokens per second)
total time = 18844.31 ms / 6002 tokens
slot release: id 2 | task 9343 | stop processing: n_tokens = 6001, truncated = 0
srv update_slots: all slots are idle
srv log_server_r: request: POST /v1/chat/completions 127.0.0.1 200
srv params_from_: Chat format: peg-constructed
slot get_availabl: id 2 | task -1 | selected slot by LCP similarity, sim_best = 0.663 (> 0.100 thold), f_keep = 0.942
slot launch_slot_: id 2 | task -1 | sampler chain: logits -> penalties -> dry -> top-n-sigma -> top-k -> typical -> top-p -> min-p -> xtc -> temp-ext -> dist
slot launch_slot_: id 2 | task 9695 | processing task
slot update_slots: id 2 | task 9695 | new prompt, n_ctx_slot = 262144, n_keep = 0, task.n_tokens = 8530
slot update_slots: id 2 | task 9695 | n_past = 5654, slot.prompt.tokens.size() = 6001, seq_id = 2, pos_min = 6000, n_swa = 1
slot update_slots: id 2 | task 9695 | restored context checkpoint (pos_min = 5590, pos_max = 5590, size = 47.618 MiB)
slot update_slots: id 2 | task 9695 | n_tokens = 5591, memory_seq_rm [5591, end)
slot update_slots: id 2 | task 9695 | prompt processing progress, n_tokens = 7216, batch.n_tokens = 1625, progress = 0.845955
slot update_slots: id 2 | task 9695 | n_tokens = 7216, memory_seq_rm [7216, end)
slot update_slots: id 2 | task 9695 | prompt processing progress, n_tokens = 8466, batch.n_tokens = 1250, progress = 0.992497
slot update_slots: id 2 | task 9695 | n_tokens = 8466, memory_seq_rm [8466, end)
slot update_slots: id 2 | task 9695 | prompt processing progress, n_tokens = 8530, batch.n_tokens = 64, progress = 1.000000
slot update_slots: id 2 | task 9695 | prompt done, n_tokens = 8530, batch.n_tokens = 64
slot update_slots: id 2 | task 9695 | created context checkpoint 2 of 8 (pos_min = 8465, pos_max = 8465, size = 47.618 MiB)
slot print_timing: id 2 | task 9695 |
prompt eval time = 5219.43 ms / 2939 tokens ( 1.78 ms per token, 563.09 tokens per second)
eval time = 70685.57 ms / 2444 tokens ( 28.92 ms per token, 34.58 tokens per second)
total time = 75904.99 ms / 5383 tokens
slot release: id 2 | task 9695 | stop processing: n_tokens = 10973, truncated = 0
srv update_slots: all slots are idle
srv log_server_r: request: POST /v1/chat/completions 127.0.0.1 200
srv log_server_r: request: GET /api/tags 127.0.0.1 200
srv update_slots: all slots are idle
srv log_server_r: request: GET /slots 127.0.0.1 200
srv log_server_r: request: GET /api/tags 127.0.0.1 200
srv update_slots: all slots are idle
srv log_server_r: request: GET /slots 127.0.0.1 200
srv params_from_: Chat format: peg-constructed
slot get_availabl: id 2 | task -1 | selected slot by LCP similarity, sim_best = 0.792 (> 0.100 thold), f_keep = 0.777
slot launch_slot_: id 2 | task -1 | sampler chain: logits -> penalties -> dry -> top-n-sigma -> top-k -> typical -> top-p -> min-p -> xtc -> temp-ext -> dist
slot launch_slot_: id 2 | task 12144 | processing task
slot update_slots: id 2 | task 12144 | new prompt, n_ctx_slot = 262144, n_keep = 0, task.n_tokens = 10769
slot update_slots: id 2 | task 12144 | n_past = 8529, slot.prompt.tokens.size() = 10973, seq_id = 2, pos_min = 10972, n_swa = 1
slot update_slots: id 2 | task 12144 | restored context checkpoint (pos_min = 8465, pos_max = 8465, size = 47.618 MiB)
slot update_slots: id 2 | task 12144 | n_tokens = 8466, memory_seq_rm [8466, end)
slot update_slots: id 2 | task 12144 | prompt processing progress, n_tokens = 10091, batch.n_tokens = 1625, progress = 0.937042
slot update_slots: id 2 | task 12144 | n_tokens = 10091, memory_seq_rm [10091, end)
slot update_slots: id 2 | task 12144 | prompt processing progress, n_tokens = 10705, batch.n_tokens = 614, progress = 0.994057
slot update_slots: id 2 | task 12144 | n_tokens = 10705, memory_seq_rm [10705, end)
