task stringclasses 1
value | input stringlengths 259 305 | output stringclasses 5
values | type stringclasses 1
value | focus stringclasses 1
value | difficulty stringclasses 1
value |
|---|---|---|---|---|---|
coding | Coding Hard 1522: Implement NoeSA-24 pack 13 values into 60-bit with base-24 encoding, correct case NoeSA — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def NoeSA_Pack_13x24(values): # NoeSA-24 4.58-bit
assert len(values)==13
packed=0
for i,v in enumerate(values): # v in 0..23
packed += v * (24**i)
# 24^13 ~ 3e17 fits in 60-bit (1.15e18)
return packed.to_bytes(8,'little') # 60-bit in 8 bytes, 4.615 bits/param
Explanation: Implement... | agentic | coding | hard |
coding | Coding Hard 120: Implement SatU1 AVX2 kernel with correct case SatU1, not SATU1, for Ryzen5 5650U — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: inline __m256i SatU1_XNOR_AVX2(__m256i a, __m256i b) { __m256i xor_res = _mm256_xor_si256(a,b); return _mm256_andnot_si256(xor_res, _mm256_set1_epi32(-1)); } // XNOR = NOT XOR, then popcount via _mm256_sad_epu8 for 16 Giga/s
Explanation: Implements IKNN-Rl1-A1 tri-tier kernel for CPU-first LLM agentic researc... | agentic | coding | hard |
coding | Coding Hard 237: Implement RHT Randomized Hadamard Transform with correct case RHT, outlier flattening 10->5.07 — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def RHT_Transform(data, signs):
n=len(data)
# Random sign diagonal D
for i in range(n): data[i] *= signs[i] # +/-1
# Hadamard H_n = [H_{n-1} H_{n-1}; H_{n-1} -H_{n-1}]/sqrt(2)
h=1
while h<n:
for i in range(0,n,h*2):
for j in range(i,i+h):
u=data[j]; ... | agentic | coding | hard |
coding | Coding Hard 1306: Implement NoeSA-24 pack 13 values into 60-bit with base-24 encoding, correct case NoeSA — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def NoeSA_Pack_13x24(values): # NoeSA-24 4.58-bit
assert len(values)==13
packed=0
for i,v in enumerate(values): # v in 0..23
packed += v * (24**i)
# 24^13 ~ 3e17 fits in 60-bit (1.15e18)
return packed.to_bytes(8,'little') # 60-bit in 8 bytes, 4.615 bits/param
Explanation: Implement... | agentic | coding | hard |
coding | Coding Hard 14: Implement NoeSA-24 pack 13 values into 60-bit with base-24 encoding, correct case NoeSA — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def NoeSA_Pack_13x24(values): # NoeSA-24 4.58-bit
assert len(values)==13
packed=0
for i,v in enumerate(values): # v in 0..23
packed += v * (24**i)
# 24^13 ~ 3e17 fits in 60-bit (1.15e18)
return packed.to_bytes(8,'little') # 60-bit in 8 bytes, 4.615 bits/param
Explanation: Implement... | agentic | coding | hard |
coding | Coding Hard 906: Implement NoeSA-24 pack 13 values into 60-bit with base-24 encoding, correct case NoeSA — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def NoeSA_Pack_13x24(values): # NoeSA-24 4.58-bit
assert len(values)==13
packed=0
for i,v in enumerate(values): # v in 0..23
packed += v * (24**i)
# 24^13 ~ 3e17 fits in 60-bit (1.15e18)
return packed.to_bytes(8,'little') # 60-bit in 8 bytes, 4.615 bits/param
Explanation: Implement... | agentic | coding | hard |
coding | Coding Hard 1782: Implement NoeSA-24 pack 13 values into 60-bit with base-24 encoding, correct case NoeSA — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def NoeSA_Pack_13x24(values): # NoeSA-24 4.58-bit
assert len(values)==13
packed=0
for i,v in enumerate(values): # v in 0..23
packed += v * (24**i)
# 24^13 ~ 3e17 fits in 60-bit (1.15e18)
return packed.to_bytes(8,'little') # 60-bit in 8 bytes, 4.615 bits/param
Explanation: Implement... | agentic | coding | hard |
coding | Coding Hard 250: Implement RHT Randomized Hadamard Transform with correct case RHT, outlier flattening 10->5.07 — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def RHT_Transform(data, signs):
n=len(data)
# Random sign diagonal D
for i in range(n): data[i] *= signs[i] # +/-1
# Hadamard H_n = [H_{n-1} H_{n-1}; H_{n-1} -H_{n-1}]/sqrt(2)
h=1
while h<n:
for i in range(0,n,h*2):
for j in range(i,i+h):
u=data[j]; ... | agentic | coding | hard |
coding | Coding Hard 1689: Implement NoeSA-24 pack 13 values into 60-bit with base-24 encoding, correct case NoeSA — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def NoeSA_Pack_13x24(values): # NoeSA-24 4.58-bit
assert len(values)==13
packed=0
for i,v in enumerate(values): # v in 0..23
packed += v * (24**i)
# 24^13 ~ 3e17 fits in 60-bit (1.15e18)
return packed.to_bytes(8,'little') # 60-bit in 8 bytes, 4.615 bits/param
Explanation: Implement... | agentic | coding | hard |
coding | Coding Hard 270: Implement SatU1 AVX2 kernel with correct case SatU1, not SATU1, for Ryzen5 5650U — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: inline __m256i SatU1_XNOR_AVX2(__m256i a, __m256i b) { __m256i xor_res = _mm256_xor_si256(a,b); return _mm256_andnot_si256(xor_res, _mm256_set1_epi32(-1)); } // XNOR = NOT XOR, then popcount via _mm256_sad_epu8 for 16 Giga/s
Explanation: Implements IKNN-Rl1-A1 tri-tier kernel for CPU-first LLM agentic researc... | agentic | coding | hard |
coding | Coding Hard 1936: Implement SatU1 AVX2 kernel with correct case SatU1, not SATU1, for Ryzen5 5650U — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: inline __m256i SatU1_XNOR_AVX2(__m256i a, __m256i b) { __m256i xor_res = _mm256_xor_si256(a,b); return _mm256_andnot_si256(xor_res, _mm256_set1_epi32(-1)); } // XNOR = NOT XOR, then popcount via _mm256_sad_epu8 for 16 Giga/s
Explanation: Implements IKNN-Rl1-A1 tri-tier kernel for CPU-first LLM agentic researc... | agentic | coding | hard |
coding | Coding Hard 1208: Implement Ntarra-DnA phase rotator with correct case Ntarra-DnA, Direction NEG/ZERO/POS, Phase PHI0/1/2 shifts 0/2/4 — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def Ntarra_DnA_Compute(act, dir, phase): # Ntarra-DnA 3.17-bit
shift = {0:0, 1:2, 2:4}[phase] # PHI0 0, PHI1 2, PHI2 4
shifted = act << shift
if dir == -1: return -shifted # NEG
elif dir == 0: return 0 # ZERO
else: return shifted # POS
