#### in llama.cpp distribution directory, this shell script: go_test_huge_model.sh
cmake -B build
cmake --build build --config Release -j 4
cmake --build build --config Release -t llama-server
cmake --build build --config Release -t llama-cli

# example from https://huggingface.co/unsloth/DeepSeek-R1-Distill-Qwen-32B-GGUF



# model that easily fits in RAM:
MODEL=/aimodels/cache/DeepSeek-R1-Distill-Qwen-14B-Q5_K_M.gguf

# models that don't easily fit into RAM:
MODEL=/u03/aimodels/cache/Llama-3.3-70B-Instruct-Q4_K_L.gguf

# model that is an order of magnitude larger than RAM:
MODEL=/u03/aimodels/deepseek/DeepSeek-V3-GGUF/DeepSeek-V3-BF16-256x20B-Q8_0-00001-of-00016.gguf



# clear Linux cache for test
echo dropping caches ...
echo 3 | sudo tee /proc/sys/vm/drop_caches
echo caches dropped
# sudo sh -c "/usr/bin/echo 3 > /proc/sys/vm/drop_caches"

SEED=12

./build/bin/llama-cli \
    --model $MODEL \
    --no-warmup  \
    --cache-type-k q8_0 \
    --threads 4 \
    --seed $SEED \
    --temp 0.7 \
    --top-p 0.95 \
    --no-warmup \
    --ctx-size 2048 \
    --prompt '<｜User｜>
    Specifications: Using PyGame, open a window that is 4-inches square, with a black background, and create an equilateral triangle that is 2-inches per side, hollow, white lines. It is rotating clockwise at 6 degrees per second - to match that of a second-hand on a watch. Inside the equilateral triangle is a red ball, 1/8" in diameter. Unaffected by gravity, it is moving with an arbitrary speed, that is configurable in a variable. The ball has proper hitbox checking and can detect when it hits a wall. Hitting a wall should deflect/redirect the ball into the proper direction based on the angle of the wall that it hit and how it should deflect given the angle and speed of the wall. Be careful of bounds checking such that the ball does not get stuck outside a wall. Remember this is all in 2D space. Start with the ball located in the center of the triangle, initially moving in a random direction and with an initial speed of about one inch per second. To clarify, the triangle has three equal 2-inch sides, rotating at 6 degrees per second (not 6 degrees per frame). Be careful of pygame functions that may return zero at times (e.g. framerate) and avoid divide-by-zero errors.

Assumptions you may make:
Screen size is adequate with 154 pixels per inch resolution.
Frame rate is nominally 60 frames per second but may vary with system load.
Pygame is installed.

Window title requirement: Please title the pygame window "Triangle-Ball '$(basename $MODEL)-$SEED'".

Code block requirement: Emit a fully functioning python program within a single code block. Enclose your python program in a single codeblock as such:
\\n
```python
# your program here
```

Your code will be evaluated with the following rubric:
Does the program fail to execute due to a Python error? 0 points.
Does the program load a pygame window with a slowly rotating equilateral triangle and a red ball? +1 point.
Are the dimensions, ball speed, and triangle rotation exactly as specified? +1 point.
Does the ball interact with the rotating sides? +1 point.
Does the ball always correctly bounce inside the triangle and never get caught in a side or vertex? +1 point.
Does the program function correctly in extreme varying frame rate conditions down to 4 fps? +1 point.

Review all the specifications and requirements, then begin coding. If you would like to think in text before beginning your codeblock, that is perfectly acceptable and may help improve your answer. Only the code in your codeblock will be evaluated.

Thanks!
<｜Assistant｜>' \
	-no-cnv 2>&1 | tee $(basename $MODEL)_hugemodel.out.$SEED.txt

cat $(basename $MODEL)_hugemodel.out.$SEED.txt |
sed 's/.*```/```/;' |
sed -n '/^```python/,/^```/p' |
sed '1d; $d' |
sed 's/.*```.*/#&/' > $(basename $MODEL)_hugemodel.out.$SEED.py