slot update_slots: id 2 | task 12144 | prompt processing progress, n_tokens = 10769, batch.n_tokens = 64, progress = 1.000000
slot update_slots: id 2 | task 12144 | prompt done, n_tokens = 10769, batch.n_tokens = 64
slot update_slots: id 2 | task 12144 | created context checkpoint 3 of 8 (pos_min = 10704, pos_max = 10704, size = 47.618 MiB)
slot print_timing: id 2 | task 12144 |
prompt eval time = 4906.92 ms / 2303 tokens ( 2.13 ms per token, 469.34 tokens per second)
eval time = 28626.52 ms / 985 tokens ( 29.06 ms per token, 34.41 tokens per second)
total time = 33533.44 ms / 3288 tokens
slot release: id 2 | task 12144 | stop processing: n_tokens = 11753, truncated = 0
srv update_slots: all slots are idle
srv log_server_r: request: POST /v1/chat/completions 127.0.0.1 200
srv params_from_: Chat format: peg-constructed
slot get_availabl: id 2 | task -1 | selected slot by LCP similarity, sim_best = 0.791 (> 0.100 thold), f_keep = 0.916
slot launch_slot_: id 2 | task -1 | sampler chain: logits -> penalties -> dry -> top-n-sigma -> top-k -> typical -> top-p -> min-p -> xtc -> temp-ext -> dist
slot launch_slot_: id 2 | task 13132 | processing task
slot update_slots: id 2 | task 13132 | new prompt, n_ctx_slot = 262144, n_keep = 0, task.n_tokens = 13614
slot update_slots: id 2 | task 13132 | n_past = 10768, slot.prompt.tokens.size() = 11753, seq_id = 2, pos_min = 11752, n_swa = 1
slot update_slots: id 2 | task 13132 | restored context checkpoint (pos_min = 10704, pos_max = 10704, size = 47.618 MiB)
slot update_slots: id 2 | task 13132 | n_tokens = 10705, memory_seq_rm [10705, end)
slot update_slots: id 2 | task 13132 | prompt processing progress, n_tokens = 12330, batch.n_tokens = 1625, progress = 0.905685
slot update_slots: id 2 | task 13132 | n_tokens = 12330, memory_seq_rm [12330, end)
slot update_slots: id 2 | task 13132 | prompt processing progress, n_tokens = 13550, batch.n_tokens = 1220, progress = 0.995299
slot update_slots: id 2 | task 13132 | n_tokens = 13550, memory_seq_rm [13550, end)
slot update_slots: id 2 | task 13132 | prompt processing progress, n_tokens = 13614, batch.n_tokens = 64, progress = 1.000000
slot update_slots: id 2 | task 13132 | prompt done, n_tokens = 13614, batch.n_tokens = 64
slot update_slots: id 2 | task 13132 | created context checkpoint 4 of 8 (pos_min = 13549, pos_max = 13549, size = 47.618 MiB)
slot print_timing: id 2 | task 13132 |
prompt eval time = 5195.99 ms / 2909 tokens ( 1.79 ms per token, 559.85 tokens per second)
eval time = 8079.72 ms / 276 tokens ( 29.27 ms per token, 34.16 tokens per second)
total time = 13275.70 ms / 3185 tokens
slot release: id 2 | task 13132 | stop processing: n_tokens = 13889, truncated = 0
srv update_slots: all slots are idle
srv log_server_r: request: POST /v1/chat/completions 127.0.0.1 200
srv params_from_: Chat format: peg-constructed
slot get_availabl: id 2 | task -1 | selected slot by LCP similarity, sim_best = 0.825 (> 0.100 thold), f_keep = 0.980
slot launch_slot_: id 2 | task -1 | sampler chain: logits -> penalties -> dry -> top-n-sigma -> top-k -> typical -> top-p -> min-p -> xtc -> temp-ext -> dist
slot launch_slot_: id 2 | task 13411 | processing task
slot update_slots: id 2 | task 13411 | new prompt, n_ctx_slot = 262144, n_keep = 0, task.n_tokens = 16495
slot update_slots: id 2 | task 13411 | n_past = 13613, slot.prompt.tokens.size() = 13889, seq_id = 2, pos_min = 13888, n_swa = 1
slot update_slots: id 2 | task 13411 | restored context checkpoint (pos_min = 13549, pos_max = 13549, size = 47.618 MiB)
slot update_slots: id 2 | task 13411 | n_tokens = 13550, memory_seq_rm [13550, end)
slot update_slots: id 2 | task 13411 | prompt processing progress, n_tokens = 15175, batch.n_tokens = 1625, progress = 0.919976
slot update_slots: id 2 | task 13411 | n_tokens = 15175, memory_seq_rm [15175, end)