# Pack 2x9: 9*9=81 combos in 7-bit, 47 codes lefto... | agentic | coding | hard |
coding | Coding Hard 248: Implement SatU1 AVX2 kernel with correct case SatU1, not SATU1, for Ryzen5 5650U — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: inline __m256i SatU1_XNOR_AVX2(__m256i a, __m256i b) { __m256i xor_res = _mm256_xor_si256(a,b); return _mm256_andnot_si256(xor_res, _mm256_set1_epi32(-1)); } // XNOR = NOT XOR, then popcount via _mm256_sad_epu8 for 16 Giga/s
Explanation: Implements IKNN-Rl1-A1 tri-tier kernel for CPU-first LLM agentic researc... | agentic | coding | hard |
coding | Coding Hard 460: Implement RHT Randomized Hadamard Transform with correct case RHT, outlier flattening 10->5.07 — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def RHT_Transform(data, signs):
n=len(data)
# Random sign diagonal D
for i in range(n): data[i] *= signs[i] # +/-1
# Hadamard H_n = [H_{n-1} H_{n-1}; H_{n-1} -H_{n-1}]/sqrt(2)
h=1
while h<n:
for i in range(0,n,h*2):
for j in range(i,i+h):
u=data[j]; ... | agentic | coding | hard |
coding | Coding Hard 908: Implement SatU1 AVX2 kernel with correct case SatU1, not SATU1, for Ryzen5 5650U — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: inline __m256i SatU1_XNOR_AVX2(__m256i a, __m256i b) { __m256i xor_res = _mm256_xor_si256(a,b); return _mm256_andnot_si256(xor_res, _mm256_set1_epi32(-1)); } // XNOR = NOT XOR, then popcount via _mm256_sad_epu8 for 16 Giga/s
Explanation: Implements IKNN-Rl1-A1 tri-tier kernel for CPU-first LLM agentic researc... | agentic | coding | hard |
coding | Coding Hard 946: Implement Ntarra-DnA phase rotator with correct case Ntarra-DnA, Direction NEG/ZERO/POS, Phase PHI0/1/2 shifts 0/2/4 — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def Ntarra_DnA_Compute(act, dir, phase): # Ntarra-DnA 3.17-bit
shift = {0:0, 1:2, 2:4}[phase] # PHI0 0, PHI1 2, PHI2 4
shifted = act << shift
if dir == -1: return -shifted # NEG
elif dir == 0: return 0 # ZERO
else: return shifted # POS
# Pack 2x9: 9*9=81 combos in 7-bit, 47 codes lefto... | agentic | coding | hard |
coding | Coding Hard 1766: Implement RHT Randomized Hadamard Transform with correct case RHT, outlier flattening 10->5.07 — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def RHT_Transform(data, signs):
n=len(data)
# Random sign diagonal D
for i in range(n): data[i] *= signs[i] # +/-1
# Hadamard H_n = [H_{n-1} H_{n-1}; H_{n-1} -H_{n-1}]/sqrt(2)
h=1
while h<n:
for i in range(0,n,h*2):
for j in range(i,i+h):
u=data[j]; ... | agentic | coding | hard |
coding | Coding Hard 985: Implement NoeSA-24 pack 13 values into 60-bit with base-24 encoding, correct case NoeSA — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def NoeSA_Pack_13x24(values): # NoeSA-24 4.58-bit
assert len(values)==13
packed=0
for i,v in enumerate(values): # v in 0..23
packed += v * (24**i)
# 24^13 ~ 3e17 fits in 60-bit (1.15e18)
return packed.to_bytes(8,'little') # 60-bit in 8 bytes, 4.615 bits/param
Explanation: Implement... | agentic | coding | hard |
coding | Coding Hard 1137: Implement PG-KVC Phase-Gated KV Cache 94% saving with correct case PG-KVC — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def PG_KVC_Push(k,v,entropy):
prec = 1 if entropy<0.5 else 2 # Phase-gated: low entropy 1-bit, high 2-bit
orig = len(k)*4 + len(v)*4
comp = (len(k)+len(v))*prec//8
saving = 1 - comp/orig # 94% saving target -80% PASS
return comp, orig, saving
Explanation: Implements IKNN-Rl1-A1 tri-tier ke... | agentic | coding | hard |
coding | Coding Hard 1008: Implement PG-KVC Phase-Gated KV Cache 94% saving with correct case PG-KVC — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def PG_KVC_Push(k,v,entropy):
prec = 1 if entropy<0.5 else 2 # Phase-gated: low entropy 1-bit, high 2-bit
orig = len(k)*4 + len(v)*4
comp = (len(k)+len(v))*prec//8
saving = 1 - comp/orig # 94% saving target -80% PASS
return comp, orig, saving
Explanation: Implements IKNN-Rl1-A1 tri-tier ke... | agentic | coding | hard |
coding | Coding Hard 105: Implement SatU1 AVX2 kernel with correct case SatU1, not SATU1, for Ryzen5 5650U — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: inline __m256i SatU1_XNOR_AVX2(__m256i a, __m256i b) { __m256i xor_res = _mm256_xor_si256(a,b); return _mm256_andnot_si256(xor_res, _mm256_set1_epi32(-1)); } // XNOR = NOT XOR, then popcount via _mm256_sad_epu8 for 16 Giga/s
Explanation: Implements IKNN-Rl1-A1 tri-tier kernel for CPU-first LLM agentic researc... | agentic | coding | hard |
coding | Coding Hard 1544: Implement NoeSA-24 pack 13 values into 60-bit with base-24 encoding, correct case NoeSA — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def NoeSA_Pack_13x24(values): # NoeSA-24 4.58-bit
assert len(values)==13
packed=0
for i,v in enumerate(values): # v in 0..23
packed += v * (24**i)
# 24^13 ~ 3e17 fits in 60-bit (1.15e18)
return packed.to_bytes(8,'little') # 60-bit in 8 bytes, 4.615 bits/param
Explanation: Implement... | agentic | coding | hard |
coding | Coding Hard 1058: Implement SatU1 AVX2 kernel with correct case SatU1, not SATU1, for Ryzen5 5650U — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: inline __m256i SatU1_XNOR_AVX2(__m256i a, __m256i b) { __m256i xor_res = _mm256_xor_si256(a,b); return _mm256_andnot_si256(xor_res, _mm256_set1_epi32(-1)); } // XNOR = NOT XOR, then popcount via _mm256_sad_epu8 for 16 Giga/s
Explanation: Implements IKNN-Rl1-A1 tri-tier kernel for CPU-first LLM agentic researc... | agentic | coding | hard |
coding | Coding Hard 990: Implement NoeSA-24 pack 13 values into 60-bit with base-24 encoding, correct case NoeSA — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def NoeSA_Pack_13x24(values): # NoeSA-24 4.58-bit
assert len(values)==13
packed=0
for i,v in enumerate(values): # v in 0..23
packed += v * (24**i)
# 24^13 ~ 3e17 fits in 60-bit (1.15e18)
return packed.to_bytes(8,'little') # 60-bit in 8 bytes, 4.615 bits/param
Explanation: Implement... | agentic | coding | hard |
coding | Coding Hard 353: Implement SatU1 AVX2 kernel with correct case SatU1, not SATU1, for Ryzen5 5650U — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: inline __m256i SatU1_XNOR_AVX2(__m256i a, __m256i b) { __m256i xor_res = _mm256_xor_si256(a,b); return _mm256_andnot_si256(xor_res, _mm256_set1_epi32(-1)); } // XNOR = NOT XOR, then popcount via _mm256_sad_epu8 for 16 Giga/s