########### END OF TEST SCRIPT ###########

########### OUTPUT ###########

build: 4735 (73e2ed3c) with cc (Ubuntu 9.4.0-1ubuntu1~20.04.2) 9.4.0 for x86_64-linux-gnu
main: llama backend init
main: load the model and apply lora adapter, if any
llama_model_loader: additional 15 GGUFs metadata loaded.
llama_model_loader: loaded meta data with 46 key-value pairs and 1025 tensors from /u03/aimodels/deepseek/DeepSeek-V3-GGUF/DeepSeek-V3-BF16-256x20B-Q8_0-00001-of-00016.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              = deepseek2
llama_model_loader: - kv   1:                               general.type str              = model
llama_model_loader: - kv   2:                               general.name str              = DeepSeek V3 BF16
llama_model_loader: - kv   3:                         general.size_label str              = 256x20B
llama_model_loader: - kv   4:                      deepseek2.block_count u32              = 61
llama_model_loader: - kv   5:                   deepseek2.context_length u32              = 163840
llama_model_loader: - kv   6:                 deepseek2.embedding_length u32              = 7168
llama_model_loader: - kv   7:              deepseek2.feed_forward_length u32              = 18432
llama_model_loader: - kv   8:             deepseek2.attention.head_count u32              = 128
llama_model_loader: - kv   9:          deepseek2.attention.head_count_kv u32              = 128
llama_model_loader: - kv  10:                   deepseek2.rope.freq_base f32              = 10000.000000
llama_model_loader: - kv  11: deepseek2.attention.layer_norm_rms_epsilon f32              = 0.000001
llama_model_loader: - kv  12:                deepseek2.expert_used_count u32              = 8
llama_model_loader: - kv  13:                          general.file_type u32              = 7
llama_model_loader: - kv  14:        deepseek2.leading_dense_block_count u32              = 3
llama_model_loader: - kv  15:                       deepseek2.vocab_size u32              = 129280
llama_model_loader: - kv  16:            deepseek2.attention.q_lora_rank u32              = 1536
llama_model_loader: - kv  17:           deepseek2.attention.kv_lora_rank u32              = 512
llama_model_loader: - kv  18:             deepseek2.attention.key_length u32              = 192
llama_model_loader: - kv  19:           deepseek2.attention.value_length u32              = 128
llama_model_loader: - kv  20:       deepseek2.expert_feed_forward_length u32              = 2048
llama_model_loader: - kv  21:                     deepseek2.expert_count u32              = 256
llama_model_loader: - kv  22:              deepseek2.expert_shared_count u32              = 1
llama_model_loader: - kv  23:             deepseek2.expert_weights_scale f32              = 2.500000
llama_model_loader: - kv  24:              deepseek2.expert_weights_norm bool             = true
llama_model_loader: - kv  25:               deepseek2.expert_gating_func u32              = 2
llama_model_loader: - kv  26:             deepseek2.rope.dimension_count u32              = 64
llama_model_loader: - kv  27:                deepseek2.rope.scaling.type str              = yarn
llama_model_loader: - kv  28:              deepseek2.rope.scaling.factor f32              = 40.000000
llama_model_loader: - kv  29: deepseek2.rope.scaling.original_context_length u32              = 4096
llama_model_loader: - kv  30: deepseek2.rope.scaling.yarn_log_multiplier f32              = 0.100000
llama_model_loader: - kv  31:                       tokenizer.ggml.model str              = gpt2
llama_model_loader: - kv  32:                         tokenizer.ggml.pre str              = deepseek-v3
llama_model_loader: - kv  33:                      tokenizer.ggml.tokens arr[str,129280]  = ["<｜begin▁of▁sentence｜>", "<�...
llama_model_loader: - kv  34:                  tokenizer.ggml.token_type arr[i32,129280]  = [3, 3, 3, 1, 1, 1, 1, 1, 1, 1, 1, 1, ...
llama_model_loader: - kv  35:                      tokenizer.ggml.merges arr[str,127741]  = ["Ġ t", "Ġ a", "i n", "Ġ Ġ", "h e...
llama_model_loader: - kv  36:                tokenizer.ggml.bos_token_id u32              = 0
llama_model_loader: - kv  37:                tokenizer.ggml.eos_token_id u32              = 1
llama_model_loader: - kv  38:            tokenizer.ggml.padding_token_id u32              = 1
llama_model_loader: - kv  39:               tokenizer.ggml.add_bos_token bool             = true
llama_model_loader: - kv  40:               tokenizer.ggml.add_eos_token bool             = false
llama_model_loader: - kv  41:                    tokenizer.chat_template str              = {% if not add_generation_prompt is de...
llama_model_loader: - kv  42:               general.quantization_version u32              = 2
llama_model_loader: - kv  43:                                   split.no u16              = 0
llama_model_loader: - kv  44:                                split.count u16              = 16
llama_model_loader: - kv  45:                        split.tensors.count i32              = 1025