slot update_slots: id 2 | task 13411 | prompt processing progress, n_tokens = 16431, batch.n_tokens = 1256, progress = 0.996120
slot update_slots: id 2 | task 13411 | n_tokens = 16431, memory_seq_rm [16431, end)
slot update_slots: id 2 | task 13411 | prompt processing progress, n_tokens = 16495, batch.n_tokens = 64, progress = 1.000000
slot update_slots: id 2 | task 13411 | prompt done, n_tokens = 16495, batch.n_tokens = 64
slot update_slots: id 2 | task 13411 | created context checkpoint 5 of 8 (pos_min = 16430, pos_max = 16430, size = 47.618 MiB)
slot print_timing: id 2 | task 13411 |
prompt eval time = 5301.19 ms / 2945 tokens ( 1.80 ms per token, 555.54 tokens per second)
eval time = 6638.55 ms / 222 tokens ( 29.90 ms per token, 33.44 tokens per second)
total time = 11939.73 ms / 3167 tokens
slot release: id 2 | task 13411 | stop processing: n_tokens = 16716, truncated = 0
srv update_slots: all slots are idle
srv log_server_r: request: POST /v1/chat/completions 127.0.0.1 200
srv params_from_: Chat format: peg-constructed
slot get_availabl: id 2 | task -1 | selected slot by LCP similarity, sim_best = 0.880 (> 0.100 thold), f_keep = 0.987
slot launch_slot_: id 2 | task -1 | sampler chain: logits -> penalties -> dry -> top-n-sigma -> top-k -> typical -> top-p -> min-p -> xtc -> temp-ext -> dist
slot launch_slot_: id 2 | task 13636 | processing task
slot update_slots: id 2 | task 13636 | new prompt, n_ctx_slot = 262144, n_keep = 0, task.n_tokens = 18753
slot update_slots: id 2 | task 13636 | n_past = 16494, slot.prompt.tokens.size() = 16716, seq_id = 2, pos_min = 16715, n_swa = 1
slot update_slots: id 2 | task 13636 | restored context checkpoint (pos_min = 16430, pos_max = 16430, size = 47.618 MiB)
slot update_slots: id 2 | task 13636 | n_tokens = 16431, memory_seq_rm [16431, end)
slot update_slots: id 2 | task 13636 | prompt processing progress, n_tokens = 18056, batch.n_tokens = 1625, progress = 0.962833
slot update_slots: id 2 | task 13636 | n_tokens = 18056, memory_seq_rm [18056, end)
slot update_slots: id 2 | task 13636 | prompt processing progress, n_tokens = 18689, batch.n_tokens = 633, progress = 0.996587
slot update_slots: id 2 | task 13636 | n_tokens = 18689, memory_seq_rm [18689, end)
slot update_slots: id 2 | task 13636 | prompt processing progress, n_tokens = 18753, batch.n_tokens = 64, progress = 1.000000
slot update_slots: id 2 | task 13636 | prompt done, n_tokens = 18753, batch.n_tokens = 64
slot update_slots: id 2 | task 13636 | created context checkpoint 6 of 8 (pos_min = 18688, pos_max = 18688, size = 47.618 MiB)
slot print_timing: id 2 | task 13636 |
prompt eval time = 5169.56 ms / 2322 tokens ( 2.23 ms per token, 449.17 tokens per second)
eval time = 9686.29 ms / 322 tokens ( 30.08 ms per token, 33.24 tokens per second)
total time = 14855.85 ms / 2644 tokens
slot release: id 2 | task 13636 | stop processing: n_tokens = 19074, truncated = 0
srv update_slots: all slots are idle
srv log_server_r: request: POST /v1/chat/completions 127.0.0.1 200
srv params_from_: Chat format: peg-constructed
slot get_availabl: id 2 | task -1 | selected slot by LCP similarity, sim_best = 0.865 (> 0.100 thold), f_keep = 0.983
slot launch_slot_: id 2 | task -1 | sampler chain: logits -> penalties -> dry -> top-n-sigma -> top-k -> typical -> top-p -> min-p -> xtc -> temp-ext -> dist
slot launch_slot_: id 2 | task 13961 | processing task
slot update_slots: id 2 | task 13961 | new prompt, n_ctx_slot = 262144, n_keep = 0, task.n_tokens = 21668
slot update_slots: id 2 | task 13961 | n_past = 18752, slot.prompt.tokens.size() = 19074, seq_id = 2, pos_min = 19073, n_swa = 1
slot update_slots: id 2 | task 13961 | restored context checkpoint (pos_min = 18688, pos_max = 18688, size = 47.618 MiB)