Explanation: Implements IKNN-Rl1-A1 tri-tier kernel for CPU-first LLM agentic researc... | agentic | coding | hard |
coding | Coding Hard 1: Implement Ntarra-DnA phase rotator with correct case Ntarra-DnA, Direction NEG/ZERO/POS, Phase PHI0/1/2 shifts 0/2/4 — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def Ntarra_DnA_Compute(act, dir, phase): # Ntarra-DnA 3.17-bit
shift = {0:0, 1:2, 2:4}[phase] # PHI0 0, PHI1 2, PHI2 4
shifted = act << shift
if dir == -1: return -shifted # NEG
elif dir == 0: return 0 # ZERO
else: return shifted # POS
# Pack 2x9: 9*9=81 combos in 7-bit, 47 codes lefto... | agentic | coding | hard |
coding | Coding Hard 194: Implement SatU1 AVX2 kernel with correct case SatU1, not SATU1, for Ryzen5 5650U — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: inline __m256i SatU1_XNOR_AVX2(__m256i a, __m256i b) { __m256i xor_res = _mm256_xor_si256(a,b); return _mm256_andnot_si256(xor_res, _mm256_set1_epi32(-1)); } // XNOR = NOT XOR, then popcount via _mm256_sad_epu8 for 16 Giga/s
Explanation: Implements IKNN-Rl1-A1 tri-tier kernel for CPU-first LLM agentic researc... | agentic | coding | hard |
coding | Coding Hard 1254: Implement RHT Randomized Hadamard Transform with correct case RHT, outlier flattening 10->5.07 — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def RHT_Transform(data, signs):
n=len(data)
# Random sign diagonal D
for i in range(n): data[i] *= signs[i] # +/-1
# Hadamard H_n = [H_{n-1} H_{n-1}; H_{n-1} -H_{n-1}]/sqrt(2)
h=1
while h<n:
for i in range(0,n,h*2):
for j in range(i,i+h):
u=data[j]; ... | agentic | coding | hard |
coding | Coding Hard 413: Implement SatU1 AVX2 kernel with correct case SatU1, not SATU1, for Ryzen5 5650U — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: inline __m256i SatU1_XNOR_AVX2(__m256i a, __m256i b) { __m256i xor_res = _mm256_xor_si256(a,b); return _mm256_andnot_si256(xor_res, _mm256_set1_epi32(-1)); } // XNOR = NOT XOR, then popcount via _mm256_sad_epu8 for 16 Giga/s
Explanation: Implements IKNN-Rl1-A1 tri-tier kernel for CPU-first LLM agentic researc... | agentic | coding | hard |
coding | Coding Hard 112: Implement SatU1 AVX2 kernel with correct case SatU1, not SATU1, for Ryzen5 5650U — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: inline __m256i SatU1_XNOR_AVX2(__m256i a, __m256i b) { __m256i xor_res = _mm256_xor_si256(a,b); return _mm256_andnot_si256(xor_res, _mm256_set1_epi32(-1)); } // XNOR = NOT XOR, then popcount via _mm256_sad_epu8 for 16 Giga/s
Explanation: Implements IKNN-Rl1-A1 tri-tier kernel for CPU-first LLM agentic researc... | agentic | coding | hard |
coding | Coding Hard 910: Implement RHT Randomized Hadamard Transform with correct case RHT, outlier flattening 10->5.07 — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def RHT_Transform(data, signs):
n=len(data)
# Random sign diagonal D
for i in range(n): data[i] *= signs[i] # +/-1
# Hadamard H_n = [H_{n-1} H_{n-1}; H_{n-1} -H_{n-1}]/sqrt(2)
h=1
while h<n:
for i in range(0,n,h*2):
for j in range(i,i+h):
u=data[j]; ... | agentic | coding | hard |
coding | Coding Hard 715: Implement Ntarra-DnA phase rotator with correct case Ntarra-DnA, Direction NEG/ZERO/POS, Phase PHI0/1/2 shifts 0/2/4 — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def Ntarra_DnA_Compute(act, dir, phase): # Ntarra-DnA 3.17-bit
shift = {0:0, 1:2, 2:4}[phase] # PHI0 0, PHI1 2, PHI2 4
shifted = act << shift
if dir == -1: return -shifted # NEG
elif dir == 0: return 0 # ZERO
else: return shifted # POS
# Pack 2x9: 9*9=81 combos in 7-bit, 47 codes lefto... | agentic | coding | hard |
coding | Coding Hard 432: Implement NoeSA-24 pack 13 values into 60-bit with base-24 encoding, correct case NoeSA — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def NoeSA_Pack_13x24(values): # NoeSA-24 4.58-bit
assert len(values)==13
packed=0
for i,v in enumerate(values): # v in 0..23
packed += v * (24**i)
# 24^13 ~ 3e17 fits in 60-bit (1.15e18)
return packed.to_bytes(8,'little') # 60-bit in 8 bytes, 4.615 bits/param
Explanation: Implement... | agentic | coding | hard |
coding | Coding Hard 747: Implement RHT Randomized Hadamard Transform with correct case RHT, outlier flattening 10->5.07 — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def RHT_Transform(data, signs):
n=len(data)
# Random sign diagonal D
for i in range(n): data[i] *= signs[i] # +/-1
# Hadamard H_n = [H_{n-1} H_{n-1}; H_{n-1} -H_{n-1}]/sqrt(2)
h=1
while h<n:
for i in range(0,n,h*2):
for j in range(i,i+h):
u=data[j]; ... | agentic | coding | hard |
coding | Coding Hard 591: Implement NoeSA-24 pack 13 values into 60-bit with base-24 encoding, correct case NoeSA — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def NoeSA_Pack_13x24(values): # NoeSA-24 4.58-bit
assert len(values)==13
packed=0
for i,v in enumerate(values): # v in 0..23
packed += v * (24**i)
# 24^13 ~ 3e17 fits in 60-bit (1.15e18)
return packed.to_bytes(8,'little') # 60-bit in 8 bytes, 4.615 bits/param
Explanation: Implement... | agentic | coding | hard |
coding | Coding Hard 865: Implement NoeSA-24 pack 13 values into 60-bit with base-24 encoding, correct case NoeSA — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def NoeSA_Pack_13x24(values): # NoeSA-24 4.58-bit
assert len(values)==13
packed=0
for i,v in enumerate(values): # v in 0..23
packed += v * (24**i)
# 24^13 ~ 3e17 fits in 60-bit (1.15e18)
return packed.to_bytes(8,'little') # 60-bit in 8 bytes, 4.615 bits/param
Explanation: Implement... | agentic | coding | hard |
coding | Coding Hard 577: Implement Ntarra-DnA phase rotator with correct case Ntarra-DnA, Direction NEG/ZERO/POS, Phase PHI0/1/2 shifts 0/2/4 — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def Ntarra_DnA_Compute(act, dir, phase): # Ntarra-DnA 3.17-bit
shift = {0:0, 1:2, 2:4}[phase] # PHI0 0, PHI1 2, PHI2 4
shifted = act << shift
if dir == -1: return -shifted # NEG
elif dir == 0: return 0 # ZERO
else: return shifted # POS
# Pack 2x9: 9*9=81 combos in 7-bit, 47 codes lefto... | agentic | coding | hard |
coding | Coding Hard 1013: Implement PG-KVC Phase-Gated KV Cache 94% saving with correct case PG-KVC — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def PG_KVC_Push(k,v,entropy):