llama_model_loader: - type  f32:  361 tensors
llama_model_loader: - type q8_0:  664 tensors
print_info: file format = GGUF V3 (latest)
print_info: file type   = Q8_0
print_info: file size   = 664.29 GiB (8.50 BPW)
load: special_eos_id is not in special_eog_ids - the tokenizer config may be incorrect
load: special tokens cache size = 818
load: token to piece cache size = 0.8223 MB
print_info: arch             = deepseek2
print_info: vocab_only       = 0
print_info: n_ctx_train      = 163840
print_info: n_embd           = 7168
print_info: n_layer          = 61
print_info: n_head           = 128
print_info: n_head_kv        = 128
print_info: n_rot            = 64
print_info: n_swa            = 0
print_info: n_embd_head_k    = 192
print_info: n_embd_head_v    = 128
print_info: n_gqa            = 1
print_info: n_embd_k_gqa     = 24576
print_info: n_embd_v_gqa     = 16384
print_info: f_norm_eps       = 0.0e+00
print_info: f_norm_rms_eps   = 1.0e-06
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: n_ff             = 18432
print_info: n_expert         = 256
print_info: n_expert_used    = 8
print_info: causal attn      = 1
print_info: pooling type     = 0
print_info: rope type        = 0
print_info: rope scaling     = yarn
print_info: freq_base_train  = 10000.0
print_info: freq_scale_train = 0.025
print_info: n_ctx_orig_yarn  = 4096
print_info: rope_finetuned   = unknown
print_info: ssm_d_conv       = 0
print_info: ssm_d_inner      = 0
print_info: ssm_d_state      = 0
print_info: ssm_dt_rank      = 0
print_info: ssm_dt_b_c_rms   = 0
print_info: model type       = 671B
print_info: model params     = 671.03 B
print_info: general.name     = DeepSeek V3 BF16
print_info: n_layer_dense_lead   = 3
print_info: n_lora_q             = 1536
print_info: n_lora_kv            = 512
print_info: n_ff_exp             = 2048
print_info: n_expert_shared      = 1
print_info: expert_weights_scale = 2.5
print_info: expert_weights_norm  = 1
print_info: expert_gating_func   = sigmoid
print_info: rope_yarn_log_mul    = 0.1000
print_info: vocab type       = BPE
print_info: n_vocab          = 129280
print_info: n_merges         = 127741
print_info: BOS token        = 0 '<｜begin▁of▁sentence｜>'
print_info: EOS token        = 1 '<｜end▁of▁sentence｜>'
print_info: EOT token        = 1 '<｜end▁of▁sentence｜>'
print_info: PAD token        = 1 '<｜end▁of▁sentence｜>'
print_info: LF token         = 201 'Ċ'
print_info: FIM PRE token    = 128801 '<｜fim▁begin｜>'
print_info: FIM SUF token    = 128800 '<｜fim▁hole｜>'
print_info: FIM MID token    = 128802 '<｜fim▁end｜>'
print_info: EOG token        = 1 '<｜end▁of▁sentence｜>'
print_info: max token length = 256
load_tensors: loading model tensors, this can take a while... (mmap = true)
load_tensors:   CPU_Mapped model buffer size = 41757.90 MiB
load_tensors:   CPU_Mapped model buffer size = 42853.11 MiB
load_tensors:   CPU_Mapped model buffer size = 42611.83 MiB
load_tensors:   CPU_Mapped model buffer size = 42853.11 MiB
load_tensors:   CPU_Mapped model buffer size = 42853.11 MiB
load_tensors:   CPU_Mapped model buffer size = 42611.83 MiB
load_tensors:   CPU_Mapped model buffer size = 42853.11 MiB
load_tensors:   CPU_Mapped model buffer size = 42853.11 MiB
load_tensors:   CPU_Mapped model buffer size = 42611.83 MiB
load_tensors:   CPU_Mapped model buffer size = 42853.11 MiB
load_tensors:   CPU_Mapped model buffer size = 42853.11 MiB
load_tensors:   CPU_Mapped model buffer size = 42611.83 MiB
load_tensors:   CPU_Mapped model buffer size = 42853.11 MiB
load_tensors:   CPU_Mapped model buffer size = 42853.11 MiB
load_tensors:   CPU_Mapped model buffer size = 42611.83 MiB
load_tensors:   CPU_Mapped model buffer size = 39742.87 MiB
....................................................................................................
llama_init_from_model: n_seq_max     = 1
llama_init_from_model: n_ctx         = 2048
llama_init_from_model: n_ctx_per_seq = 2048
llama_init_from_model: n_batch       = 2048
llama_init_from_model: n_ubatch      = 512
llama_init_from_model: flash_attn    = 0
llama_init_from_model: freq_base     = 10000.0
llama_init_from_model: freq_scale    = 0.025
llama_init_from_model: n_ctx_per_seq (2048) < n_ctx_train (163840) -- the full capacity of the model will not be utilized
llama_kv_cache_init: kv_size = 2048, offload = 1, type_k = 'q8_0', type_v = 'f16', n_layer = 61, can_shift = 0
llama_kv_cache_init:        CPU KV buffer size =  7015.00 MiB
llama_init_from_model: KV self size  = 7015.00 MiB, K (q8_0): 3111.00 MiB, V (f16): 3904.00 MiB
llama_init_from_model:        CPU  output buffer size =     0.49 MiB
llama_init_from_model:        CPU compute buffer size =   670.01 MiB
llama_init_from_model: graph nodes  = 5025
llama_init_from_model: graph splits = 1
common_init_from_params: KV cache shifting is not supported for this model, disabling KV cache shifting
common_init_from_params: setting dry_penalty_last_n to ctx_size = 2048
main: llama threadpool init, n_threads = 4