slot update_slots: id 2 | task 13961 | n_tokens = 18689, memory_seq_rm [18689, end)
slot update_slots: id 2 | task 13961 | prompt processing progress, n_tokens = 20314, batch.n_tokens = 1625, progress = 0.937512
slot update_slots: id 2 | task 13961 | n_tokens = 20314, memory_seq_rm [20314, end)
slot update_slots: id 2 | task 13961 | prompt processing progress, n_tokens = 21604, batch.n_tokens = 1290, progress = 0.997046
slot update_slots: id 2 | task 13961 | n_tokens = 21604, memory_seq_rm [21604, end)
slot update_slots: id 2 | task 13961 | prompt processing progress, n_tokens = 21668, batch.n_tokens = 64, progress = 1.000000
slot update_slots: id 2 | task 13961 | prompt done, n_tokens = 21668, batch.n_tokens = 64
slot update_slots: id 2 | task 13961 | created context checkpoint 7 of 8 (pos_min = 21603, pos_max = 21603, size = 47.618 MiB)
slot print_timing: id 2 | task 13961 |
prompt eval time = 5810.13 ms / 2979 tokens ( 1.95 ms per token, 512.73 tokens per second)
eval time = 66240.13 ms / 2182 tokens ( 30.36 ms per token, 32.94 tokens per second)
total time = 72050.26 ms / 5161 tokens
slot release: id 2 | task 13961 | stop processing: n_tokens = 23849, truncated = 0
srv update_slots: all slots are idle
srv log_server_r: request: POST /v1/chat/completions 127.0.0.1 200
srv params_from_: Chat format: peg-constructed
slot get_availabl: id 2 | task -1 | selected slot by LCP similarity, sim_best = 0.924 (> 0.100 thold), f_keep = 0.909
slot launch_slot_: id 2 | task -1 | sampler chain: logits -> penalties -> dry -> top-n-sigma -> top-k -> typical -> top-p -> min-p -> xtc -> temp-ext -> dist
slot launch_slot_: id 2 | task 16146 | processing task
slot update_slots: id 2 | task 16146 | new prompt, n_ctx_slot = 262144, n_keep = 0, task.n_tokens = 23447
slot update_slots: id 2 | task 16146 | n_past = 21667, slot.prompt.tokens.size() = 23849, seq_id = 2, pos_min = 23848, n_swa = 1
slot update_slots: id 2 | task 16146 | restored context checkpoint (pos_min = 21603, pos_max = 21603, size = 47.618 MiB)
slot update_slots: id 2 | task 16146 | n_tokens = 21604, memory_seq_rm [21604, end)
slot update_slots: id 2 | task 16146 | prompt processing progress, n_tokens = 23229, batch.n_tokens = 1625, progress = 0.990702
slot update_slots: id 2 | task 16146 | n_tokens = 23229, memory_seq_rm [23229, end)
slot update_slots: id 2 | task 16146 | prompt processing progress, n_tokens = 23383, batch.n_tokens = 154, progress = 0.997270
slot update_slots: id 2 | task 16146 | n_tokens = 23383, memory_seq_rm [23383, end)
slot update_slots: id 2 | task 16146 | prompt processing progress, n_tokens = 23447, batch.n_tokens = 64, progress = 1.000000
slot update_slots: id 2 | task 16146 | prompt done, n_tokens = 23447, batch.n_tokens = 64
slot update_slots: id 2 | task 16146 | created context checkpoint 8 of 8 (pos_min = 23382, pos_max = 23382, size = 47.618 MiB)
slot print_timing: id 2 | task 16146 |
prompt eval time = 4699.71 ms / 1843 tokens ( 2.55 ms per token, 392.15 tokens per second)
eval time = 72890.12 ms / 2384 tokens ( 30.57 ms per token, 32.71 tokens per second)
total time = 77589.83 ms / 4227 tokens
slot release: id 2 | task 16146 | stop processing: n_tokens = 25830, truncated = 0
srv update_slots: all slots are idle
srv log_server_r: request: POST /v1/chat/completions 127.0.0.1 200
srv params_from_: Chat format: peg-constructed
slot get_availabl: id 2 | task -1 | selected slot by LCP similarity, sim_best = 0.910 (> 0.100 thold), f_keep = 0.908
slot launch_slot_: id 2 | task -1 | sampler chain: logits -> penalties -> dry -> top-n-sigma -> top-k -> typical -> top-p -> min-p -> xtc -> temp-ext -> dist
slot launch_slot_: id 2 | task 18533 | processing task
slot update_slots: id 2 | task 18533 | new prompt, n_ctx_slot = 262144, n_keep = 0, task.n_tokens = 25755