prec = 1 if entropy<0.5 else 2 # Phase-gated: low entropy 1-bit, high 2-bit
orig = len(k)*4 + len(v)*4
comp = (len(k)+len(v))*prec//8
saving = 1 - comp/orig # 94% saving target -80% PASS
return comp, orig, saving
Explanation: Implements IKNN-Rl1-A1 tri-tier ke... | agentic | coding | hard |
coding | Coding Hard 1749: Implement PG-KVC Phase-Gated KV Cache 94% saving with correct case PG-KVC — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def PG_KVC_Push(k,v,entropy):
prec = 1 if entropy<0.5 else 2 # Phase-gated: low entropy 1-bit, high 2-bit
orig = len(k)*4 + len(v)*4
comp = (len(k)+len(v))*prec//8
saving = 1 - comp/orig # 94% saving target -80% PASS
return comp, orig, saving
Explanation: Implements IKNN-Rl1-A1 tri-tier ke... | agentic | coding | hard |
coding | Coding Hard 1223: Implement PG-KVC Phase-Gated KV Cache 94% saving with correct case PG-KVC — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def PG_KVC_Push(k,v,entropy):
prec = 1 if entropy<0.5 else 2 # Phase-gated: low entropy 1-bit, high 2-bit
orig = len(k)*4 + len(v)*4
comp = (len(k)+len(v))*prec//8
saving = 1 - comp/orig # 94% saving target -80% PASS
return comp, orig, saving
Explanation: Implements IKNN-Rl1-A1 tri-tier ke... | agentic | coding | hard |
coding | Coding Hard 880: Implement Ntarra-DnA phase rotator with correct case Ntarra-DnA, Direction NEG/ZERO/POS, Phase PHI0/1/2 shifts 0/2/4 — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def Ntarra_DnA_Compute(act, dir, phase): # Ntarra-DnA 3.17-bit
shift = {0:0, 1:2, 2:4}[phase] # PHI0 0, PHI1 2, PHI2 4
shifted = act << shift
if dir == -1: return -shifted # NEG
elif dir == 0: return 0 # ZERO
else: return shifted # POS
# Pack 2x9: 9*9=81 combos in 7-bit, 47 codes lefto... | agentic | coding | hard |
coding | Coding Hard 1582: Implement Ntarra-DnA phase rotator with correct case Ntarra-DnA, Direction NEG/ZERO/POS, Phase PHI0/1/2 shifts 0/2/4 — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def Ntarra_DnA_Compute(act, dir, phase): # Ntarra-DnA 3.17-bit
shift = {0:0, 1:2, 2:4}[phase] # PHI0 0, PHI1 2, PHI2 4
shifted = act << shift
if dir == -1: return -shifted # NEG
elif dir == 0: return 0 # ZERO
else: return shifted # POS
# Pack 2x9: 9*9=81 combos in 7-bit, 47 codes lefto... | agentic | coding | hard |
coding | Coding Hard 1309: Implement SatU1 AVX2 kernel with correct case SatU1, not SATU1, for Ryzen5 5650U — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: inline __m256i SatU1_XNOR_AVX2(__m256i a, __m256i b) { __m256i xor_res = _mm256_xor_si256(a,b); return _mm256_andnot_si256(xor_res, _mm256_set1_epi32(-1)); } // XNOR = NOT XOR, then popcount via _mm256_sad_epu8 for 16 Giga/s
Explanation: Implements IKNN-Rl1-A1 tri-tier kernel for CPU-first LLM agentic researc... | agentic | coding | hard |
coding | Coding Hard 380: Implement NoeSA-24 pack 13 values into 60-bit with base-24 encoding, correct case NoeSA — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def NoeSA_Pack_13x24(values): # NoeSA-24 4.58-bit
assert len(values)==13
packed=0
for i,v in enumerate(values): # v in 0..23
packed += v * (24**i)
# 24^13 ~ 3e17 fits in 60-bit (1.15e18)
return packed.to_bytes(8,'little') # 60-bit in 8 bytes, 4.615 bits/param
Explanation: Implement... | agentic | coding | hard |
coding | Coding Hard 1467: Implement PG-KVC Phase-Gated KV Cache 94% saving with correct case PG-KVC — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def PG_KVC_Push(k,v,entropy):
prec = 1 if entropy<0.5 else 2 # Phase-gated: low entropy 1-bit, high 2-bit
orig = len(k)*4 + len(v)*4
comp = (len(k)+len(v))*prec//8
saving = 1 - comp/orig # 94% saving target -80% PASS
return comp, orig, saving
Explanation: Implements IKNN-Rl1-A1 tri-tier ke... | agentic | coding | hard |
coding | Coding Hard 304: Implement SatU1 AVX2 kernel with correct case SatU1, not SATU1, for Ryzen5 5650U — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: inline __m256i SatU1_XNOR_AVX2(__m256i a, __m256i b) { __m256i xor_res = _mm256_xor_si256(a,b); return _mm256_andnot_si256(xor_res, _mm256_set1_epi32(-1)); } // XNOR = NOT XOR, then popcount via _mm256_sad_epu8 for 16 Giga/s
Explanation: Implements IKNN-Rl1-A1 tri-tier kernel for CPU-first LLM agentic researc... | agentic | coding | hard |
coding | Coding Hard 388: Implement Ntarra-DnA phase rotator with correct case Ntarra-DnA, Direction NEG/ZERO/POS, Phase PHI0/1/2 shifts 0/2/4 — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def Ntarra_DnA_Compute(act, dir, phase): # Ntarra-DnA 3.17-bit
shift = {0:0, 1:2, 2:4}[phase] # PHI0 0, PHI1 2, PHI2 4
shifted = act << shift
if dir == -1: return -shifted # NEG
elif dir == 0: return 0 # ZERO
else: return shifted # POS
# Pack 2x9: 9*9=81 combos in 7-bit, 47 codes lefto... | agentic | coding | hard |
coding | Coding Hard 912: Implement SatU1 AVX2 kernel with correct case SatU1, not SATU1, for Ryzen5 5650U — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: inline __m256i SatU1_XNOR_AVX2(__m256i a, __m256i b) { __m256i xor_res = _mm256_xor_si256(a,b); return _mm256_andnot_si256(xor_res, _mm256_set1_epi32(-1)); } // XNOR = NOT XOR, then popcount via _mm256_sad_epu8 for 16 Giga/s
Explanation: Implements IKNN-Rl1-A1 tri-tier kernel for CPU-first LLM agentic researc... | agentic | coding | hard |
coding | Coding Hard 200: Implement SatU1 AVX2 kernel with correct case SatU1, not SATU1, for Ryzen5 5650U — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: inline __m256i SatU1_XNOR_AVX2(__m256i a, __m256i b) { __m256i xor_res = _mm256_xor_si256(a,b); return _mm256_andnot_si256(xor_res, _mm256_set1_epi32(-1)); } // XNOR = NOT XOR, then popcount via _mm256_sad_epu8 for 16 Giga/s
Explanation: Implements IKNN-Rl1-A1 tri-tier kernel for CPU-first LLM agentic researc... | agentic | coding | hard |
coding | Coding Hard 1346: Implement NoeSA-24 pack 13 values into 60-bit with base-24 encoding, correct case NoeSA — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def NoeSA_Pack_13x24(values): # NoeSA-24 4.58-bit
assert len(values)==13
packed=0
for i,v in enumerate(values): # v in 0..23
packed += v * (24**i)
# 24^13 ~ 3e17 fits in 60-bit (1.15e18)
return packed.to_bytes(8,'little') # 60-bit in 8 bytes, 4.615 bits/param