system_info: n_threads = 4 (n_threads_batch = 4) / 4 | CPU : SSE3 = 1 | SSSE3 = 1 | AVX = 1 | AVX2 = 1 | F16C = 1 | FMA = 1 | LLAMAFILE = 1 | OPENMP = 1 | AARCH64_REPACK = 1 |

sampler seed: 12
sampler params:
	repeat_last_n = 64, repeat_penalty = 1.000, frequency_penalty = 0.000, presence_penalty = 0.000
	dry_multiplier = 0.000, dry_base = 1.750, dry_allowed_length = 2, dry_penalty_last_n = 2048
	top_k = 40, top_p = 0.950, min_p = 0.050, xtc_probability = 0.000, xtc_threshold = 0.100, typical_p = 1.000, top_n_sigma = -1.000, temp = 0.700
	mirostat = 0, mirostat_lr = 0.100, mirostat_ent = 5.000
sampler chain: logits -> logit-bias -> penalties -> dry -> top-k -> typical -> top-p -> min-p -> xtc -> temp-ext -> dist
generate: n_ctx = 2048, n_batch = 2048, n_predict = -1, n_keep = 1


    Specifications: Using PyGame, open a window that is 4-inches square, with a black background, and create an equilateral triangle that is 2-inches per side, hollow, white lines. It is rotating clockwise at 6 degrees per second - to match that of a second-hand on a watch. Inside the equilateral triangle is a red ball, 1/8" in diameter. Unaffected by gravity, it is moving with an arbitrary speed, that is configurable in a variable. The ball has proper hitbox checking and can detect when it hits a wall. Hitting a wall should deflect/redirect the ball into the proper direction based on the angle of the wall that it hit and how it should deflect given the angle and speed of the wall. Be careful of bounds checking such that the ball does not get stuck outside a wall. Remember this is all in 2D space. Start with the ball located in the center of the triangle, initially moving in a random direction and with an initial speed of about one inch per second. To clarify, the triangle has three equal 2-inch sides, rotating at 6 degrees per second (not 6 degrees per frame). Be careful of pygame functions that may return zero at times (e.g. framerate) and avoid divide-by-zero errors.

Assumptions you may make:
Screen size is adequate with 154 pixels per inch resolution.
Frame rate is nominally 60 frames per second but may vary with system load.
Pygame is installed.

Window title requirement: Please title the pygame window "Triangle-Ball DeepSeek-V3-BF16-256x20B-Q8_0-00001-of-00016.gguf-12".

Code block requirement: Emit a fully functioning python program within a single code block. Enclose your python program in a single codeblock as such:
\n
```python
# your program here
```

Your code will be evaluated with the following rubric:
Does the program fail to execute due to a Python error? 0 points.
Does the program load a pygame window with a slowly rotating equilateral triangle and a red ball? +1 point.
Are the dimensions, ball speed, and triangle rotation exactly as specified? +1 point.
Does the ball interact with the rotating sides? +1 point.
Does the ball always correctly bounce inside the triangle and never get caught in a side or vertex? +1 point.
Does the program function correctly in extreme varying frame rate conditions down to 4 fps? +1 point.

Review all the specifications and requirements, then begin coding. If you would like to think in text before beginning your codeblock, that is perfectly acceptable and may help improve your answer. Only the code in your codeblock will be evaluated.

