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 56: 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 831: 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 122: 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 1183: 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 6: 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 1439: 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 1293: 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 861: 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 402: 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 1839: 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 1935: 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 1831: 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 1115: 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 688: 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 1209: 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 284: 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 1815: 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 1196: 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 223: 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 1065: 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 201: 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 1139: 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 1902: 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 547: 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 1676: 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 1112: 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 1822: 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 741: 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 1051: 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 22: 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 1805: 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 1216: 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 1038: 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 1848: 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 291: 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 1591: 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 1807: 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 1563: 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 562: 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 969: 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 854: 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 1580: 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 1788: 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 465: 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 1322: 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 691: 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 211: 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 814: 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 476: 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 459: 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 769: 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 1691: 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 765: 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 1856: 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 1981: 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 542: 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 1584: 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 829: 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 1526: 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 903: 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 228: 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 216: 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 66: 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 190: 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 1419: 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 349: 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 1829: 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 882: 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 420: 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 64: 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 322: 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 1022: 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 1379: 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 247: 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 1145: 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 1797: 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 1894: 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 1670: 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 1108: 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 919: 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 538: 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 417: 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 258: 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 1549: 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 608: 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 1498: 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 1879: 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 851: 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 1176: 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 1464: 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 163: 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 1246: 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 835: 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 1992: 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 428: 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 1162: 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 1924: 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 207: 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 1415: 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 492: 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 |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.