slot update_slots: id 2 | task 18533 | n_past = 23446, slot.prompt.tokens.size() = 25830, seq_id = 2, pos_min = 25829, n_swa = 1
slot update_slots: id 2 | task 18533 | restored context checkpoint (pos_min = 23382, pos_max = 23382, size = 47.618 MiB)
slot update_slots: id 2 | task 18533 | n_tokens = 23383, memory_seq_rm [23383, end)
slot update_slots: id 2 | task 18533 | prompt processing progress, n_tokens = 25008, batch.n_tokens = 1625, progress = 0.970996
slot update_slots: id 2 | task 18533 | n_tokens = 25008, memory_seq_rm [25008, end)
slot update_slots: id 2 | task 18533 | prompt processing progress, n_tokens = 25691, batch.n_tokens = 683, progress = 0.997515
slot update_slots: id 2 | task 18533 | n_tokens = 25691, memory_seq_rm [25691, end)
slot update_slots: id 2 | task 18533 | prompt processing progress, n_tokens = 25755, batch.n_tokens = 64, progress = 1.000000
slot update_slots: id 2 | task 18533 | prompt done, n_tokens = 25755, batch.n_tokens = 64
slot update_slots: id 2 | task 18533 | erasing old context checkpoint (pos_min = 5590, pos_max = 5590, size = 47.618 MiB)
slot update_slots: id 2 | task 18533 | created context checkpoint 8 of 8 (pos_min = 25690, pos_max = 25690, size = 47.618 MiB)
slot print_timing: id 2 | task 18533 |
prompt eval time = 5397.80 ms / 2372 tokens ( 2.28 ms per token, 439.44 tokens per second)
eval time = 97035.56 ms / 3141 tokens ( 30.89 ms per token, 32.37 tokens per second)
total time = 102433.36 ms / 5513 tokens
slot release: id 2 | task 18533 | stop processing: n_tokens = 28895, truncated = 0
srv update_slots: all slots are idle
srv log_server_r: request: POST /v1/chat/completions 127.0.0.1 200
srv params_from_: Chat format: peg-constructed
slot get_availabl: id 2 | task -1 | selected slot by LCP similarity, sim_best = 0.902 (> 0.100 thold), f_keep = 0.891
slot launch_slot_: id 2 | task -1 | sampler chain: logits -> penalties -> dry -> top-n-sigma -> top-k -> typical -> top-p -> min-p -> xtc -> temp-ext -> dist
slot launch_slot_: id 2 | task 21677 | processing task
slot update_slots: id 2 | task 21677 | new prompt, n_ctx_slot = 262144, n_keep = 0, task.n_tokens = 28550
slot update_slots: id 2 | task 21677 | n_past = 25754, slot.prompt.tokens.size() = 28895, seq_id = 2, pos_min = 28894, n_swa = 1
slot update_slots: id 2 | task 21677 | restored context checkpoint (pos_min = 25690, pos_max = 25690, size = 47.618 MiB)
slot update_slots: id 2 | task 21677 | n_tokens = 25691, memory_seq_rm [25691, end)
slot update_slots: id 2 | task 21677 | prompt processing progress, n_tokens = 27316, batch.n_tokens = 1625, progress = 0.956778
slot update_slots: id 2 | task 21677 | n_tokens = 27316, memory_seq_rm [27316, end)
slot update_slots: id 2 | task 21677 | prompt processing progress, n_tokens = 28486, batch.n_tokens = 1170, progress = 0.997758
slot update_slots: id 2 | task 21677 | n_tokens = 28486, memory_seq_rm [28486, end)
slot update_slots: id 2 | task 21677 | prompt processing progress, n_tokens = 28550, batch.n_tokens = 64, progress = 1.000000
slot update_slots: id 2 | task 21677 | prompt done, n_tokens = 28550, batch.n_tokens = 64
slot update_slots: id 2 | task 21677 | erasing old context checkpoint (pos_min = 8465, pos_max = 8465, size = 47.618 MiB)
slot update_slots: id 2 | task 21677 | created context checkpoint 8 of 8 (pos_min = 28485, pos_max = 28485, size = 47.618 MiB)
slot print_timing: id 2 | task 21677 |
prompt eval time = 6025.06 ms / 2859 tokens ( 2.11 ms per token, 474.52 tokens per second)
eval time = 108551.40 ms / 3474 tokens ( 31.25 ms per token, 32.00 tokens per second)
total time = 114576.45 ms / 6333 tokens
slot release: id 2 | task 21677 | stop processing: n_tokens = 32023, truncated = 0
srv update_slots: all slots are idle