Thanks!
```python
import pygame
import math
import random

# Constants
PIXELS_PER_INCH = 154
WINDOW_SIZE = 4 * PIXELS_PER_INCH
TRIANGLE_SIDE = 2 * PIXELS_PER_INCH
BALL_DIAMETER = (1/8) * PIXELS_PER_INCH
BALL_RADIUS = BALL_DIAMETER / 2
TRIANGLE_ROTATION_SPEED = 6  # degrees per second
BALL_SPEED = 1 * PIXELS_PER_INCH  # pixels per second
FRAME_RATE = 60

# Colors
BLACK = (0, 0, 0)
WHITE = (255, 255, 255)
RED = (255, 0, 0)

# Initialize Pygame
pygame.init()
screen = pygame.display.set_mode((WINDOW_SIZE, WINDOW_SIZE))
pygame.display.set_caption("Triangle-Ball DeepSeek-V3-BF16-256x20B-Q8_0-00001-of-00016.gguf-12")
clock = pygame.time.Clock()

# Triangle vertices
def get_triangle_vertices(angle):
    center = (WINDOW_SIZE // 2, WINDOW_SIZE // 2)
    vertices = []
    for i in range(3):
        theta = math.radians(angle + 120 * i)
        x = center[0] + (TRIANGLE_SIDE / math.sqrt(3)) * math.cos(theta)
        y = center[1] + (TRIANGLE_SIDE / math.sqrt(3)) * math.sin(theta)
        vertices.append((x, y))
    return vertices

# Ball properties
ball_pos = [WINDOW_SIZE // 2, WINDOW_SIZE // 2]
ball_velocity = [random.uniform(-1, 1), random.uniform(-1, 1)]
ball_velocity = [v / math.hypot(*ball_velocity) * BALL_SPEED for v in ball_velocity]

# Rotation angle
angle = 0

# Main loop
running = True
while running:
    dt = clock.tick(FRAME_RATE) / 1000  # Time since last frame in seconds

    # Handle events
    for event in pygame.event.get():
        if event.type == pygame.QUIT:
            running = False

    # Update triangle rotation
    angle += TRIANGLE_ROTATION_SPEED * dt
    triangle_vertices = get_triangle_vertices(angle)

    # Update ball position
    ball_pos[0] += ball_velocity[0] * dt
    ball_pos[1] += ball_velocity[1] * dt

    # Ball-wall collision detection
    for i in range(3):
        p1 = triangle_vertices[i]
        p2 = triangle_vertices[(i + 1) % 3]

        # Calculate the normal vector of the wall
        wall_vector = (p2[0] - p1[0], p2[1] - p1[1])
        normal_vector = (-wall_vector[1], wall_vector[0])
        normal_length = math.hypot(*normal_vector)
        normal_vector = (normal_vector[0] / normal_length, normal_vector[1] / normal_length)

        # Calculate the distance from the ball to the wall
        ball_to_wall = (ball_pos[0] - p1[0], ball_pos[1] - p1[1])
        distance = abs(ball_to_wall[0] * normal_vector[0] + ball_to_wall[1] * normal_vector[1])

        if distance <= BALL_RADIUS:
            # Reflect the ball's velocity
            dot_product = ball_velocity[0] * normal_vector[0] + ball_velocity[1] * normal_vector[1]
            ball_velocity[0] -= 2 * dot_product * normal_vector[0]
            ball_velocity[1] -= 2 * dot_product * normal_vector[1]

            # Move the ball away from the wall to prevent sticking
            ball_pos[0] += normal_vector[0] * (BALL_RADIUS - distance)
            ball_pos[1] += normal_vector[1] * (BALL_RADIUS - distance)

    # Draw everything
    screen.fill(BLACK)
    pygame.draw.polygon(screen, WHITE, triangle_vertices, 1)
    pygame.draw.circle(screen, RED, (int(ball_pos[0]), int(ball_pos[1])), int(BALL_RADIUS))
    pygame.display.flip()

pygame.quit()
``` [end of text]


llama_perf_sampler_print:    sampling time =     331.37 ms /  1585 runs   (    0.21 ms per token,  4783.22 tokens per second)
llama_perf_context_print:        load time =  872359.94 ms
llama_perf_context_print: prompt eval time =  601031.08 ms /   570 tokens ( 1054.44 ms per token,     0.95 tokens per second)
llama_perf_context_print:        eval time = 12235806.10 ms /  1014 runs   (12066.87 ms per token,     0.08 tokens per second)
llama_perf_context_print:       total time = 13109331.82 ms /  1584 tokens