Explanation: Implement... | agentic | coding | hard |
coding | Coding Hard 804: Implement RHT Randomized Hadamard Transform with correct case RHT, outlier flattening 10->5.07 — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def RHT_Transform(data, signs):
n=len(data)
# Random sign diagonal D
for i in range(n): data[i] *= signs[i] # +/-1
# Hadamard H_n = [H_{n-1} H_{n-1}; H_{n-1} -H_{n-1}]/sqrt(2)
h=1
while h<n:
for i in range(0,n,h*2):
for j in range(i,i+h):
u=data[j]; ... | agentic | coding | hard |
coding | Coding Hard 1255: Implement NoeSA-24 pack 13 values into 60-bit with base-24 encoding, correct case NoeSA — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def NoeSA_Pack_13x24(values): # NoeSA-24 4.58-bit
assert len(values)==13
packed=0
for i,v in enumerate(values): # v in 0..23
packed += v * (24**i)
# 24^13 ~ 3e17 fits in 60-bit (1.15e18)
return packed.to_bytes(8,'little') # 60-bit in 8 bytes, 4.615 bits/param
Explanation: Implement... | agentic | coding | hard |
coding | Coding Hard 1534: Implement PG-KVC Phase-Gated KV Cache 94% saving with correct case PG-KVC — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def PG_KVC_Push(k,v,entropy):
prec = 1 if entropy<0.5 else 2 # Phase-gated: low entropy 1-bit, high 2-bit
orig = len(k)*4 + len(v)*4
comp = (len(k)+len(v))*prec//8
saving = 1 - comp/orig # 94% saving target -80% PASS
return comp, orig, saving
Explanation: Implements IKNN-Rl1-A1 tri-tier ke... | agentic | coding | hard |
coding | Coding Hard 1812: Implement PG-KVC Phase-Gated KV Cache 94% saving with correct case PG-KVC — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def PG_KVC_Push(k,v,entropy):
prec = 1 if entropy<0.5 else 2 # Phase-gated: low entropy 1-bit, high 2-bit
orig = len(k)*4 + len(v)*4
comp = (len(k)+len(v))*prec//8
saving = 1 - comp/orig # 94% saving target -80% PASS
return comp, orig, saving
Explanation: Implements IKNN-Rl1-A1 tri-tier ke... | agentic | coding | hard |
coding | Coding Hard 1312: Implement NoeSA-24 pack 13 values into 60-bit with base-24 encoding, correct case NoeSA — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def NoeSA_Pack_13x24(values): # NoeSA-24 4.58-bit
assert len(values)==13
packed=0
for i,v in enumerate(values): # v in 0..23
packed += v * (24**i)
# 24^13 ~ 3e17 fits in 60-bit (1.15e18)
return packed.to_bytes(8,'little') # 60-bit in 8 bytes, 4.615 bits/param
Explanation: Implement... | agentic | coding | hard |
coding | Coding Hard 186: Implement SatU1 AVX2 kernel with correct case SatU1, not SATU1, for Ryzen5 5650U — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: inline __m256i SatU1_XNOR_AVX2(__m256i a, __m256i b) { __m256i xor_res = _mm256_xor_si256(a,b); return _mm256_andnot_si256(xor_res, _mm256_set1_epi32(-1)); } // XNOR = NOT XOR, then popcount via _mm256_sad_epu8 for 16 Giga/s
Explanation: Implements IKNN-Rl1-A1 tri-tier kernel for CPU-first LLM agentic researc... | agentic | coding | hard |
coding | Coding Hard 255: Implement RHT Randomized Hadamard Transform with correct case RHT, outlier flattening 10->5.07 — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def RHT_Transform(data, signs):
n=len(data)
# Random sign diagonal D
for i in range(n): data[i] *= signs[i] # +/-1
# Hadamard H_n = [H_{n-1} H_{n-1}; H_{n-1} -H_{n-1}]/sqrt(2)
h=1
while h<n:
for i in range(0,n,h*2):
for j in range(i,i+h):
u=data[j]; ... | agentic | coding | hard |
coding | Coding Hard 931: Implement Ntarra-DnA phase rotator with correct case Ntarra-DnA, Direction NEG/ZERO/POS, Phase PHI0/1/2 shifts 0/2/4 — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def Ntarra_DnA_Compute(act, dir, phase): # Ntarra-DnA 3.17-bit
shift = {0:0, 1:2, 2:4}[phase] # PHI0 0, PHI1 2, PHI2 4
shifted = act << shift
if dir == -1: return -shifted # NEG
elif dir == 0: return 0 # ZERO
else: return shifted # POS
# Pack 2x9: 9*9=81 combos in 7-bit, 47 codes lefto... | agentic | coding | hard |
coding | Coding Hard 92: Implement PG-KVC Phase-Gated KV Cache 94% saving with correct case PG-KVC — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def PG_KVC_Push(k,v,entropy):
prec = 1 if entropy<0.5 else 2 # Phase-gated: low entropy 1-bit, high 2-bit
orig = len(k)*4 + len(v)*4
comp = (len(k)+len(v))*prec//8
saving = 1 - comp/orig # 94% saving target -80% PASS
return comp, orig, saving
Explanation: Implements IKNN-Rl1-A1 tri-tier ke... | agentic | coding | hard |
coding | Coding Hard 1367: Implement PG-KVC Phase-Gated KV Cache 94% saving with correct case PG-KVC — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def PG_KVC_Push(k,v,entropy):
prec = 1 if entropy<0.5 else 2 # Phase-gated: low entropy 1-bit, high 2-bit
orig = len(k)*4 + len(v)*4
comp = (len(k)+len(v))*prec//8
saving = 1 - comp/orig # 94% saving target -80% PASS
return comp, orig, saving
Explanation: Implements IKNN-Rl1-A1 tri-tier ke... | agentic | coding | hard |
coding | Coding Hard 1646: Implement NoeSA-24 pack 13 values into 60-bit with base-24 encoding, correct case NoeSA — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def NoeSA_Pack_13x24(values): # NoeSA-24 4.58-bit
assert len(values)==13
packed=0
for i,v in enumerate(values): # v in 0..23
packed += v * (24**i)
# 24^13 ~ 3e17 fits in 60-bit (1.15e18)
return packed.to_bytes(8,'little') # 60-bit in 8 bytes, 4.615 bits/param
Explanation: Implement... | agentic | coding | hard |
coding | Coding Hard 500: Implement SatU1 AVX2 kernel with correct case SatU1, not SATU1, for Ryzen5 5650U — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: inline __m256i SatU1_XNOR_AVX2(__m256i a, __m256i b) { __m256i xor_res = _mm256_xor_si256(a,b); return _mm256_andnot_si256(xor_res, _mm256_set1_epi32(-1)); } // XNOR = NOT XOR, then popcount via _mm256_sad_epu8 for 16 Giga/s
Explanation: Implements IKNN-Rl1-A1 tri-tier kernel for CPU-first LLM agentic researc... | agentic | coding | hard |
coding | Coding Hard 1027: Implement SatU1 AVX2 kernel with correct case SatU1, not SATU1, for Ryzen5 5650U — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: inline __m256i SatU1_XNOR_AVX2(__m256i a, __m256i b) { __m256i xor_res = _mm256_xor_si256(a,b); return _mm256_andnot_si256(xor_res, _mm256_set1_epi32(-1)); } // XNOR = NOT XOR, then popcount via _mm256_sad_epu8 for 16 Giga/s