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Codex Local Token Counter Fix - Instructional Guide for llama.cpp #18847

Uh oh!

{{title}}

Uh oh!

Replies: 0 comments

Select a reply

Uh oh!

Codex Local Token Counter Fix - Instructional Guide for llama.cpp #18847

Uh oh!

jbulger82 Jan 14, 2026

Codex Local Token Counter Fix - Instructional Guide

🎯 What This Fixes

📋 Prerequisites

🔧 The Fix (4 Files to Modify)

File 1: codex-api/src/requests/chat.rs

File 2: codex-api/src/sse/chat.rs

Step 2a: Add Import

Step 2b: Add Tracking Variable

Step 2c: Modify flush_and_complete Function

Step 2d: Extract Usage from JSON

Step 2e: Update All flush_and_complete Calls

File 3: core/src/models_manager/model_info.rs

File 4: config.toml

🔨 Build & Test

🔍 How It Works

⚠️ Troubleshooting

Still showing 0 tokens?

Context percentage wrong?

Auto-compact not triggering?

📊 Token Count Breakdown (For Reference)

🙏 Credits

📜 License

Codex Benchmark (12GB VRAM)

Replies: 0 comments

jbulger82
Jan 14, 2026

File 1: `codex-api/src/requests/chat.rs`

File 2: `codex-api/src/sse/chat.rs`

Step 2c: Modify `flush_and_complete` Function

Step 2e: Update All `flush_and_complete` Calls

File 3: `core/src/models_manager/model_info.rs`

File 4: `config.toml`