Explanation: Implements IKNN-Rl1-A1 tri-tier kernel for CPU-first LLM agentic researc... | agentic | coding | hard |
coding | Coding Hard 1654: Implement PG-KVC Phase-Gated KV Cache 94% saving with correct case PG-KVC — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def PG_KVC_Push(k,v,entropy):
prec = 1 if entropy<0.5 else 2 # Phase-gated: low entropy 1-bit, high 2-bit
orig = len(k)*4 + len(v)*4
comp = (len(k)+len(v))*prec//8
saving = 1 - comp/orig # 94% saving target -80% PASS
return comp, orig, saving
Explanation: Implements IKNN-Rl1-A1 tri-tier ke... | agentic | coding | hard |
coding | Coding Hard 126: Implement SatU1 AVX2 kernel with correct case SatU1, not SATU1, for Ryzen5 5650U — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: inline __m256i SatU1_XNOR_AVX2(__m256i a, __m256i b) { __m256i xor_res = _mm256_xor_si256(a,b); return _mm256_andnot_si256(xor_res, _mm256_set1_epi32(-1)); } // XNOR = NOT XOR, then popcount via _mm256_sad_epu8 for 16 Giga/s
Explanation: Implements IKNN-Rl1-A1 tri-tier kernel for CPU-first LLM agentic researc... | agentic | coding | hard |
coding | Coding Hard 1318: Implement RHT Randomized Hadamard Transform with correct case RHT, outlier flattening 10->5.07 — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def RHT_Transform(data, signs):
n=len(data)
# Random sign diagonal D
for i in range(n): data[i] *= signs[i] # +/-1
# Hadamard H_n = [H_{n-1} H_{n-1}; H_{n-1} -H_{n-1}]/sqrt(2)
h=1
while h<n:
for i in range(0,n,h*2):
for j in range(i,i+h):
u=data[j]; ... | agentic | coding | hard |
coding | Coding Hard 1215: Implement PG-KVC Phase-Gated KV Cache 94% saving with correct case PG-KVC — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def PG_KVC_Push(k,v,entropy):
prec = 1 if entropy<0.5 else 2 # Phase-gated: low entropy 1-bit, high 2-bit
orig = len(k)*4 + len(v)*4
comp = (len(k)+len(v))*prec//8
saving = 1 - comp/orig # 94% saving target -80% PASS
return comp, orig, saving
Explanation: Implements IKNN-Rl1-A1 tri-tier ke... | agentic | coding | hard |
coding | Coding Hard 1259: Implement PG-KVC Phase-Gated KV Cache 94% saving with correct case PG-KVC — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def PG_KVC_Push(k,v,entropy):
prec = 1 if entropy<0.5 else 2 # Phase-gated: low entropy 1-bit, high 2-bit
orig = len(k)*4 + len(v)*4
comp = (len(k)+len(v))*prec//8
saving = 1 - comp/orig # 94% saving target -80% PASS
return comp, orig, saving
Explanation: Implements IKNN-Rl1-A1 tri-tier ke... | agentic | coding | hard |
coding | Coding Hard 514: Implement Ntarra-DnA phase rotator with correct case Ntarra-DnA, Direction NEG/ZERO/POS, Phase PHI0/1/2 shifts 0/2/4 — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def Ntarra_DnA_Compute(act, dir, phase): # Ntarra-DnA 3.17-bit
shift = {0:0, 1:2, 2:4}[phase] # PHI0 0, PHI1 2, PHI2 4
shifted = act << shift
if dir == -1: return -shifted # NEG
elif dir == 0: return 0 # ZERO
else: return shifted # POS
# Pack 2x9: 9*9=81 combos in 7-bit, 47 codes lefto... | agentic | coding | hard |
coding | Coding Hard 499: Implement Ntarra-DnA phase rotator with correct case Ntarra-DnA, Direction NEG/ZERO/POS, Phase PHI0/1/2 shifts 0/2/4 — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def Ntarra_DnA_Compute(act, dir, phase): # Ntarra-DnA 3.17-bit
shift = {0:0, 1:2, 2:4}[phase] # PHI0 0, PHI1 2, PHI2 4
shifted = act << shift
if dir == -1: return -shifted # NEG
elif dir == 0: return 0 # ZERO
else: return shifted # POS
# Pack 2x9: 9*9=81 combos in 7-bit, 47 codes lefto... | agentic | coding | hard |
coding | Coding Hard 1571: Implement PG-KVC Phase-Gated KV Cache 94% saving with correct case PG-KVC — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def PG_KVC_Push(k,v,entropy):
prec = 1 if entropy<0.5 else 2 # Phase-gated: low entropy 1-bit, high 2-bit
orig = len(k)*4 + len(v)*4
comp = (len(k)+len(v))*prec//8
saving = 1 - comp/orig # 94% saving target -80% PASS
return comp, orig, saving
Explanation: Implements IKNN-Rl1-A1 tri-tier ke... | agentic | coding | hard |
coding | Coding Hard 497: Implement PG-KVC Phase-Gated KV Cache 94% saving with correct case PG-KVC — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def PG_KVC_Push(k,v,entropy):
prec = 1 if entropy<0.5 else 2 # Phase-gated: low entropy 1-bit, high 2-bit
orig = len(k)*4 + len(v)*4
comp = (len(k)+len(v))*prec//8
saving = 1 - comp/orig # 94% saving target -80% PASS
return comp, orig, saving
Explanation: Implements IKNN-Rl1-A1 tri-tier ke... | agentic | coding | hard |
coding | Coding Hard 151: Implement Ntarra-DnA phase rotator with correct case Ntarra-DnA, Direction NEG/ZERO/POS, Phase PHI0/1/2 shifts 0/2/4 — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def Ntarra_DnA_Compute(act, dir, phase): # Ntarra-DnA 3.17-bit
shift = {0:0, 1:2, 2:4}[phase] # PHI0 0, PHI1 2, PHI2 4
shifted = act << shift
if dir == -1: return -shifted # NEG
elif dir == 0: return 0 # ZERO
else: return shifted # POS
# Pack 2x9: 9*9=81 combos in 7-bit, 47 codes lefto... | agentic | coding | hard |
coding | Coding Hard 1834: Implement Ntarra-DnA phase rotator with correct case Ntarra-DnA, Direction NEG/ZERO/POS, Phase PHI0/1/2 shifts 0/2/4 — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def Ntarra_DnA_Compute(act, dir, phase): # Ntarra-DnA 3.17-bit
shift = {0:0, 1:2, 2:4}[phase] # PHI0 0, PHI1 2, PHI2 4
shifted = act << shift
if dir == -1: return -shifted # NEG
elif dir == 0: return 0 # ZERO
else: return shifted # POS
# Pack 2x9: 9*9=81 combos in 7-bit, 47 codes lefto... | agentic | coding | hard |
coding | Coding Hard 1059: Implement SatU1 AVX2 kernel with correct case SatU1, not SATU1, for Ryzen5 5650U — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: inline __m256i SatU1_XNOR_AVX2(__m256i a, __m256i b) { __m256i xor_res = _mm256_xor_si256(a,b); return _mm256_andnot_si256(xor_res, _mm256_set1_epi32(-1)); } // XNOR = NOT XOR, then popcount via _mm256_sad_epu8 for 16 Giga/s
Explanation: Implements IKNN-Rl1-A1 tri-tier kernel for CPU-first LLM agentic researc... | agentic | coding | hard |
coding | Coding Hard 387: Implement SatU1 AVX2 kernel with correct case SatU1, not SATU1, for Ryzen5 5650U — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: inline __m256i SatU1_XNOR_AVX2(__m256i a, __m256i b) { __m256i xor_res = _mm256_xor_si256(a,b); return _mm256_andnot_si256(xor_res, _mm256_set1_epi32(-1)); } // XNOR = NOT XOR, then popcount via _mm256_sad_epu8 for 16 Giga/s
Explanation: Implements IKNN-Rl1-A1 tri-tier kernel for CPU-first LLM agentic researc... | agentic | coding | hard |
coding | Coding Hard 734: Implement RHT Randomized Hadamard Transform with correct case RHT, outlier flattening 10->5.07 — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def RHT_Transform(data, signs):
n=len(data)
# Random sign diagonal D
for i in range(n): data[i] *= signs[i] # +/-1
# Hadamard H_n = [H_{n-1} H_{n-1}; H_{n-1} -H_{n-1}]/sqrt(2)
h=1
while h<n:
for i in range(0,n,h*2):
for j in range(i,i+h):
u=data[j]; ... | agentic | coding | hard |
coding | Coding Hard 1801: Implement PG-KVC Phase-Gated KV Cache 94% saving with correct case PG-KVC — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def PG_KVC_Push(k,v,entropy):
prec = 1 if entropy<0.5 else 2 # Phase-gated: low entropy 1-bit, high 2-bit
orig = len(k)*4 + len(v)*4
comp = (len(k)+len(v))*prec//8
saving = 1 - comp/orig # 94% saving target -80% PASS
return comp, orig, saving
Explanation: Implements IKNN-Rl1-A1 tri-tier ke... | agentic | coding | hard |
coding | Coding Hard 1232: Implement NoeSA-24 pack 13 values into 60-bit with base-24 encoding, correct case NoeSA — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def NoeSA_Pack_13x24(values): # NoeSA-24 4.58-bit
assert len(values)==13
packed=0
for i,v in enumerate(values): # v in 0..23
packed += v * (24**i)
# 24^13 ~ 3e17 fits in 60-bit (1.15e18)
return packed.to_bytes(8,'little') # 60-bit in 8 bytes, 4.615 bits/param
Explanation: Implement... | agentic | coding | hard |
coding | Coding Hard 1400: Implement SatU1 AVX2 kernel with correct case SatU1, not SATU1, for Ryzen5 5650U — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: inline __m256i SatU1_XNOR_AVX2(__m256i a, __m256i b) { __m256i xor_res = _mm256_xor_si256(a,b); return _mm256_andnot_si256(xor_res, _mm256_set1_epi32(-1)); } // XNOR = NOT XOR, then popcount via _mm256_sad_epu8 for 16 Giga/s
Explanation: Implements IKNN-Rl1-A1 tri-tier kernel for CPU-first LLM agentic researc... | agentic | coding | hard |
coding | Coding Hard 127: Implement Ntarra-DnA phase rotator with correct case Ntarra-DnA, Direction NEG/ZERO/POS, Phase PHI0/1/2 shifts 0/2/4 — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def Ntarra_DnA_Compute(act, dir, phase): # Ntarra-DnA 3.17-bit
shift = {0:0, 1:2, 2:4}[phase] # PHI0 0, PHI1 2, PHI2 4
shifted = act << shift
if dir == -1: return -shifted # NEG
elif dir == 0: return 0 # ZERO
else: return shifted # POS
# Pack 2x9: 9*9=81 combos in 7-bit, 47 codes lefto... | agentic | coding | hard |
coding | Coding Hard 1789: Implement RHT Randomized Hadamard Transform with correct case RHT, outlier flattening 10->5.07 — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def RHT_Transform(data, signs):
n=len(data)
# Random sign diagonal D
for i in range(n): data[i] *= signs[i] # +/-1
# Hadamard H_n = [H_{n-1} H_{n-1}; H_{n-1} -H_{n-1}]/sqrt(2)
h=1
while h<n:
for i in range(0,n,h*2):
for j in range(i,i+h):
u=data[j]; ... | agentic | coding | hard |
coding | Coding Hard 1776: Implement PG-KVC Phase-Gated KV Cache 94% saving with correct case PG-KVC — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def PG_KVC_Push(k,v,entropy):
prec = 1 if entropy<0.5 else 2 # Phase-gated: low entropy 1-bit, high 2-bit
orig = len(k)*4 + len(v)*4
comp = (len(k)+len(v))*prec//8
saving = 1 - comp/orig # 94% saving target -80% PASS
return comp, orig, saving
Explanation: Implements IKNN-Rl1-A1 tri-tier ke... | agentic | coding | hard |
coding | Coding Hard 1114: Implement NoeSA-24 pack 13 values into 60-bit with base-24 encoding, correct case NoeSA — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def NoeSA_Pack_13x24(values): # NoeSA-24 4.58-bit
assert len(values)==13
packed=0
for i,v in enumerate(values): # v in 0..23
packed += v * (24**i)
# 24^13 ~ 3e17 fits in 60-bit (1.15e18)
return packed.to_bytes(8,'little') # 60-bit in 8 bytes, 4.615 bits/param
Explanation: Implement... | agentic | coding | hard |
coding | Coding Hard 1901: Implement RHT Randomized Hadamard Transform with correct case RHT, outlier flattening 10->5.07 — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def RHT_Transform(data, signs):
n=len(data)
# Random sign diagonal D
for i in range(n): data[i] *= signs[i] # +/-1
# Hadamard H_n = [H_{n-1} H_{n-1}; H_{n-1} -H_{n-1}]/sqrt(2)
h=1
while h<n:
for i in range(0,n,h*2):
for j in range(i,i+h):
u=data[j]; ... | agentic | coding | hard |
coding | Coding Hard 412: Implement PG-KVC Phase-Gated KV Cache 94% saving with correct case PG-KVC — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def PG_KVC_Push(k,v,entropy):
prec = 1 if entropy<0.5 else 2 # Phase-gated: low entropy 1-bit, high 2-bit
orig = len(k)*4 + len(v)*4
comp = (len(k)+len(v))*prec//8
saving = 1 - comp/orig # 94% saving target -80% PASS
return comp, orig, saving
Explanation: Implements IKNN-Rl1-A1 tri-tier ke... | agentic | coding | hard |
coding | Coding Hard 374: Implement PG-KVC Phase-Gated KV Cache 94% saving with correct case PG-KVC — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def PG_KVC_Push(k,v,entropy):
prec = 1 if entropy<0.5 else 2 # Phase-gated: low entropy 1-bit, high 2-bit
orig = len(k)*4 + len(v)*4
comp = (len(k)+len(v))*prec//8
saving = 1 - comp/orig # 94% saving target -80% PASS
return comp, orig, saving
Explanation: Implements IKNN-Rl1-A1 tri-tier ke... | agentic | coding | hard |
coding | Coding Hard 490: Implement SatU1 AVX2 kernel with correct case SatU1, not SATU1, for Ryzen5 5650U — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: inline __m256i SatU1_XNOR_AVX2(__m256i a, __m256i b) { __m256i xor_res = _mm256_xor_si256(a,b); return _mm256_andnot_si256(xor_res, _mm256_set1_epi32(-1)); } // XNOR = NOT XOR, then popcount via _mm256_sad_epu8 for 16 Giga/s
Explanation: Implements IKNN-Rl1-A1 tri-tier kernel for CPU-first LLM agentic researc... | agentic | coding | hard |
coding | Coding Hard 1927: Implement SatU1 AVX2 kernel with correct case SatU1, not SATU1, for Ryzen5 5650U — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: inline __m256i SatU1_XNOR_AVX2(__m256i a, __m256i b) { __m256i xor_res = _mm256_xor_si256(a,b); return _mm256_andnot_si256(xor_res, _mm256_set1_epi32(-1)); } // XNOR = NOT XOR, then popcount via _mm256_sad_epu8 for 16 Giga/s
Explanation: Implements IKNN-Rl1-A1 tri-tier kernel for CPU-first LLM agentic researc... | agentic | coding | hard |
coding | Coding Hard 1629: Implement Ntarra-DnA phase rotator with correct case Ntarra-DnA, Direction NEG/ZERO/POS, Phase PHI0/1/2 shifts 0/2/4 — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def Ntarra_DnA_Compute(act, dir, phase): # Ntarra-DnA 3.17-bit
shift = {0:0, 1:2, 2:4}[phase] # PHI0 0, PHI1 2, PHI2 4
shifted = act << shift
if dir == -1: return -shifted # NEG
elif dir == 0: return 0 # ZERO
else: return shifted # POS
# Pack 2x9: 9*9=81 combos in 7-bit, 47 codes lefto... | agentic | coding | hard |
coding | Coding Hard 431: Implement RHT Randomized Hadamard Transform with correct case RHT, outlier flattening 10->5.07 — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def RHT_Transform(data, signs):
n=len(data)
# Random sign diagonal D
for i in range(n): data[i] *= signs[i] # +/-1
# Hadamard H_n = [H_{n-1} H_{n-1}; H_{n-1} -H_{n-1}]/sqrt(2)
h=1
while h<n:
for i in range(0,n,h*2):
for j in range(i,i+h):
u=data[j]; ... | agentic | coding | hard |
coding | Coding Hard 1277: Implement RHT Randomized Hadamard Transform with correct case RHT, outlier flattening 10->5.07 — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def RHT_Transform(data, signs):
n=len(data)
# Random sign diagonal D
for i in range(n): data[i] *= signs[i] # +/-1
# Hadamard H_n = [H_{n-1} H_{n-1}; H_{n-1} -H_{n-1}]/sqrt(2)
h=1
while h<n:
for i in range(0,n,h*2):
for j in range(i,i+h):
u=data[j]; ... | agentic | coding | hard |
coding | Coding Hard 10: Implement Ntarra-DnA phase rotator with correct case Ntarra-DnA, Direction NEG/ZERO/POS, Phase PHI0/1/2 shifts 0/2/4 — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def Ntarra_DnA_Compute(act, dir, phase): # Ntarra-DnA 3.17-bit
shift = {0:0, 1:2, 2:4}[phase] # PHI0 0, PHI1 2, PHI2 4
shifted = act << shift
if dir == -1: return -shifted # NEG
elif dir == 0: return 0 # ZERO
else: return shifted # POS
# Pack 2x9: 9*9=81 combos in 7-bit, 47 codes lefto... | agentic | coding | hard |
coding | Coding Hard 375: Implement SatU1 AVX2 kernel with correct case SatU1, not SATU1, for Ryzen5 5650U — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: inline __m256i SatU1_XNOR_AVX2(__m256i a, __m256i b) { __m256i xor_res = _mm256_xor_si256(a,b); return _mm256_andnot_si256(xor_res, _mm256_set1_epi32(-1)); } // XNOR = NOT XOR, then popcount via _mm256_sad_epu8 for 16 Giga/s
Explanation: Implements IKNN-Rl1-A1 tri-tier kernel for CPU-first LLM agentic researc... | agentic | coding | hard |
coding | Coding Hard 169: Implement RHT Randomized Hadamard Transform with correct case RHT, outlier flattening 10->5.07 — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def RHT_Transform(data, signs):
n=len(data)
# Random sign diagonal D
for i in range(n): data[i] *= signs[i] # +/-1
# Hadamard H_n = [H_{n-1} H_{n-1}; H_{n-1} -H_{n-1}]/sqrt(2)
h=1
while h<n:
for i in range(0,n,h*2):
for j in range(i,i+h):
u=data[j]; ... | agentic | coding | hard |
coding | Coding Hard 1532: Implement PG-KVC Phase-Gated KV Cache 94% saving with correct case PG-KVC — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def PG_KVC_Push(k,v,entropy):
prec = 1 if entropy<0.5 else 2 # Phase-gated: low entropy 1-bit, high 2-bit
orig = len(k)*4 + len(v)*4
comp = (len(k)+len(v))*prec//8
saving = 1 - comp/orig # 94% saving target -80% PASS
return comp, orig, saving
Explanation: Implements IKNN-Rl1-A1 tri-tier ke... | agentic | coding | hard |
coding | Coding Hard 1214: Implement NoeSA-24 pack 13 values into 60-bit with base-24 encoding, correct case NoeSA — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def NoeSA_Pack_13x24(values): # NoeSA-24 4.58-bit
assert len(values)==13
packed=0
for i,v in enumerate(values): # v in 0..23
packed += v * (24**i)
# 24^13 ~ 3e17 fits in 60-bit (1.15e18)
return packed.to_bytes(8,'little') # 60-bit in 8 bytes, 4.615 bits/param
Explanation: Implement... | agentic | coding | hard |
coding | Coding Hard 4: Implement Ntarra-DnA phase rotator with correct case Ntarra-DnA, Direction NEG/ZERO/POS, Phase PHI0/1/2 shifts 0/2/4 — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def Ntarra_DnA_Compute(act, dir, phase): # Ntarra-DnA 3.17-bit
shift = {0:0, 1:2, 2:4}[phase] # PHI0 0, PHI1 2, PHI2 4
shifted = act << shift
if dir == -1: return -shifted # NEG
elif dir == 0: return 0 # ZERO
else: return shifted # POS
# Pack 2x9: 9*9=81 combos in 7-bit, 47 codes lefto... | agentic | coding | hard |
coding | Coding Hard 1658: Implement PG-KVC Phase-Gated KV Cache 94% saving with correct case PG-KVC — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def PG_KVC_Push(k,v,entropy):
prec = 1 if entropy<0.5 else 2 # Phase-gated: low entropy 1-bit, high 2-bit
orig = len(k)*4 + len(v)*4
comp = (len(k)+len(v))*prec//8
saving = 1 - comp/orig # 94% saving target -80% PASS
return comp, orig, saving
Explanation: Implements IKNN-Rl1-A1 tri-tier ke... | agentic | coding | hard |
coding | Coding Hard 199: Implement NoeSA-24 pack 13 values into 60-bit with base-24 encoding, correct case NoeSA — repo deeprcurs/IKNN-Rl1-A1 — model IKNN-Rl1-A1 — file IKNN-Rl1-A1-150M.iknn at root beside iknn.cpp — tidy — correct case SatU1 NoeSA Ntarra-DnA MoE RHT PG-KVC PEP ADLP | Solution: def NoeSA_Pack_13x24(values): # NoeSA-24 4.58-bit
assert len(values)==13
packed=0
for i,v in enumerate(values): # v in 0..23
packed += v * (24**i)
# 24^13 ~ 3e17 fits in 60-bit (1.15e18)
return packed.to_bytes(8,'little') # 60-bit in 8 bytes, 4.615 bits/param
Explanation: Implement... | agentic | coding | hard |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.